06:15Albert Olgaard · Tutorial
Higgsfield MCP + Claude = Insane Videos
A 6-minute install-and-demo: wire Higgsfield's official MCP into Claude, then generate thumbnails and a finished UGC ad without ever touching the Higgsfield UI.
May 3rdThe argument in one line.
Claude Code has collapsed the cost of doing software work to near-zero for a solo operator, which means the entire barrier to an AI services business is now marketing and sales, not technical skill.
Read if. Skip if.
READ IF YOU ARE…
- You want to start a freelance or agency business and plan to use AI to deliver the technical work.
- You have tried Claude Code but have never landed a paid client and need a full system from outreach to invoicing.
- You run any service business and want to systemize delivery with AI tools.
- You are considering a career pivot and want to go from zero to first paying client as fast as possible.
SKIP IF…
- You already run an established agency with a team and existing clients — this is a zero-to-one guide.
- You are building a SaaS or product business; this course is specifically about selling client services.
- You are looking for passive income; this is an active service-delivery model.
The full version, fast.
The course argues that Claude Code has made the technical part of running a software services business nearly free, shifting the real constraint to marketing and sales. Albert walks through the full stack: installing Claude Code, building an Upwork profile with AI assistance, setting up a cold email system (Instantly AI + Apollo + custom domains), closing clients on calls without mentioning price until the end, delivering work via an Explore-Plan-Build workflow in Claude, and pricing services as a $2,000 upfront satisfaction-guaranteed retainer plus $500/month. The final module gives a 5-step roadmap: land your first client free or cheap, sustain yourself at $2-3K/month, then charge more — and never quit your day job until you can afford to.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →Where the time goes.
00:00 – 01:50
01 · Introduction — the opportunity
Market framing: 84% of people have never used AI, 0.3% pay for it. One-person AI companies worth billions are already happening.
01:50 – 04:00
02 · Why learn AI in 2026
The excitement curve — most will plateau, a small group will capitalize. Positioning for the top 0.3%.
04:00 – 10:00
03 · Claude Code setup in VS Code
Installation walkthrough, CLAUDE.md setup, skills installation, GitHub integration.
10:00 – 19:50
04 · How to land clients — mindset and 3 service levels
Client-getting mindset, the 3 levels of AI service delivery, positioning against low-quality competition.
25:03 – 57:00
05 · Getting clients with Upwork
Full Upwork profile build with Claude, proposal strategy, Connects bidding, Loom video demos, getting first $15/hr job for the reviews.
57:00 – 1:18:00
06 · Getting clients with cold email
Instantly AI setup, custom domains, mailbox warmup, Apollo lead lists, AI-generated personalized sequences built inside Claude Code.
1:18:00 – 1:37:33
07 · Closing clients on calls
20/80 rule (80% listening), never mention price before the call, objection handling, Stripe and invoicing setup.
1:37:33 – 1:50:00
08 · Delivering work — websites with Claude
Live build of a real client landscape architecture website using Claude Code with Explore-Plan-Build workflow.
1:50:00 – 2:06:54
09 · Building automations with Trigger.dev
Email automation system: Claude builds the full backend (Next.js + Trigger.dev + Resend). Live debugging session included.
2:06:54 – 2:36:48
10 · Building full AI systems
Lead enrichment and AI email-response system. Multi-agent architecture with Claude Code as orchestrator.
2:36:48 – 2:59:08
11 · How to build with no experience
Explore-Plan-Build methodology in detail. How to scope projects, get architecture right, and avoid rebuilds.
2:59:08 – 3:29:38
12 · How to price AI services
Three pricing tiers, satisfaction guarantee structure ($2K upfront + $500/month), never compete on hours.
3:29:38 – 3:38:26
13 · Roadmap — what to do now
5-step roadmap from zero to scaled agency. Build from 5 AM to 9 AM alongside a day job.
Lines worth screenshotting.
- The real bottleneck in an AI services business is marketing and sales — not coding ability. Spend four hours a day outreaching before worrying about what to build.
- Upwork jobs at $15/hr are wins early on because the profile rating and review history are the actual asset being built.
- Cold email infrastructure is itself a Claude Code project: custom domain, separate mailboxes, warmup sequence, Apollo lead list, all built and configured without leaving the terminal.
- Never mention price before the sales call. If a client asks beforehand, delay — premature price discussion kills deals before they start.
- Explore first, then plan, then build: one of the most common Claude Code mistakes is starting to build before gathering enough context about what to build.
- A $2,000 upfront satisfaction-guaranteed fee plus $500/month recurring beats hourly pricing because it removes risk for the client and creates predictable income for you.
- Nine out of ten businesses fail at step one or two — before the founder reaches the point of self-sustainability. Getting to $2-3K/month is the survival threshold, not the goal.
- 84% of people have never used AI and only 0.3% pay for it — the opportunity window is still wide open.
- Do not over-automate a bottleneck. Sometimes hiring one salesperson is faster and more effective than spending weeks building an AI solution.
- Scaling a business is just an infinite loop: find the bottleneck, fix it, find the next one. There is no final stage.
- AI does not make client results irrelevant — clients do not care that AI built their website, they care whether it converts and saves them money.
- Building a Loom walkthrough video for every Upwork proposal dramatically increases response rate by proving you are a real human, not a template-spammer.
- It is a myth that you must quit your job to start a business. Four hours before work and weekend time is enough to land the first clients.
The business bottleneck is marketing, not code.
WHAT TO LEARN
Claude Code removes the technical barrier to building software entirely — which means the one thing that still requires human effort is convincing people to pay for it.
01Introduction — the opportunity
- 84% of people have never used AI at all — the market is not saturated, it is barely started.
- A small percentage of operators will capitalize on the opportunity; the rest will plateau at curiosity.
05Getting clients with Upwork
- Getting your first Upwork reviews at $15/hr matters more than the money — the profile rating is what unlocks access to better-paying jobs.
- A Loom video walkthrough in every proposal is the fastest signal to a client that you are a real human, not a template spammer.
06Getting clients with cold email
- Cold email that works requires proper infrastructure first: a dedicated domain, separate mailboxes, a warmup period, and a lead list before a single email goes out.
- The entire cold email system can be built without leaving Claude Code — domain setup, mailbox config, lead list import, sequence automation.
07Closing clients on calls
- Pricing with a satisfaction guarantee moves the risk from the buyer to the seller, which is why clients say yes to $2,000 upfront when they would hesitate at $500/hour.
- Client objections about price usually resolve when you shift from cost to outcome: the question is not what the service costs but what it saves.
11How to build with no experience
- Gathering context before touching the keyboard prevents the most expensive kind of rework.
- The Explore-Plan-Build sequence works because architecture decisions made in plain language are reversible; decisions made in code are not.
12How to price AI services
- Over-automating a bottleneck is a real failure mode — sometimes the answer is hiring a person, not building more AI.
- Three pricing tiers give clients a range that anchors the mid-tier as the obvious choice.
13Roadmap — what to do now
- The survival threshold for a service business is not profit — it is covering rent, because that is the point where a founder stops having a reason to quit.
- Building alongside a day job is the lower-risk path to the first client.
- Scaling means finding the current bottleneck, removing it, and immediately finding the next one.
Terms worth knowing.
- Claude Code
- Anthropic's terminal-based AI coding agent that can read, write, and execute code across an entire codebase with a single natural-language prompt.
- CLAUDE.md
- A project-level instruction file Claude Code reads at the start of every session to maintain persistent context, coding standards, and workflow rules.
- Instantly AI
- A cold email platform that manages multiple mailboxes, email warmup, sequence automation, and deliverability — used here to send personalized outreach at scale.
- Explore-Plan-Build
- A three-phase Claude Code workflow: first gather all context without writing code, then produce a written architecture plan, then build — reducing costly rebuilds from misunderstood requirements.
- Satisfaction guarantee
- A pricing structure where the client pays upfront but receives a full refund if unsatisfied, removing their perceived risk and making premium pricing easier to close.
- Trigger.dev
- An open-source background job and automation platform used here as the backend engine for scheduled tasks, email sequences, and webhook-triggered workflows.
- Warmup (email)
- An automated process that gradually increases a new email domain's sending volume and generates positive engagement signals to avoid spam filters before cold outreach begins.
Things they pointed at.
Lines you could clip.
3:31:24
“If you can just sustain yourself — just pay yourself a small salary every single month that covers your rent — that means you can go all in and you will never give up.”
3:29:00
“People massively underestimate how much time they should actually be spending on marketing and selling.”
3:29:27
“Don't focus on it being AI. Focus on the outcome. Will they save $10,000 a month? Yes or no? They don't care if it's AI doing it.”
3:35:36
“It took me four months of straight grind outreaching every single day before I landed my first client that paid me $400.”
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
00:00If I were me two right now and graduating college, would feel like the luckiest kid in all of history. Why?
00:08Wake up. AI is here. And if you assume any rate of improvement over any reasonable time period, learning how to use AI should become your number one priority.
00:18I think it is possible now to start a company that is a one person company that will go on to be worth more than a billion dollars. The New York Times just ran a story about a guy who built a billion dollar company with AI. We all knew it would happen.
00:30One person creating a $1,000,000,000 company. You have access to tools that can let you do what used to take teams of hundreds.
00:37And you just have to learn how to use these tools and come up with a great idea. You've probably heard it about a thousand times.
00:45Oh, AI is going to make you rich. Oh, I just made Claude Code wipe my ass. Or I just replaced my entire family with Claude Code.
00:53There's a bunch of those videos out there, and they have taken the Internet by storm. And I know what you're thinking. Oh, another one of these drop shipping, NFT, get rich quick schemes.
01:02But that is not what this course is about. And if you're looking for the next get rich quick thing, then you might as well click off this video, because this video is not for you. And I know that, because I actually thought the exact same when I got into this space a little over three years ago.
01:16If you don't know me, my name is Albert, and I am by no means the smartest. But I still somehow managed to build two AI companies that combined have done over a million dollars, which is weird, because I don't have a university degree in computer science.
01:32I actually never went to university. So I'm kinda on my, what, like seventh gap year or something. The only reason that this was possible was because of AI, and I give full credit to Tradjibouti and Claude, like, 100%.
01:45I couldn't have done that without these tools. But let me show you how this was even possible. You might have seen this graph going around.
01:52Each dot on this graph represents 3,200,000 people.
01:57So in total, we have 2,500 dots, which amounts to 8,100,000,000 humans.
02:02And what this shows is how many people in 2026 that have even used AI. And what you'll find is that 84% of people, 84%, which is 6,800,000,000 people, have never used AI in their life.
02:18They've never even given chattypity a regular prompt, which is hard to believe because if you're like me, you see AI everywhere. Every time you open your phone, you hear AI news, but that is because your algorithm knows that you're interested in that stuff.
02:3116% of people have used free chatbots. So that is free ChatGPT or free Clawd. That is 1,300,000,000 people.
02:39But if you are just a little into AI, you know that the free models, you almost can't do anything with those. And then if we zoom in, we have the people that pays $20 a month for AI. That is 25,000,000 people, represented by these small yellow squares right here.
02:550.3% of the population. And my guess is that's probably where you are right now.
03:00Or maybe you are part of the 0.04%, which is the max users, so the people using the most powerful models like ClaudeCode and Codex.
03:10That is only 3,600,000 people or 0.04 The reason that I'm showing you this is that I want to show you how far behind the world still is when it comes to AI.
03:22Many say that AI is going to be the greatest productivity boost that the world has ever seen. And even then, after AI has existed for a couple of years now, 84% of people has still never even used it.
03:34I hope you see how this creates a massive opportunity, because this allows us, the 0.04%, to provide services to the rest of the 8,000,000,000 people that don't know how to use AI effectively yet.
03:49To these people, your AI services is going to seem like magic. And the reason that I'm dropping this full course right now is that the world is finally starting to understand AI. When I started providing AI services three years ago, the only real markets that we could go after was The US and Canada, The UK and Australia.
04:09Every other market simply wasn't ready yet. But because of the mass adoption that's happening right now in AI, all of a sudden, most other countries are starting to pick up as well.
04:19And this opens up markets like South America. Europe is one of the biggest up and coming markets in AI right now. Countries in Asia like India is also starting to pick up on AI, which means that companies are ready and they know that they need AI implemented into their business.
04:34And it's not going to take long before the entire world is at the same wavelength. You might be thinking, well, AI companies already implementing AI. I might be too late.
04:43But I showed this to you to make you understand that it's not too late yet. And by the end of this course, when you stick around, you will 100% be in the 0.04% of AI users that will actually be able to capitalize on AI.
04:57But don't get too excited yet, because if you've ever tried to start a business before, you might know or definitely have filled the excitement curve before. When you've just heard about a new business model or you want to start something or a new project or a new business, you're gonna be extremely excited. Your excitement is going to be through the roof.
05:16You're gonna be thinking about all of the good things. So for example, building an AI business, this would be, oh, I can automate my entire product delivery. Oh, I can get unlimited clients.
05:24Everyone's gonna be interested in my services. I'm going to be rich. That's the thoughts that are going through your head when you're just starting out.
05:30And the reason that I told you right at the start that you shouldn't expect this to be a get rich thing, is because your excitement will then drop. This is 100%. All of a sudden, you realize all of the problems that are in the business, that it's actually not easy to get your first client, that you have to get really, really good at what you do, you need to watch videos like this, you need to learn before you can actually start selling these services, and that you probably won't land your first client a couple of weeks after starting.
05:56That is the reality of every single business model. There are always issues, there are always things, and things are never as easy as they seem. So you reach a low of excitement.
06:07And this right here is where ninety five percent of people give up. They don't wanna go through the pain of figuring out, okay, how do I actually land clients?
06:17They don't wanna spend all of their time learning the skills that it actually takes in order to provide good AI services. So they do one of two things. Either they say, okay, this AI thing, scam, onto the next, they try drop shipping or NFTs or crypto or some other thing, or they simply just stop business altogether and they go back and just focus on the regular thing that they're doing, the nine to five or whatever.
06:38But a small percent of people, 5% or so, are going to work through the excitement drop, and it's not a steep curve.
06:48It takes time, and all of a sudden, they fix a couple of issues, and then they land their first client, so it might go up like this. But then they realize, okay, it's actually not as easy as I thought provided the service, so it stalls again. And then they maybe figure out, okay, how do I actually provide the service?
07:02So it goes up a bit again. And then their first client drops, and all of a sudden they are back with no clients. But as they keep working, this curve starts going up, and all of a sudden they crack how it works.
07:12And after long enough time, they're gonna be very excited. They're gonna know, okay, this is how I actually scale this business. This is how I do it.
07:19But getting to here takes an immense amount of time and work. I'm setting these expectations right now, because if you're not ready to do this, then it doesn't make sense for you to spend hours watching this, and you can just click off the video.
07:31But if you're actually committed to do this, if you are the top 5% that does give up, then you are the person that I made this video for. The worst thing you can do is to get stuck in this loop of trying a new business model, getting very excited, getting the drop in excitement when you actually try it, and say, okay, this doesn't work, and go to the next one.
07:50Because all of a sudden, you get stuck in a loop of just trying stuff, but never actually committing and going all in and actually staying with it, which is what it takes to get success in anything in life. I hope you're ready to put in the work, and if you are, then good, Listen up. These are the things that we're gonna go through in this full course.
08:06Don't worry if you are a beginner in all of this Claude stuff. We're gonna start all the way from scratch designing and building our AI operating system from nothing. So if you're just starting out, this is the perfect video for you.
08:18Most other guides on YouTube show you how to build the tech, but they actually never show you how to land your first client. So that's actually the first thing that we're gonna do in this course. We're gonna focus on getting clients, getting clients, getting clients, because that is really what matters and really what moves the needle when you're just starting your AI business.
08:35The best way to land clients when you're just a beginner is to do something where you can reach out to a lot of people, but do it very cheaply. So I'm gonna show you how you set up cold email campaigns, and I'm gonna show you how you sign up to Upwork to get your first couple of freelance clients. Then I'm going to show you how you take those interested leads that we got from cold email and Upwork, and how you book them in on meetings, and how you close them.
08:57And I'm even going to show you a secret trick for how we can get Claude to listen in on our meetings, so it gets the context for exactly what we need to build, and can even start the building process before we even end the meeting. So stay around for that, because that's absolutely going to blow your mind. Then I'm going to show you how you actually do the service delivery.
09:14I'm going to show you how to build websites, automations, mini apps, and the process that I use for basically being able to build anything with no university degree where Claude Code does the heavy lifting for us. And then we're gonna go over a very important step which a lot of AI business owners, they miss.
09:30We're gonna take the cool things that we have built, and we're gonna show them to the world. We're gonna post this on LinkedIn. We're gonna post it to YouTube.
09:37And this will have a compounding effect where people all of a sudden see the proof of the stuff that you have built, which is going to get a ball rolling to get you even more clients. And at the end, I'm going to show you how not only to automate your clients' businesses, but also how you automate your own business so you truly achieve the one person AI business using Cloud Code.
09:55This is probably going to be the most valuable thing that I've ever released, and I'm releasing it for free. So I'm very excited. Let's get it.
10:01The beauty of plot code is that we can manage our entire business with it. So let's set that up, shall we? If you have already set up plot code inside of Visual Studio Code, then you can just skip to the next chapter of the video, but let's set up plot code the right way.
10:17The first thing you want to do is to open a browser, then you want to search for Visual Studio Code, click the top link, and then download it for your computer, either Mac or Windows.
10:29Visual Studio Code is what's called a code editor, but don't worry. We're basically not gonna create any code ourselves. Cloud Code is going to do absolutely everything for us.
10:39We then click on the installer, then we drag it over to other applications. We let it install, and now we can open up Visual Studio Code. You're gonna land on this page right here.
10:50Before we do anything inside of here, let's create a folder that we can work within. Excuse the mess on my desktop.
10:58I promise you we will get that cleaned up. I'm going to create a new folder, and I'm going to call it shiny, which is the name of our agency. This is going to be where our AI operating system lives inside of this folder.
11:11So call it something that you can remember. Then head back under Visual Studio Code and click this open button right here, then go to wherever you saved the folder. I saved it on my desktop, so I'm going to open this shiny folder.
11:22Click open, and there we go. We'll then click yes, I trust the authors, and we can close down this welcome message.
11:30Now we are inside of our code editor, and we are ready to start setting up plot code. What you want to do is that you want to open the terminal, so I'm going to right click inside of this space and click new terminal, and this is going to open up this terminal view right here.
11:44Don't worry. This is going to look a bit technical, but you can just watch exactly what I do one to one, and it's really not going to be that hard. You don't need to be that technical for this.
11:52Then I'm going to open a new tab. I'm going to search for ClaudeCode install.
11:58Click the first one that's called quick start ClaudeCode docs, and then we have free commands right here for installing it on Mac or on Windows. So I'm going to choose the Mac install that fits with my computer, then I'm going to paste in this command, and that's going to install Claude code on our computer. And there we go.
12:14It now says setting up Claude code, installing Claude code native build latest. It then says Claude code successfully installed. It shows the version.
12:23It shows where it was installed. So now we can write clear, that's going to clear our terminal. When we now write Claude, you can see it's going to say welcome to Claude code, and we can choose what style we want.
12:34The way that we navigate is with the arrow keys up and down, I'm and definitely going to want dark mode, so I'm just going to hit enter to dark mode, and now it gives us three options, either a cloud account with a subscription or using an API key through the console.
12:49I'll definitely recommend you use cloud with a subscription, either pro or max. I'm personally on the max plan right now, because I need that bit more usage.
12:58But if you're just starting out, then I recommend that you start on the pro plan. So just go to claude.ai, sign up to the pro plan, and you can see that gives you access to Claude code directly in your code base.
13:08When you've signed up with an account and you have the pro plan, you now go to Claude account with the subscription and you hit enter. And this is going to ask you to authorize with your Claude account. You can see it says Claude code would like to connect to your Claude chat account.
13:20So we just click authorize. Now it says build something great. How fitting.
13:25And you can see it says login successful. Press enter to continue. So that's what we're gonna do.
13:30We're gonna click enter, and then we're gonna hit enter again. And then Claud is asking us if we want to use recommended settings. I'm going to hit enter, and there we go.
13:37Now we are inside of Claude Code. And then we can write just Claude like this in our terminal after we've installed it. That's going to open up Claude Code on our computer.
13:46It says, quick safety check. Is this a project you created or want your trust? So I'm going to just hit enter.
13:51The way that you navigate inside of these options is that you can use your arrow keys up and down to go up and down. I'm going to click yes. I trust the author, and then hit enter.
14:00And there we go. Now we are inside our Claude code view, and we have successfully installed Claude code. The next thing we're going to need is some Claude skills.
14:08Inside of our completely free group, you can grab my Claude skills. I'll leave a link right below this video, and then you can find those on the classroom, and then go into AI learning hub. Scroll down until you find Claude skills right here, and this gives us this drive folder with all of our Claude skills.
14:25So I'm going to click Claude skills right here. I'm going to click download, and now you can see it's downloading all of these skills in a folder. Claude skills are basically portable skills that you can give to Claude code.
14:38So inside of these skills, we have some documents that explains how to use something. It's basically just a list of instructions. But by having the right skills and having the right instructions, you can make Claude code even smarter than it already is.
14:50So if we go back into Claude code and we then double click on the zip file that we just created, this gives us our Claude skills. I'm now going to write exactly this inside of my folder Claude skills two.
15:04This was the folder name. Right? The folder name is Claude skills two.
15:08In my downloads, because it's inside of our downloads folder on our computer, I want you to install these skills in this project folder, and then I'm going to call the folder shiny, where we are right now.
15:26Then we're going to write exactly this. I want you to research the documentation for installing Claude skills in a project.
15:38Please research that and install these skills in this current folder.
15:44Hit enter, and now what we're basically forcing Claude to do is that we're forcing it to go out and research first before we have it do anything, and that is also how you usually work with Claude. Sometimes it doesn't have all of the context that it needs.
15:58So by telling it to go out and find that context, we're gonna get much better results. That took about well, exactly forty three seconds, And if we go inside of this dot cloud folder that was now created on the left hand side and inside of the skills folder, we can now see all of the skills that we have inside of this folder.
16:18What you want to do now is that you want to write clear inside of Cloud Code, then we can hit control c, which is basically just going to start Cloud Code, and then we're gonna force it to restart.
16:30After installing skills, you need to restart the session for the skills to actually be enabled. If I now write slash and then front end design, you can now see we have a front end design skill right here that creates distinctive production grade front end interfaces with high design quality.
16:48Use this skill when the user were asked to build web components, pages, or applications, blah blah blah. So now we have a front end design skill, and we have all of these other skills as well, like composure, or customer support, or research, or scalability, or security, etcetera.
17:03All of these skills are now inside of this dot claud folder inside of our workspace. The next thing you want to do is that you want to be able to save this to the cloud. Right now, all of this information, all of these skills are living on our computer, but we want to have a copy of this in the cloud as well.
17:19That way, if I smashed my computer and for some reason it was destroyed, we would still be able to recover our workspace, which if you have been working in a workspace for, let's say, a couple of weeks, you definitely don't want to lose that workspace and all of those skills. Let me show you how to do that. Go back into your browser, then search for github.com and create a user if you don't have one.
17:42After you've created a user, you're gonna land on this page right here. Then you wanna click new. I'm just going to call this shiny workspace.
17:50I'm going to choose myself as the owner, and then I'm going to make it not public, because we don't want everyone to be able to see our workspace, that is a high value asset. So I'm going to choose private and then click create repository.
18:04This will basically create a folder inside of GitHub, and with GitHub, we can basically publish our code that allows us to save a version of it in the cloud. The next thing you want to do is that you wanna copy this URL, paste this in, and say, please push this current folder we are in to this repo. And now what Cloud is going to do is that it's going to take these Cloud skills, which is everything we have inside of the repo yet, and it's going to take all of that and push it into the cloud, so again, if something happens to our computer, if we lost access to it, we don't lose our workspace.
18:38Very important. My GitHub account is already connected to my computer. When you're doing this for the first time, it's going to pop up and ask you to authorize with your GitHub account.
18:47So just make sure to do that when you're doing this for yourself. You can now see it says the push was blocked by an auto mode classifier. That's because we have Claude in auto mode right now.
18:56What I'll do is that I'm going to set it to just like default mode, right, try again, and then we just have to accept that it actually takes this workspace and pushes it into the cloud. Do we want to proceed?
19:06I'm going to go down and say yes and don't ask again. Hit enter and this will now be pushed into the cloud. Push successfully to alpha shiny shiny workspace main branch 52 files.
19:16If we go back into GitHub, you can see this is what a blank workspace looks like. If you now hit a refresh, you'll be able to see our Claude skills right here. This is the exact same files that are inside of this folder right here.
19:27Now they're just also in the cloud. Great. Congratulations on having all of this set up.
19:32Now comes probably the most valuable part of this entire course, and it's really what no one else is showing you on YouTube. Everyone is talking about the tech, but the hard part for almost everyone is, okay, now I know how to build stuff, how do I actually sell it, how do I go out and actually land clients?
19:51A wise man once said that you have to get good before you can get rich. What do I mean with that?
19:58What you've probably seen on Instagram and all the other socials as well is people saying, go out, learn how to build websites, and sell your first one for $10,000. And while it is true that you can sell websites for this much, is it realistic that you're going to go out and close a 10 k deal as your first one?
20:17Well, probably not. And that is also what I mentioned earlier when we talked about expectations. It's probably not going to go that way.
20:23Why is that? Well, there's a couple of reasons. The first one is that you have absolutely no trust.
20:29And how do you build trust? Well, you need testimonials and you need previous experience.
20:34So when you hop on a call with someone that you actually know what you're talking about, that you're actually certain in yourself that, okay, you can help this person get better results. In this course, I'll show you how to get both of those things.
20:46And what I'll also show you is the next part that you're lacking, which is skills. For some reason, these days, people think that you can just go out and learn something in five minutes and all of a sudden charge thousands of dollars for it. I'm sorry, but that is not reality.
21:00You've gotta understand that what people are paying you for is the time and effort you've put in to gather skills, skills that this person that are paying you doesn't have themselves. If it was easy, the thing that you were doing, the thing that you had learned, then no one would pay you for it.
21:16So you have to do exactly what it is that you're doing right now when watching this video. You have to go out and you have to learn, you have to build these skills in order to get paid.
21:24And that's exactly what I mean when I say you have to get good before get rich. And of course, that's also the entire point of this video, is to get you to build those skills. When I started my AI business, it didn't take a week, it didn't take a month, it didn't take two months, it took four months before I closed my first paying client for $400.
21:45And guess what I did before that? I did not charge $10,000 for a website.
21:49I actually did free work in order to actually build trust, to gain experience, and to gain testimonials. And by doing the free work, I also gained the skills that I could later monetize and actually charge for. So if you're thinking right now, okay, I'm gonna go out, I'm gonna learn this, and then the next week I'm gonna sell a 10 k website, leave this video.
22:07This video is not for you. Understand that you have to build the skills before you can charge anything, and understand that this takes months. The way that I think about it is that there's kinda like three different levels to what you can be building when you're just starting out.
22:20The first level that I recommend that you start with is AI generated websites, Specifically with Cloud Code, it's extremely good at building websites. And let me show you something very, very interesting. If I head over to Claude and I ask how many small businesses still don't have a website?
22:36While I then drink my coffee, Claude is now going to research the web and find out the number. And now we can see that roughly 27 to 30% of businesses in 2026 have no website.
22:49So roughly one in three small businesses still don't have a website. So if you learn that skill of building websites, all of a sudden you can provide services to that 30%, and that 30% does not have a website. You can also build better websites and sell it to the rest of the 70% if you provide something that's better than what they currently have.
23:07So I can write, okay, so if it's 30%, how many businesses in that in, let's say, The US? Just to put a number for how many potential businesses that you could actually be working with. I'll take another sip of the coffee while Claude is researching.
23:19So in The US there are 36,000,000 small businesses, and 30% of that is then 10,000,000 businesses without a website.
23:30And that is, in my opinion, the best place to start. Now you have millions of businesses that you could service with your skills. They don't know how to build websites with something like Cloud Code.
23:39They are not going to learn it. They are focused on their craft. So when you in this course is going to learn how to build websites, all of a sudden you can service all of those businesses and provide a valuable service.
23:50That is level one out of three. The next thing we're gonna take a look at is the step up. Now you're not only building websites, you're also automating their business with automations and agents.
24:00This requires some more knowledge into integrations and APIs, but don't worry, you're gonna be learning how to do that as well. And I would say the final level is building out full AI systems for businesses that not only integrate into what they already have, but also where you are connecting all of the automations that you have built from level two, and then maybe creating a dashboard or somehow building it into a full AI system.
24:22But of course, we don't start at level three. We start at level one, websites. And stay around because later in the video, I'm going to show you exactly how you build those out in just a couple of minutes.
24:31I wanted to go over these three levels so you know what to actually sell because in the next part of the course, we're gonna go into, okay, how do I actually land clients. One of the biggest mistakes I see beginners make is that they focus all of their time on building and not enough time actually outreaching and talking to clients.
24:45We're gonna flip that on its head. So before you even know how to build anything, we're gonna start with the outreach, because that means we can have the outreach running while we are then building stuff. So let's get into probably the most valuable part of the course, how to actually get clients.
24:59Let's get into it. To do our outreach and actually build up the marketing system, we will of course also be using Cloud Code. We're basically gonna be using Cloud Code throughout the entire video, and Claude Code will probably do 90 to 95% of the work in this business model that we are building up right now.
25:18I'll be showing a bunch of cool stuff that you can do inside of Claude Code. If you're looking for a full guide to Claude Code, then I'll definitely recommend that you check out the Claude Code masterclass inside of our closed community, the 1% in AI. This is also where you can get help directly from me.
25:33You have a full masterclass in absolutely everything inside of Claude Code that takes you from a complete beginner to a master in AI coding models. This community is also where you can get help from me.
25:43I answer every single post inside of this community. If I click on these icons right here and then go to new, then you can see that I answer absolutely every single post in here. This right here is my face, and you can see I have commented and answered every single post inside of this community.
25:58I can actually see a post was just made one minute ago and just now, so let me answer those real quick. That was one. Let me answer this as well.
26:04There we go. And the best thing about this community is that right now we are running the thirty day AI challenge. What this is is basically thirty days of videos, and when you finish those videos and stay consistent for thirty days, then you actually get your first month completely refunded.
26:18So if you're consistent, you can start the thirty day AI challenge that takes you from a complete beginner to setting up an agency in thirty days, and then you can get to try all of this completely for free, because when you finish, then you get all of your money back. If that sounds interesting, then watch this video right here.
26:32This explains everything. I'll leave a link right below this video. Alright.
26:36Let's continue with the course. Let me show you how you can actually get clients. Alright.
26:39Let's get into it. This is how I'd recommend that you get your first couple of clients, and how you build trust, and how you build testimonials. The first ones are always the hardest.
26:49It's like rolling a big stone. It's going to require a lot of force when you're just starting out, but as soon as you get the ball rolling, then you can use the momentum, and it will start rolling faster. It's the exact same when it comes to landing clients and building the business.
27:01The first couple of months, the first couple of clients are an absolute pain. But let me show you how I recommend that you get your first couple of clients. To do this, we are going to use Upwork.
27:11If you don't know what Upwork is, Upwork is basically a freelancing platform that allows you to post yourself as a freelancer to say, okay, I can help with these things, and then businesses will also go to the platform and say, I need help with these things. So this platform basically connects freelancers with actual businesses.
27:27The good thing about Upwork is that you can find leads directly that need your exact services. So if you're building websites, then you can find people that need something improved on their website. If you're doing AI automation system, or GoHighLevel, or NNN, you can find people that are specifically looking for that service.
27:43This doesn't mean that Upwork is just like free clients and that you're gonna get a client as soon as you sign up to Upwork. There are really levels in the game when it comes to Upwork, how you build your profile, how you send your applications to post, but I'll cover all of that in this module. So stay around because there's a high chance that this I'm about to tell you right now will get you your first couple of clients.
28:01First thing you want to do, go to upwork.com, click the top right corner, and click sign up, and then sign up as a freelancer. I'm then going to use my Google account to sign up.
28:09There we go. I write in my first name. I write in my last name, and then I choose the country, which is going to be Denmark.
28:15Then I'm going to check this box right here and click create my account, and there we go. Congratulations. You have been signed up.
28:21Now we need to create our profile, so I'm gonna click get started. I recommend that you write, I have some experience right here. Click next.
28:28What's your biggest goal for freelancing? To earn my main income. Click next.
28:32Then check both these on, and check this right here, and click next create a profile. Here you can do a couple of different things. If you have a LinkedIn that already now shows your experience, then you can connect your LinkedIn already.
28:43So that's what I'm going to do. I'm going to click import from LinkedIn, then I will go to LinkedIn, go to my profile, click these three dots, click save to PDF.
28:51That basically creates a PDF of our LinkedIn profile. Then I can click upload, disable LinkedIn PDF, choose this, and this will now upload our PDF, so Upwork gets all of our information from our LinkedIn.
29:02Don't worry if you are not set up on LinkedIn yet. If not, just click fill out manually right here and go through the process. But for now, I'm gonna click continue editing your profile.
29:10Now you're choosing what kind of work we're in. I'm gonna go inside of IT and network, and then choose database management and administration, and CRM software right here.
29:19This is what best fits like a classic automation agency. Now we click next, add your skills. Now we need to add a couple of skills.
29:26We could add CRM software already here. We could add our platform that we're working, like, it in, for example. We could add Claude.
29:33We could also add things like marketing. If you're doing Facebook ads, for example, you could choose Facebook advertising. You could choose things like Google ads.
29:41You could choose things like building websites. You can always change this later, but for now it's good to show, okay, these are the things that you are usually working with. You could also choose AI agent development, and sometimes it's also good to choose some coding languages like TypeScript, Python, etcetera, that can sometimes help you get some jobs.
29:59You could also choose things like Sabia, for example. This can help out as well. So basically, like, automation platforms.
30:05You can also choose make.com, and there we go. Now we have added a bunch of skills that shows what we can offer.
30:11We can then click next and add a profile title. In our professional role, we can write something like this. AI agents in Sabia go high level make in it in AI consulting and implementation.
30:21Then click next, add your experience. You can see it has found that I'm cofounder of Shiny AI Solutions. You can click edit right here, and then we can add more information, like location, for example, is in Denmark.
30:34Click I'm currently working in this role. Save right here. Now we can add our education, then you can add it right here.
30:40For now, we're not going to do that. Now we're going to click save for now and then add languages. I definitely recommend that four languages that you choose native or bilingual right here.
30:48You then choose all of the languages that you know. I'm low Danish, for example. I know a bit of German, and I know a bit of Swedish as well.
30:56I can then write next, write an overview. We're gonna come back to this later. I actually already like the this AI version right here.
31:03As a cofounder with four years of experience in tech landscape. We can change this in AI automation.
31:09AI agents, I specialize in developing solutions to drive efficiency and productivity and solve real world problems. And then we end it off with, if you need someone who is an expert in automation, I am your guy. This one I probably want to delete.
31:20We don't want too big of a description. I like this one right here. It's also a good idea to write in the tools that you use.
31:26So I specialize in developing. I'll change that to building. I specialize in building solutions, specialize in building helpful solutions that also made it improve real world problems.
31:36I use tools such as go high level, savior, make, n n like this.
31:45If you need someone who's an expert in automation, I'm your guy. We can even also write plot code. I use tools such as go high level savior make n n n and code slash codex.
31:56If you need someone who's an expert in automation, I am your guy. There we go. Now we go next, set your rate.
32:02I'd recommend not charging too much in the start. What you'll see the experts on Upwork charge is something around like $50 an hour. For now you just want to go out and land your first couple of jobs, because those are going to show the success rate is going to show on your profile.
32:16So I recommend that you set this to something like $15 an hour. The service fee, you cannot change that, which means you will get 13 and a half dollars an hour. Now we click next, add your photo and location.
32:25For the profile photo, you want to add a professional photo of yourself, like this one right here. So I'm going to click upload photo, and then upload this photo of ourselves right here.
32:35Click attach photo, then you wanna write in your birthday, your country, fill out all of this information, and then review our profile. There we go.
32:42AI agents and say, we're gonna high level make in an AI consulting implementation. Set our hourly rate, set our skills, set our work history, and then submit the profile. Now Upwork is asking us to buy some connects.
32:54I'll definitely recommend that we do do that at one point. For now, I'm gonna click browse without bidding, and we're gonna go to this page right here, and you can see we are 60% complete now.
33:03We still want to add a bunch of stuff to our Upwork profile. It's important that you fill all of this information out as much as possible, because this helps build trust with Upwork, which will help you land more jobs. So inside of profile settings, I'm going to set myself as an expert in the space.
33:18We can link our accounts, like if you want to link our GitHub, we could do that. Going to click authorize. There we go.
33:24And I recommend that you complete this working style right here, which is basically just a small quiz. I click start assessment. Just fill out these questions with what you believe.
33:32Stay organized. I like to communicate my schedule. When scheduling my week, I usually plan detailed steps, block time.
33:39When I deliver work to a new client for the first time, I usually double check every part before setting it out. When timelines are tight, I prefer to focus on the key tasks that ensures quality.
33:49Before submitting work, I usually review the brief. I confirm it meets the client's intent. When collaborating with other freelancers, I usually cross check shared work for consistency.
33:57When project deadlines change suddenly, I talk with clients to confirm what's more most important to finish first. I feel most productive when I'm solutions.
34:06When I'm faced with something new, I usually start by brainstorming. When I need to make a decision, I simply make the call and take responsibility. Landing the golden key so I can make this.
34:16Before I start to work with a new client, I prefer to identify potential chances early and plan away to handle. There we go. Clear communicator is now visible on our profile.
34:24All of these things basically just helps you build trust with both Upwork and people looking at your profile. All of these are really important. The next thing you want to do is verify your identity.
34:32You do that by clicking this button right next to your name. Click on portfolio, and then you can add a project title,
34:39like a solar sales AI agent that I've done, for example. Reactivation of old solar leads. This could literally be anything.
34:46Also, if you've had a previous job, write in what you did and what you accomplished. Then I have this picture right here from our website, so let me take a screenshot of that. Now we can add images.
34:55I also have a picture right here of a completed booked calendar that I worked on. I can write book Google Calendar booked by an AI agent. I can add more images like this one.
35:03Future appointments booked in a month. Next, preview. And there we have it.
35:07Solar sales agent developer. Reactivation of all solar leads. Next, thumbnail.
35:12Let's choose this as the thumbnail, because that's probably the most catching picture, and there we go. Our portfolio is published. Now you can see we are 75% done.
35:21We can also create a video introduction if you wanted to. We could add certifications. If you have any certifications that you've done that are slightly similar to what it is that you're offering now, make sure to show that on your Upwork profile as well.
35:32Including an employment history is also smart. So if you had a previous job, include that too. Basically, fill out as much information on this Upwork profile that you can.
35:40One of the things that can make or break your Opioid profile as well is the location right here. So you want to make sure to go inside of profile settings, go inside of contact info, and then change this to a location that people know. Right now, I've set it to a city called Vaixper in Denmark, but I'm going to better results if I set it to something that people know, like Copenhagen.
35:59Then we can click update, and now you can see it says Copenhagen, Denmark, and people know what Copenhagen is, which one, sparks new conversations, and two, builds credibility, which is what all of this is about.
36:10But we can improve this Upwork profile a whole lot more, and of course, we're going to use Cloud for that. Let's get into it. Before we continue, let's take a look at the people that are absolutely crushing it on Upwork.
36:20Like this guy, for example, v per d, Total earnings, a million dollars on Upwork,
36:26two thousand hours worked. If we take a look at his profile, we can see it says Istanbul, Turkey, you know, a capital big city that people know.
36:3398% job success and top rated. And the great thing about these Upwork profiles is that they are 100% public, so we can just go in and steal what we want.
36:43You can see he's offering a consultation first, and he's showing some work history with his reviews. You can see that he has a bunch of languages right here that he's showing.
36:52He's ID verified. I'll show you how to do that later as well. It's showing education, and then it's showing a bunch of reviews.
36:57It's showing his portfolio of things that he's already worked on. It's showing certifications, employment history.
37:02You can see how he has a very stacked Upwork profile. The most important thing for an Upwork profile is this job success right here, and also this text right here that says available now. For the job success, we literally just need to get some jobs done, and that also means that we don't go after the $75 an hour jobs just yet.
37:20If we can get a job for $15 an hour, even if it doesn't make us much money, it's still a massive win because it's going to show on our Upwork profile. Let's move on to this guy. You can see also available now, 100% job success.
37:31He's charging $62 an hour. He also has a book a consultation, so there's definitely a pattern here.
37:37He has worked four thousand hours in total. He has perfect reviews, a portfolio, a bunch of skills, and a project catalog.
37:45Same with this guy, also available now, 100% job success rate, 57 dot $63 an hour, 200 k earned, also booked consultation, so we should probably add this, and then a bunch of work history and a bunch of things that he has already worked on. It's always important when you're starting something new that you look at the people that are actually crushing it to steal all you can.
38:04What I've done is that I've actually taken a bunch of these high performing Upwork profiles. I've given them to Claude and created a Claude skill, which means that we can literally just pass in our own Upwork profile, and Claude can improve it. Let's do that.
38:16So now we can head back into Claude code. The Upwork skill should already be installed for you from all of my skills. Just write slash Upwork, and then we want to go to our own Upwork profile.
38:26Basically, just control a to copy everything and say, improve my Upwork profile, paste this in, and because of the Upwork skill that is already trained in the best performing profiles in our space, Cloud can now take that as a reference and then improve our Upwork profile.
38:43Of course, your Upwork profile should be with whatever you want to sell. I'm selling AI lead follow-up systems, so I'm going to copy this one, actually. I kinda like this one, AI automation engineer.
38:52Let's change this. Insert one. These as well, AI automation engineer, lead follow-up systems, and go high level.
38:57Instead of only lead follow-up systems, we can write AI agent systems and go high level and then make. Save. Now you can see that Claude is auditing our description as well.
39:07This is probably pretty good. I like this. Let's copy that and insert it.
39:14Save. This is a pretty good call to action. Message me with whatever it's leaking.
39:18So let's include that at the end. I like this and I'll tell you in twenty four hours if I can fix it. That's like kind of a call to action on our Upwork profile.
39:25So we can write, message me with your project. And I'll tell you in twenty four hours if I can help. Great.
39:33I think our Upwork profile is looking a lot better. Another tip that you want right now, I just set the price to $15 an hour. You want to be more precise, like, let's set it to $16.73 an hour, for example, and click save.
39:48The reason for that is that if you just set it to, let's say, $16 straight, it doesn't seem like a precise price. It's just, like, thrown out there. When you set it to something very precise, it makes the human brain think, okay, there's a reason that it's this precise, which can build credibility.
40:01Awesome. Our profile is starting to look a lot better with the help from Claude. Now it's time to sign up to some jobs.
40:08Before we do that, you need to know about the two plans inside of Upwork. Right now, we are on the basic plan, right, which is the free plan. And here, we get 10 connects per month when we finish our Upwork profile, which if you are serious about building out your Upwork profile, probably won't be enough.
40:23They have an offer right here though. You can see the plus plan right now is $20 a month, but if you go to your profile, then you can actually get 50% off right here, get freelancer plus for 50% off for one month.
40:35So that's $10 for to try it for one month, where you get a 100 connects instead, which is very, worth it. So I'm gonna go and click upgrade now to the plus plan.
40:44Let's get started with that. There we go. You can now see we are on the freelancer plus plan, and the first thing you want to do is that you actually want to head back to your profile, and then you want to verify your identity.
40:53This was something that we didn't have before with the standard plan, and this basically just builds trust with both the Upwork and also with our clients. And this also costs 35 connects right here. Right now, have a 150, so that's definitely worth getting this done.
41:06So I'm going to click verify your identity, and now we scan it with our mobile device, then and we have to verify with some ID. I'll be back when I've done that. And there we go.
41:14That took about five minutes, and now you can see right next to my name, we now have a verified badge that says this freelancer's identity has been verified for a government ID check and a visual verification. Nice. On the left hand side of your profile right here, you also have the availability badge.
41:29That's the thing I showed you before where it says available now, which costs 14 connects per week. I would also say that that's very worth it. So I'm going to click turn on.
41:38You've added the badge. And now we can go to our public profile, and we can see how it looks. You can see now it says this, available now, e automation engineer, $16.73 dollars an hour.
41:49There we go. The skills, the employment history. And the only thing that our profile is now missing is some jobs, because that is really what matters.
41:56The thing that people really take a look at is this right here, top rated plus and 98% job success rate, and then also total learnings, total jobs, total hours. And that's also why we set our price very low compared to these other people.
42:08You can see they're charging $62 an hour. We are only charging one fourth of that.
42:12It's because we want to get jobs first, we can have 100% job success rate. Nice. It's time to go out and actually start bidding on jobs, which is what we're gonna be using these connects for.
42:22So let's get into that. Alright. Let's start actually applying to jobs.
42:26If you go to the find work tab right here and click find work, then it's going to show you the jobs that Upwork thinks that you will be interested in. We can see our connects right here, and the way that connects work is that we are basically paying a little bit every single time that we apply to a job.
42:42What does this mean? Well, it means that we need to be cautious with what we apply to. We cannot just apply it to every single job.
42:47There's a couple of rules of funds, like if a post, for example, has 50 plus proposals, it's probably not the best job for you to reach out to. 20 to 50 is in a much, much better range, and it's really a speed to lead game.
43:02When someone makes a post, you want to be instant there. You can see this was posted yesterday, posted yesterday. You want to be there instantly.
43:08This one, for example, right here, was posted three hours ago, and the proposals are five to 10. This one would be a great one to reach out to. So what you want to do is that you wanna command click on this so it opens in a new tab, and you can see this opens up the job right here.
43:23We're looking for a skilled AI automation specialist who has hands on experience with GoHighLevel. That's us, Cloud AI, and AI agent development to help streamline business operations and improve client communication workflows. There we go.
43:35This seems awesome. Remote, freelance, potential, long term opportunity. There we go.
43:39There we go. This one seems like a really, really good job. And you see they are payment verified and phone number verified as well, and they have a 98% hire rate with 910 jobs posted.
43:50And again, what we are trying to do now is not to make the most amount of money. We need to build up our Upwork profile before we can do that. So if we think we can do this, which we can, we can click apply now.
44:00You can see that this proposal requires 11 connects, and when we submit it, we have a 103 left. How do we want to be paid? I'm going to choose by milestone, and then we can choose maybe something like 25, 25, 25, and we can do initial audit implementation, verify that everything runs smoothly.
44:28And you can see when we do this, we are gonna be making $67 like this. How long will the project take?
44:35Probably less than one month. And then right here, we can add our cover letter, and this is what is important to add. This is basically our application.
44:42What I've done is that I've taken some of the best proposals that I've seen, and I've created an Upwork proposal skill like this one. Write a high converting Upwork proposal for a job post the user provides.
44:52So let's write slash proposal inside of our AI workspace, and again, you're also going to have access to this skill. It's inside of the old skills folder. We're then going to take this post right here that was submitted three hours ago.
45:03I'm going to control a everything, so we can see what it looks like. Going to paste this in, and then say, write a proposal, please.
45:11This job might not be the best one. You can see it right here. It says $10 as a fixed price.
45:16But I think what they're trying to do right now with this post is literally just to find someone that they could maybe hire. You can see they've spent in total 9.3 k. This is also what our skill flags.
45:24You can see $10 fixed price. Clients average hourly paid is $7, and they have 67 open jobs. They're probably just using this for a hire.
45:33That's actually a good flag by Claude. So this is probably not the best job to spend your connects on. You can also see what country these people are making their posters from, Nigeria right here, which might not be the best type of client.
45:42You can see this one might be interesting. This is HighLevel AI automation fulfillment partner. So this is more of, like, a long term job.
45:48You can see they've spent $50,000. It's inside of GoHighLevel as well. So let's apply to this one instead.
45:53Maybe we can click apply now, then we can copy this entire thing. I'll write here to clear it and then choose the Upwork proposal generator again.
46:02Paste this in and say, build a proposal for this. And now it'll ask for some information about ourselves since it doesn't have any saved facts.
46:11Have you built high level voice agents before? Yes. High level.
46:15I know everything about the high level, and I have built a lot of voice agents in 11 labs, not only in English, but also in many other languages.
46:40Do you actually use high level NLN today to day? Yes. All the time.
46:44Pricing approach, do you want to quote each service line by line? Blah blah blah. On hourly.
46:49One hourly rate is fine.
46:54Your unfair advantage angle for fulfillment partner, The US time zone overlap prior to fulfillment experience in house dev custom calculators. Sign off name, Albert. Yes.
47:03Your unfair advantage. Let's come up with an unfair advantage. Prior agency fulfillment used to run my own agency.
47:14There we go. We give Claude some context, and it's going to build out this entire proposal for us. Don't wanna paste this in one to one.
47:20I'm going to copy this right here. You can see this would cost 27 connects. Schedule a rate increase.
47:25Let's do never for that. You can always tell that later if you want. And then a quick tip is that you can write, this is not written by AI.
47:34Then I don't think this is good. We'll just write, I used to run my own AI automation agencies, so they use cell I fulfill model.
47:41The fulfill is not that good, so I'll just write, so I so I know the use cell I fulfill model.
47:49And when it comes to these proposals, you want to educate them. So I'll write it's important that we do good onboarding flow, snapshot deployments, sub account hygiene, flows, and snapshot deployments, and then add some information that they might not know already.
48:01We can actually use the go high level API through something like make NNN to also deploy some accounts in both GoHighLevel and in LivingLabs.
48:18How it starts? Let's first audit your current high level setup, and we build the reusable template that also deploys through in it in, then go high level and 11 labs through the API, and then take over all.
48:40Let's do like this. This is not like what I did. I would write take all then I take over all clients' world plus two names to close deals.
48:47Great. $16 now. Aviso quote fixed per client.
48:52When I see the current onboarding process, Maybe just like current onboarding.
48:59This is a pretty bad question. There's also some weird spacing right here. Let's remove that too.
49:04One question, is it custom build per project? Does it custom build per project? Do you sell productized services will help me to plan it out.
49:14Help button like this right here. Then we can add an attachment, like, we can, for for example, add that calendar right here that is completely full.
49:22And there we go. This is a really good Upwork proposal. What really can make it better is that if we also at the end right here, write made a Loom as well that describes how we can set this up.
49:33And then in this Loom, you wanna showcase that you actually know what it is that you're talking about. Let me let me show you an example of what Loom we can create for this, for example. I'm gonna go to loom.com, and then I'm also going to go to make.com, which is probably what I'll be using to set this up.
49:45And, this just becomes very easy when you already know what it is that you are doing. And, again, in these proposals, you kinda wanna educate them on what is actually possible for them to do. So what I'm going to do now is that I'm gonna make a Loom video right here.
49:58Very important that we show our face. We can click the corner Loom. We turn on our camera right here.
50:03There we go. Make us as big as possible. That's a good idea.
50:07So they see, okay, we are an actual person, not an AI. There we go. Make sure to make yourself big in the corner and then click start recording.
50:14Tile screen. Sup, guys? Just wanted to elaborate a bit about what I explained in the proposal.
50:20What you want to do is probably use either make.com or initn. It's possible in both. But many don't know that inside of GoHighLevel, and this is actually make specific, this doesn't exist in initn, you have this API call right here that says adds an account.
50:36And what this basically does is that this adds a subaccount automatically, so you don't have to create one inside of inside of Go High Level.
50:45And then what you can also do after you've used this is that you can use the Go High Level v two API, this one right here, and you have a an account update right here.
50:57So you can basically do anything you can do manually inside of Go high level. You can do this through the high level API, not something that's inside of actual MIG.
51:05You will need to create an HTTPS request like this, make a request, but then you can basically do all of these things.
51:12You can update every single thing inside of HighLevel. You can connect with other services as well. You can basically automate the entire thing.
51:20So so yeah. Just wanted to make sure that you guys know that this exists, and yeah.
51:24Let me know if you need help in setting it up. There we go. That was what a one minute fifteen second Loom.
51:30Just copy this link right here, and then paste it at the bottom. This is a perfect proposal. This is not written by AI.
51:40I used to run my own AI automation agency. So the u cell, I fulfill model. So I actually really like the u cell, fulfill model.
51:49It's important that we do good onboarding flows and snapshot deployments. We now you actually use the GoHighlevel API through something like MetaNNN to also deploy subaccounts in both GoHighlevel and 11 apps. How to start?
51:59Let's first order your current high level setup, then we build a template. It's sometimes good as well to write that we have a loom down here at the bottom. This is not written by AI.
52:06Made a loom below. That explains. But I used to run like this right here.
52:11You want to make it seem like it's not written by AI, and all of a sudden, we have changed a lot about it ourselves. But this is a perfect proposal. We have added an image of a screenshot of a built calendar.
52:23We don't wanna bid contacts, then we just click send for 27 connects. There we go. Submit.
52:28We can even see if it has been opened yet so far. There's 19 proposals, six open, 13 unopened, and all proposals are at 25 an hour.
52:37Top rated one is also at $25 an hour. So we actually have a real shot of landing this job. That's just what you want to do.
52:43You want to apply to a bunch of jobs, and you wanna stack up on connects. And you might think, well, I'm gonna have to keep buying connects all the time, and that is honestly true. But there's no other platform really like Upwork where you can find so many interested hot leads, and honestly, it's very very cheap.
52:58If you had to run Facebook ads, for example, for these type of leads, you'll probably paid a 100 or $200 per lead. Right now, for 10 connects, we're paying $1.50. So make this your routine.
53:09Go out, send a bunch of requests, actually make them good, spend some time on them, spend at least, like, ten to fifteen minutes on proposal, and of course, at least when you're starting out, and then when you get better, you, of course, can be able to do it faster. Make a Loom video to every single one. This will help you land jobs.
53:24And you might think, Albert, well, already to this, there was 20 applicants already. No one is going to choose me. Let me tell you why that's not the case.
53:32And the best way to show you that is actually not to hop on my freelancer profile right here, but to hop on my buyer profile, because I've also spent thousands of dollars on freelancers on Upwork, so I know what I like to look for when it comes to freelancers. Let me show you that perspective as well, so you actually know what your profile looks like and what buyers think about when they look for freelancers.
53:52So right here is now from my other profile, so from my company profile, not from the freelancer account, but from the company profile, and I made a post a couple of weeks ago about finding a video editor to basically cut up videos, and you can see I got 11 proposals on this post, and you can see who actually signed up.
54:14One from Pakistan, another one from Pakistan, a third one, fourth, fifth, sixth, seven, one from Nigeria, one from India, one from Pakistan, another one from Pakistan. And the reason that your competition is not as good as you think is that a lot of the freelancers already on Upwork are from third world countries, which means that if you are from Western countries, you definitely have an advantage.
54:35This doesn't mean that if you're from India that you cannot do it, you're just gonna have harder competition because there's a lot more freelancers from those countries already on the platform. And let's take a look at the actual proposals.
54:45You saw how I reworked mine to not sound like AI, but what you'll find is that almost everyone right here just has basic proposals. As a top rated plus video editor with over six years of retention, retention focused into a YouTube post production, like it just sounds AI made.
55:02Then he has inserted a bunch of videos that doesn't really concern us. In terms of AI, it's a core part of my workflows. I use tools like ChatGPT.
55:09I work in EST. We can discuss the monthly rate. That's it.
55:13I love to connect. Blah blah blah blah. You can see very AI generated, not very personal.
55:16This one as well. Portfolio at the end of the process, like, is just like a template. You can see that it's a template.
55:21It's not personalized to the actual job. That's just to show you from the buyer side as well how it actually looks, so you know, okay, this is what I should write, this is how I should catch their attention. The most important part of your cover letter as well is this part right here that we wrote.
55:36Right? This cover letter. Because this is the thing that they see and determine, okay, am I actually interested in this person?
55:41And it also shows how important it is to actually just land your first couple of jobs on Upwork. And the thing is, when you're just starting on Upwork, you're gonna have to do more effort. Right?
55:48You're gonna have to make a better cover letter. You're gonna have to make it more personalized and more looms. But what you can see right here is that this is definitely what moves the needle, like the job success rate, how many jobs completed, total hours, and also money earned.
56:00The bias on Upwork, they definitely look for that. And then to set some expectations, building an Upwork profile can take months, if not years. It's not a get rich quick thing by no means.
56:09It also takes long work. But what you'll find out is that these clients that you work with on Upwork, all of a sudden, they refer you to someone else, and someone else, and someone else. And Upwork, even though it is a freelancer platform, it's the perfect place to start.
56:21Awesome. I showed you this just to make you understand that the competition on Upwork isn't really that bad. And Upwork, even though that it's a freelancing platform, is in my opinion one of the best places to start, and here's why.
56:34You build experience. It's what I mentioned in the start. Right?
56:37Get good before you get rich. And this, by actually doing the work, is the best way of getting good, and what also happens is that all of a sudden you have clients, you talk to people, and you'll be referred to other clients. It's a snowballing effect, and all of a sudden when you have a bunch clients, you can automate a lot of the work with Claude Code, and boom, you're running a one person business with Claude Code.
56:58So if you want to do any type of outreach, recommend that you start with Upwork. And just to set some expectations, it might take a couple of weeks before you land your first freelancing client, but it's still gonna be so worth it. However, if you want to speed that up and get even more volume, then I'm gonna show you something insane that we can do with Cloud Code in the next part of the course.
57:16Let's get into it. Great. To be clear, now you just want to send as many connections as you want and land your first couple of jobs on Upwork.
57:24That's why you have the highest likelihood of getting your first client the fastest. If you want to push that further and send out a bunch of connections to every new job that is posted on Upwork that you're interested in, and you want to do even more than that, then this part of the course is for you. If not, that's completely fine, and you can skip to the next part of the course.
57:40Do you remember when we talked about this map right here, where all of these countries marked with green, they are getting very, used to AI, especially like chatbots and voice agents. All the other countries, however, are a bit behind.
57:54So that is South America, like, for example, South Africa is a very booming market. It's especially Europe, and then also many parts of Asia. If you live in one of the blue countries, then you should especially listen up in the next part of the course.
58:07This will also work in the green countries, but it will have lower reply rates. What I've seen is that this next method is absolutely insane in the blue countries right here, and what we'll be doing is something called cold email.
58:19You've probably heard about it, which is where we can send a lot a lot of cold emails out to a lot of qualified prospects, and then the people that are interested will reply back, and we can automate this entire thing.
58:31My favorite cold emailing software is instantly dot a I. I'm not sponsored to say this. That is just my favorite platform.
58:37If you want to set this up and you want to support channel, then make sure to check our affiliate link right below this video. Again, I'm not sponsored, but I've just created an affiliate link through their affiliate platform. But why is this software so good?
58:48Well, it allows us to get a bunch of email accounts very, very easily, and then send a bunch of bunch of emails to qualified leads, and then they have an MCP that allows Claude to do 90% of the work.
59:01So let's set that up first, shall we? The first thing I want you to do is actually search for the Claude desktop app, if you don't have it already, and then download it for whatever computer that you have. So I'm gonna search Mac, and then download Claude right here.
59:14Download for Mac OS, that's going to download the installer just like this. The next thing we want to do is actually not go inside of Claude code, but then go inside of the Claude desktop app that we have just installed. The reason we do this is because that's by far the easiest way to add NCPs.
59:28If you go in the bottom left corner right here, and then click on settings, and then go inside of connectors, and then click this customize right here, because they have now moved it, we can see all of the connectors that we have. So I can now click add connector right here, add custom connector, call this instantly, which is the platform, and then what we need to do now is to go back to instantly.
59:51So now's the time. If you haven't already, then sign up for an instant new account. Again, if you wanna use my affiliate link, then that would be very appreciated.
59:57We can now click on the bottom left corner right here, go to settings, then click on integrations, and then click on API keys right here. And now we want to create an API key.
1:00:07This is what allows Claude to basically have access into InstaLift. So click create API key. I'm going to call this Claude.
1:00:14I'm going to select all, so I can do everything it needs through the API, and then we click create. And this is going to create this API key right here.
1:00:22So let me copy that one, and then I'm gonna go to this docs right here. I'm gonna leave this docs right below this video, but that's basically the MCP URL that we are using for instantly. So you can see it says your API key right here.
1:00:34You will copy this, change it with your own API key, and of course, I'm going to rotate this API key after we are done. I can now copy this entire thing, go back to Claude, edit right here, and then click add.
1:00:45And there we go. The Instant MCB has now been added. I can now just go back if I wanted to test it right quick, go into Cloud Code for example, and just as it right quick, do you have access to the Instantly MCP now?
1:01:00This is just to verify that it actually has the access that it needs. You can see it says, yes, the instant MCP is available, and I can see tools like blah blah blah blah. There we go.
1:01:09I'm also just going to ask it right now, can you pull campaigns just to check the connection works. This is to ensure that it both has the MCP. We just added that, but that it also can list all of the campaigns you can see it's using this right here, list campaigns.
1:01:24Let's see if it returns, connection works, one campaign returns, blah blah blah blah, like this. Great. The reason we installed the m c p inside of the Claude app is because that makes it much much easier when we go to Claude code.
1:01:35I'm going to click control c right here a couple of times. That's going to Claude, start Claude. I'm going to write clear again and write Claude again, and that's basically how we restart Claude.
1:01:45Now I can write slash m c p, hit enter, and under Cloud AI, we can now see the MCPs that we have added, and you can see right here, Cloud dot a I instantly is connected with 38 tools.
1:01:58I can hit enter on that as well. I can hit enter to view tools, and then we can see all of the tools that it has access to, which was, what, 38 tools. Right?
1:02:07It can basically reply to emails, list emails, create lead list. It can do it all.
1:02:11And you can see the status is already connected because we added the API key right here. Cool. We can write clear.
1:02:17That clears the flawed code session, and then we can basically get flawed to create these code email campaigns for us. Before we do that, we need to do a very important step, which is we need to go into instantly and get some email accounts. Email accounts are basically the emails that we are sending from, and there's two different ways that you can do this.
1:02:33I can click add new right here. You can either choose pre warmed accounts, which is basically accounts that instantly has warmed up for us. These accounts don't have our own name.
1:02:42You can see it's Arthur. It's Charlotte, and it's using these interesting domains right here, but they are pre warmed, which means that when we send emails, it will have high deliverability. And those are the ones that I recommend that you start with, because usually you would have to warm up your own domains, which takes a lot of work and thirty days.
1:02:58Here we can just pre buy them. You could also connect existing accounts if you want to, but for now, we are just gonna choose the pre warmed accounts.
1:03:06We can click go. Now we can select some domains. We want to find something that seems very professional, like landaiimagine.org.
1:03:13That's probably pretty good for an AI agency. Right? I'm just going to choose land AI imagine.
1:03:17It's $15 right here. And here you can see we get five emails. You will set the forwarding domain.
1:03:21This is very important that you set it to your own website, because if people then go to this website, it'll go to your own. So I'm going to go to my own website, shiny.ai, and to paste that in right here where it says enter valid URL, and then click place order.
1:03:34And now we would just need to check out right here. You can see it's $65 to get these domains, but that is very, worth it.
1:03:40It's $10 per domain. When you've bought your email domains, it's going to look like this. Now you can see we have five email domains right here from this domain right here, thrivegateway.com, and we can see the health score right here next to them.
1:03:52You can see 100% health score. Each of these, you can see it says email sent, can send up to 20 emails a day.
1:03:58So when we have five emails, then we can send a 100 emails a day, which will be plenty, because we're gonna take a more targeted approach with our cold email. We want to get this right. So after you have your domains, and after they are at 100% health score, we're ready to start and create our campaigns.
1:04:14And of course, we're gonna be using Claude code for that. And I've actually already created a skill that's based on ours, most high performing campaigns, that you, of course, also have access to. So let's go start up with that.
1:04:23Before I show you how to set it up with Claude, I wanted to show you this campaign right here, which is what you should strive towards. And this is also what the Claude skill is based on. You can see this campaign right here is going on to dentists, and we are 60% through.
1:04:35We have sent a thousand emails. This is both initial emails and then follow ups. Follow ups in cold email is very important.
1:04:41And from that, we have 50 people that replied with a 5.1 reply rate. 5.1 is extremely high in cold email.
1:04:51Usually, when you send cold email campaigns, you maybe get a 0.5% reply rate with a lot of people that say, fuck you, or I don't wanna hear from you. But from those 50 replies, a lot of them were also, take me off the list or don't talk to me again, that is just a part of cold email.
1:05:04But from those 50, we have six interested leads. These are people that said, yes, I'm interested. Let's hop on a call.
1:05:10Let's talk more about this, or ask a follow-up questions that we can then answer and start a conversation with, and you can see if you can send, let's say, a 100 emails a day, you get five dead replies, that means that every other day you get an interested lead, which is extremely good, and you can run a business on that.
1:05:25So let's get into it. Let me show you how we create this campaign. Awesome.
1:05:28If we open a new Claude code chat, write instantly dash campaign like this, This is the instantly campaign creation skill.
1:05:38We can then write, I want to create a campaign for let's do roofers in Sweden.
1:05:47And if you are from one of these blue countries that I showed earlier, then usually what a lot of beginner do is that they still go after the English market, so after The US, The UK, Australia, but because AI is being adopted so massively everywhere around the world, I would actually say that it's a much better advantage that you speak some unique language.
1:06:06For example, if you speak Swedish, please go after Sweden. Just make sure that it's legal to reach out with cold email in the country that you want to go after.
1:06:15But please go after those countries. Why? Well, you're gonna have a much higher reply rate.
1:06:19For example, I don't really speak Swedish. I understand a little bit. But someone that speaks Swedish have an unfair advantage over me by knowing the language.
1:06:28Same if you're from Greece or from Portugal or from Belgium, then you speak a language that not the entire world speaks, which basically just means that you have way less competition. So just to do it as an example and to show you that Claude is going to do all of the creation of a campaign for us, I'm going to create a campaign for roofers in Sweden while not speaking Swedish, only understanding a little bit.
1:06:49And that's all we need to write. Now we hit enter. Now Claude code is going to build out the campaign using the instant VMCP that we have connected it to.
1:06:57And you can see, it says, what are you selling to Swedish roofers? Describe the offer in one paragraph. What's what does it do for them?
1:07:03Let's do new website just to show you how it looks. What's the primary pain? Slow coding loses jobs, few Google reviews.
1:07:10Let's do few let's do a couple, like, Google reviews, ranking on SEO, and also missing out on leads because they don't have the right information.
1:07:25What tangible asset will email a free offer? This becomes the main subject line. Let's do the lost revenue calculator, so they can see, okay, how much they're saving.
1:07:33What should the campaign be called? This doesn't really matters, but let's just call it Rufus Sweden, and then submit answers.
1:07:38And now, Plotcode is going to build out this entire campaign for us. And here we go. We have a draft that goes over everything.
1:07:44This means I was thinking about calling, but I wanted to write to you first, and then saying right here that I'm thinking that you're losing jobs because your website is not showing high on Google. Is this something that you would need some help for? I really like this.
1:07:56This is all absolutely awesome. So I'm going to approve. Let's create campaign.
1:08:01And now, Claude is going to build out the instantly campaign for us. We don't have to do a thing. Now you can see it is calling instantly.
1:08:07If I click control o, we can see what it does. It's making this request right here with the information. There we go.
1:08:14Campaign is created and called roofers Sweden. If we go into instantly right now, head to campaigns, we can see that the new campaign has now been created. If we click in on this, we can go to sequences, which is basically where we write the emails, but it looks completely blank for some reason.
1:08:28It looks completely blank. I don't see the emails. Let's see what happened here.
1:08:31Oh, I just had to refresh it, do a hard refresh, shift command r, that half refreshes the page. Now we have the entire campaign inside of Instantly that we just rolled out. So we need to make sure that we click save.
1:08:44There we go. Save on everyone. Now we have our campaign ready to go.
1:08:47The last thing we need is just leads to send to, and I'm gonna show you the fastest and what I found to be one of the cheapest way to get really qualified and verified leads. Let's get into it. To get qualified leads, we're gonna use a platform called Apollo.
1:09:03But don't worry, because Apollo itself is actually extremely expensive. I'll show you a cheaper way to get the same quality leads as Apollo, but using a different site.
1:09:13So let's say we want to find roofing and building companies in Sweden, for example. We then make sure to go to people right here. We go to location and we search Sweden, then we go to industry and keywords, and we could search for that will probably be construction.
1:09:28Let's maybe also do real estate. If it's like home builders, then real estate is probably also gonna You can see now we have a 102,000 people.
1:09:36You probably want to narrow it further, so you can also do things like job titles. You, of course, want to do go for someone like the owner of the business or the founder of the business.
1:09:47Add a lot of keywords to people that would fit your description for your ideal client, like CEO, founder, CEO and founder. Add all of these, because you can see that increases the list size right here.
1:09:59You can also exclude titles if you want to, but now you can see we have a list of around 3,000 people right now. What's very important that you go into email status and then make sure to click this verified right here. That's going to get you the best quality of emails, and you can see our list is now at 2,100 people.
1:10:14You can also choose if it should be b to b only, b to c only, if it should be e commerce companies, fintech companies, nonprofits, SaaS, consulting. You have a bunch of things to choose from inside of here. This one is also one of the most important filters.
1:10:27You can choose what range you want to reach out to, so how big the company should be. If you're just starting out, you're probably not gonna close an enterprise deal right off the bat.
1:10:37That would be insane. We could probably choose, like, from one to 10 to 11 to 20, maybe up to 21 to 50, and that gives us 1,800 people that we can reach out to.
1:10:48That way we leave out the biggest companies, because you can reach out to those in the future, but you need some experience first. And there we go.
1:10:54Now we have a bunch of Swedish companies, and we can access all of their emails. But, usually, if you're using Apollo, you do it directly through their Apollo platform, which is extremely expensive. What you can do instead is that you can use this platform right here that's called trustedleads.io, enterprise grade b to b lead list, starting at $0.005 per lead.
1:11:14So go in there, create a user, again, I'm not affiliated to say this, and then you can click new order right here. Here you write in your name, your email address, write in a phone number, and then you want to choose the Apollo URL right here, and the number of leads that you want to scrape. Very important that you turn on the verified leads that I showed you earlier, otherwise, you're gonna get bad lead quality, and you're gonna be paying for leads that are not good enough.
1:11:35So let's say a total of how many do we wanna scrape, like 1,500 of these leads, And you can see that's going to cost us $25, which is 67% less of what we would use on Apollo.
1:11:48So now we can click continue. Now we want to take this URL from Apollo right here and paste it in. It says that the email is not verified, but I'm almost 100 sure that we set it to verified.
1:11:59You can try and refresh this page just to make sure that you that you have the newest URL like this. Let's copy this one, paste it in, and see if it says null. That is fine.
1:12:07Let's just verify that we have the right search. We do. Then click continue.
1:12:11You could if you wanted to pay additional to actually verify the emails, but I'm gonna show you a better way of doing that. So we can now click no additional verification, do a signature, and then click proceed to payment.
1:12:22After you're done with that, then it can take a couple of hours, but then you receive an email like this. Your order is ready. Thank you for your order of b two b leads.
1:12:28Download the leads list lead list right here, and this is a previous list that I bought, and this is actually 10,000 leads. You can see how many leads we have right here.
1:12:36When you buy leads from inside of here, it is very smart to do it in bulk. You can see here we have a lead list of 10,000 leads. What I'm going to do now is that I want this as a CSV file, I can click file.
1:12:47I can click import. Click upload. I can upload this CSV file right here.
1:12:52That's gonna take some time because it's a very big file. Replace spreadsheet and import the data. And here we go.
1:12:56Here we have our big list. We have to do one more step. For now, I don't need all of the rest of these leads, so I'm just going to delete those like this.
1:13:04This is just to show you how you import those into instantly. I'm going to call them leads and maybe just the date, May 16, and I'm gonna click file, click download as CSV.
1:13:14And then before we put those into instantly, we want to run them through Million Verifier first. This is basically finding all the emails that are old. When you're sending out cold email campaigns, you don't wanna send to email domains that doesn't exist.
1:13:26That's like a tell that you are spamming with emails, which the email providers don't like. Instead, you wanna run it through something like Million Verifier first. You click select file, and we can select this one right here, and you can see we're gonna be using 969 credits for this, and we're gonna be removing the duplicates.
1:13:42We click start verify, and this will now create a job right here. You can see it says takes fifteen minutes, and this is basically going to find all of the duplicates, and it's gonna remove all of the old emails. It's basically verifying the emails to make sure that they are good to send to.
1:13:55You don't wanna skip this step. That's very important, because if you skip it, then you're gonna be sending emails and your domains will instantly be backlisted, and then you're gonna have to buy new domains. So make sure to use Millionverifier as well.
1:14:05Again, can find links to all of these platforms in the description of this video. There we go. That took around ten minutes, and you can see now we have a total of 64% of emails that were good emails, 17% that are risky, and 18% that are bad emails that doesn't exist anymore.
1:14:23You definitely don't want to be sending to these bad emails, because those emails are going to bounce and it's going to hurt your domains. Risky emails you can sometimes send to, but I would still not recommend sending to these inboxes. And then we have the 64% that are good emails.
1:14:38Those are the ones that we will be sending to. I can now click download report right here, and I want the good emails only. So I'm going to click right here.
1:14:45That's going to give me this file right here, only the good emails. There we have it. And now it's time to add these emails to instantly.
1:14:52So I'm gonna go back into instantly. I'm gonna click on the leads tab under my campaign. I'm gonna click add leads, and then from a CSV file, and then I'm going to choose the CSV file that we just downloaded, which is this one right here from berlinverify.com.
1:15:06And you can see we now have to map the fields with first name is going to be first name, last name, last name, title is going to be job title, LinkedIn URL, email. All of these, we don't wanna import it, see if there's anyone we need.
1:15:18Probably not. Company name we want. Website we want.
1:15:21Industry, probably not. And we don't wanna verify our leads because we just did that ourselves. Now we can click upload all.
1:15:28This will upload 625 contacts to your campaign. Let's click yes, and now you can see all of our leads are uploaded to this campaign right here.
1:15:37These emails I got done probably a couple of months ago. So we are, of course, not gonna send emails in Swedish to these US leads. And then when we have our leads, last thing we need to do is go inside of options and then just add our email accounts like this.
1:15:50You can see these email accounts right now are inactive, and when you've just gotten them, they're gonna be good to go. And then you can click launch and publish like this.
1:15:58So when you have your sequence in place, when you have your leads uploaded, when you have your pre warmed email accounts, then you can click start campaign up here in the right corner, and that's going to send out all of these emails. Another thing that you can change if you want to, you can see this is set from 7AM to 4PM by default and the time zone right here.
1:16:16This is a pretty good window, but if you wanted to change that, you could, of course, do that. You can also send on weekends if you want to, but I probably wouldn't recommend that. Do it in business hours.
1:16:24And here's what's gonna happen. When you start sending out, you're going to start receiving replies, and if you're sending to these countries that only you speak the where you speak the language and not that many else do, then you're gonna get a very high reply rate if you have made a good campaign with a good offer. You'll find those responses inside of Unibug.
1:16:40You can click on interested, and then you can see the responses that are interested. And you can see this is from a real campaign that we sent around a month ago. And what you can see is that we are receiving interested responses.
1:16:49This says, yes. I'm interested in a conversation. This person right here says that they're interested in an offer to see what we provide.
1:16:56And in this campaign, we have only sent to 278 leads. I hope you start to see the power of COLI billing, especially to countries where you speak a unique language.
1:17:05If you want to absolutely maximize so you get as many clients as possible, I recommend running Upwork as your main priority. Every single day, go out and send connection requests because those are the hottest leads and the easiest way to gain that initial experience, which is so crucial. And then if you wanna maximize, then also run a campaign in the background because this is just gonna be running, sending out emails as you go, and probably book you a couple of meetings a day if you have good offer and are offering an interesting service in your market.
1:17:30These two things, when you're just starting out, is the highest leverage thing that you can do. I'd honestly recommend getting started on that right now. But you, of course, also need to know how you can actually convert meetings, how you take this information that you're getting on meetings when people are actually interested in proceeding, how you take all of this information and give that to Claude, so Claude can build out and do the majority of our service delivery.
1:17:51This is going to be extremely valuable for you to watch, so let's get into it. Great. Now you have outreach going out on both Upwork and on InstaLift.
1:17:59The next part of this course is one of the most important, because this is where most beginners, they mess up, and they mess up in three places, both before the call, under the call, and after the call. So let's go through some ground rules that you need to understand before you start booking meetings and closing clients.
1:18:15The first one is that you never, and I mean it, never mention price before the call.
1:18:22And here's the reason why you don't do that. If you mention the price, let's say that the client asks for, okay, what is the price before I hop in a call, and you say, well, it's $2,000 a month. What then happens is that the client judges you on the price and not the value.
1:18:37The client doesn't know enough about your service. They don't know what work you'll be doing and why it'll give them a bigger return than the price that they'll pay. So when they hear a $2,000 price, what they think about is, okay.
1:18:49Do I want to pay $2,000 right now? Probably not. I'm gonna stop replying to this guy.
1:18:54And the only reason they do that is because they don't know your value yet. You only mention the price after you're on a call with them, after they know how you can help them, because it's a way better equation, a way better thought process if you show, let's say, $10,000 worth of value, and then show the $2,000 price point.
1:19:10All of a sudden, that becomes a no brainer deal. So many beginners make this mistake, and they throw out a bunch of meetings because they don't understand this key principle. That is rule one.
1:19:20Rule two, that you always want to send a confirmation and reminders. It happens more often than you think that people simply just forgets that they booked a meeting with you, and then they get kind of, like, embarrassed that they forgot the meeting, and you never hear from them again.
1:19:34You always want to send both the confirmation and reminders to every single call that you hop on. So the process looks like this.
1:19:41You have people interested either over cold email or over Upwork. You decide to, okay, let's hop on a meeting, and then as soon as they have agreed to a time, you send them the confirmation. Right?
1:19:50This is 4PM Monday. This is where we meet, and you send this on as many channels as possible. So if you have them on email and their phone number, then you send it on both, as many channels as you can.
1:20:01You do this instantly, and then depending how long the meeting is out in the future, you then also send reminders. I'll recommend that you send reminders at least every three days up to the meeting, then do it one day before, and then do it one hour before, and then do it five minutes before, And the way that these messages look is basically just like, hey, first name, just reminding you that we have our call tomorrow.
1:20:23Excited to see you there. If you're doing, like, a demo for them already, it's a very good idea to mention that you've also put work in before the meeting because that then makes them feel guilty if they don't show up. So if you build a demo or a test website for them or something ahead of time, then say, I've prepared a demo for you based on your own company so you can see exactly how this will look for your business.
1:20:42If they then don't show up, they're gonna feel guilty. You wanna include that in the reminders, and then the last reminder, like, five minutes before, did you just write, hey, first name, I'm just finishing up another call.
1:20:53I'll see you in our call in, like, five minutes. That's a really good message for this last reminder. Something I also see, and this is connected to never mention price, is that some people wanna skip this step of hopping on a call.
1:21:05The reality is that that's just making your life 10 times harder. Not everyone is going to be ready to buy right in that moment. They might need more information, so you need to hop on a call with them.
1:21:14You wanna give them limited amount of information, so where they really get the information is on the call with you. And it's very simple. If you have an issue right now where people don't show up to the calls, it's just an equation of do they think that this call would be valuable enough for their time?
1:21:28Do they think it can actually help them? And if you have an issue where people don't show up to the calls, the reality is that they don't think that it's gonna be worth their time. That is why you need to include the reminders.
1:21:36You need to tell them I've built a demo specifically for your business. I've prepared this and this and this. We're gonna see this and this and this, and you, of course, need to solve a painful problem.
1:21:45Okay. So before the call, you never mention the price. You always wanna be judged on value.
1:21:49You always send confirmations and reminders on as many platforms as you possibly can, and then you always never skip the call. You don't sell straight away. You hop on a call with them before you start selling.
1:22:00Okay. Now on the call. A lot of beginners make the mistake of talking too much.
1:22:05The best way I found to sell these AI services is not by being the one talking. It's by taking what I call the doctor approach. When you go to the doctor, they don't just hand out medicine the moment you come in the door.
1:22:17The first thing they do is that they ask you a bunch of questions, like, where does it hurt? How long has it been hurting? And then they try to diagnose what's wrong.
1:22:23You wanna take the same approach when you are selling, and the way you do that is that you only talk 20% of the time, and they should talk 80% of the time. And when you talk, it should mainly be questions, And when they talk, they should answer your questions.
1:22:40And here's why that works so well. When you're the one asking the questions, you're the one leading the conversation. And one thing is that you mention why your service is valuable.
1:22:49And even better thing is when the client themselves mention why your service is valuable. When you get people to say themselves that they need something, they convince themselves way better, and the way to do that is by asking questions.
1:23:01Let me give you an example. If someone has, let's say, only a couple of reviews on Google Maps, you you can then ask a question like, okay, what do you think happen when a customer compares you to your competitor? Your competitor has a 150 reviews.
1:23:14I can see it right here. You only have five reviews, and one of them is a Feaster review. Who do you think your client is gonna go with?
1:23:21They're gonna say, well, based on this only, they're probably gonna go with a client. Do you see what we do there? We get them to mention themselves that they have a problem, and when we take this doctor approach, it becomes a much stronger sell, and it doesn't even feel like selling.
1:23:34You're just saying, well, what do you think happens when when this and this and this in your current situation? They mention a problem. And then it makes it much easier later in the call to say, well, I actually have a solution to your problem.
1:23:45That's the entire point of the call. It's to position a problem that they actually do have. If someone doesn't have a lot of reviews on Google, they have a problem that's hurting their business, and then positioning you as the solution.
1:23:57That way you don't need to convince them. You don't need to be overly salesy or talk a bunch. You're literally just asking good questions, and by asking good questions, you control the conversation.
1:24:06You can steer it in any way that you want, and that leads me to the second rule, which is that you need some structure on your calls, and the structure that I like to use is initially just these doctor questions, initially just asking if they even have a problem.
1:24:19And what you'll run into sometimes let's say that you are providing more Google reviews with automation, and you run into someone that already has 500 reviews and are on top of everyone else. They probably don't need that specific service.
1:24:30You might sell them another service, but you don't wanna sell something to someone that doesn't actually need your service. That's unethical. So the structure that I like to use is one, ask a bunch of good questions that positions a pain.
1:24:42Then two, tell them that, well, we can take this pain, and we might have a solution. Let's see if it even makes sense.
1:24:49And then you can show a demo or you can explain, well, we can actually do this to fix this problem. And three is that you then explain further about this solution and say, well, the price of this solution is and only when they've seen the solution, when they've seen the value that you can provide, that's when you mention the price, and that's when you shut up.
1:25:08They now get themselves to decide, okay, this pain that I have in this solution right here, is this worth it or is it not worth it? And what will usually happen is that they will then have some kind of objections, and those come at the end.
1:25:22Objections are usually like questions. Well, what happens if we try this for a month and it doesn't work? Or well, what happens with my current website if we are changing it over?
1:25:31How would I do that? It's basically just questions from their side that show some kind of concern. So what you then have to do is just answer those questions in a good way, and then make sure that that that concern is squashed, and then at the end, that's when you close.
1:25:43That is the structure that I like to use where it doesn't really feel like a sales call at all. This is what you should do. Let me show you what you then shouldn't do.
1:25:52A lot of beginners they make the mistake of basically just hopping and call and just yapping. So a lot of beginners structure look like this. They just explain their product, and usually they basically just explain features, and then they ask, do you wanna buy after they've done that.
1:26:08Right? So what does this look like? This is maybe if you create a slideshow that's just about your product.
1:26:13This is what we do, and you just sit down and and go through it. The problem with this approach is that you don't position a pain. You don't ask good questions.
1:26:19You don't make sure that they actually need your service. You basically let them make the decision themselves of, okay, do I need this or do I not need this? And if you have a boring slideshow, chances are that they're just gonna zone out.
1:26:29All of a sudden they hear a $2,000 price point, and they say no. Alright?
1:26:34So don't do this, do this structure instead.
1:26:39Alright? Questions first that position a pain. This is questions like, well, your competitor has way more Google reviews than you.
1:26:46They have a good website as well. You have almost no reviews. You have a trash website.
1:26:50What do you think happens? Well, don't say trash website. Say, your website is quite old.
1:26:55Who do you think your client is gonna go with? And just like a first impression. You want to say themselves, well well, in this scenario, they're probably gonna go with my competitor.
1:27:02You position a pain that hurts their business. Then you position your solution, well, we can install this AI that reaches out after every single job and gets you Google reviews. We can also install this website.
1:27:11I can build it in two days, which looks way better than what you currently have. Do you think that would help increase your first impressions with new customers? Again, you even position that as a question to make sure that they understand the solution is worth the price.
1:27:23Great. Great that you now understand this. So many beginners mess up right here.
1:27:27You don't want to do that. But you now understand this, which is good. Great.
1:27:31So now you're starting to get sales. What do you do? Well, you follow rule number three, which is always book next call on this call.
1:27:44Always book next call on this call. Here's the reason. When you're just starting out and you don't do this, there's gonna be a bunch of times where someone agrees, yeah, let's go, let's do it, and then you don't book the next call on this call, and what then happens is that they ghost you.
1:27:57If they have something in their calendar, it increases chances that they actually hop on a call with you again. This rule is extremely important, and it's gonna save you a bunch of deals.
1:28:05So what you basically just do on the call is that you say, great. Happy that you are on board. Let's book it into calendar already the next time we're gonna talk, and then I'll have this and this ready for you.
1:28:14Again, you kind of sell the next call, you hype up the next call, and you make sure that they see that you've made an investment now in them, that you've spent time, so they feel bad if they don't show up. And the last rule, and this honestly surprises me that people still do this to this day, is that they don't take payment on the call.
1:28:29You always want to have some initial commitment, and even if you're doing like a free trial, you still take the payment information for after. Just like you would with a software, if you get a free trial, you still give your payment information first. Do that here as well.
1:28:42The reason for that is that they need to make a commitment, and they need to make a commitment on that call. And you can even write it in the contract that if they then no show next call, that you then charge a $500 fee for wasting your time. Like, you can write that into a contract, but you always want that payment information on file, and the way you would usually do that is with Stripe.
1:29:00And what we usually do is that we just get, a $1 payment link. Inside of Stripe, if you create a $1 a year payment link, what that does is that it saves the payment information.
1:29:10You just say that's like a $1 down. We'll refund it so you don't worry. But then you have that payment method on file, which means that you can charge it in time, and that also removes friction later on, because now you have that payment method on file.
1:29:20So let's say that you're doing a commission deal or you're just charging $2,000 a month, you can just use this payment information to set it up. They don't have to do anything on their side, which again removes friction. So you always want to make sure to take some kind of commitment on the call.
1:29:33Even if you're doing a free trial, get the Stripe $1 payment link. Let me actually show you how to do that. You go inside of your Stripe dashboard, then you click on payment links right here and click create payment link, then you create a new product.
1:29:45You just call it checking card, something like that. You do it recurring, and you do $1, and then you do a year like this, and then you click add product.
1:29:55You don't need to collect tax automatically in this because you're gonna be refunding the payment anyways. Create link. And what will then happen is that when someone signs up on this link, you get this payment link right here, which says one dollar a year.
1:30:06When someone fills out the information right here, you get their email and you also get their payment information, and then it allows you to later to just create subscription for whatever you're charging. Let's say you're charging $2,000 a month, you can then go in and set up that payment without involving them.
1:30:19So that's how you create Stripe payment links, and trust me, that's gonna help you get commitment, which is gonna help you close more deals. If you have a problem closing, you're using this format right here, and you get to the point where you mention the price, and people don't buy it, then use this last tip, which is kind of a bonus tip, because it's not always required.
1:30:36Sometimes you have such a good offer that this is not needed, but I call it the proof of concept offer.
1:30:43What is this? Well, this is an offer that has made us tens of thousands of dollars. Here's how it works.
1:30:47It's basically a satisfaction guarantee, and you can say this that, well, this is our satisfaction guarantee. What you say is that you get payment upfront. You always want to get commitment.
1:30:55Right? So let's say that you are selling a website. Website, you still get your payment or whatever you agree on, let's say it's a thousand dollars first and a thousand dollars later, you still get your thousand dollars upfront, and you take that payment on the call via Stripe.
1:31:08But then you give a satisfaction guarantee, which means that I'm going to build this entire thing out. So if we use the website example, I'm going to build out the entire website for you, so you can see everything that's going to be on your website.
1:31:20You can see the finished product. If you're not happy with your website at that point, then you'll get a full refund.
1:31:26So if you don't even wanna see how it performs when we give it to customers and when we actually launch it, you can get a full refund. No questions asked. We have given this offer a bunch of times, and the great thing about this is that, one, it removes all risk, because they can actually make an informed decision, they can see exactly what they're getting, and if they don't like it, they can get a full refund.
1:31:46And two, it makes you seem very confident and good at what you do. All of a sudden, they say, okay, this guy can put money on the line, he will refund me an entire thing if I don't like what he creates. That makes you seem very confident, which again is gonna increase your close rate.
1:31:58We have given this offer so many times, and we have never had to refund anyone, because people have been very happy with what we built. So if you're good at what you do, this is such a no brainer offer, and adding this will probably just bump up your close rate with, like, 10% instantly. And you can use this for any AI build relief.
1:32:13It doesn't matter if you're building websites. It doesn't matter if you're building automations. You just wanna hop on a call with them and show it face to face.
1:32:18Okay. Here's how it works. Look at this.
1:32:20It does this. It does this. It does this.
1:32:22Get them to try it. Get them to see how good it actually is, and then say, great. Let's launch it.
1:32:27You don't even need to mention the satisfaction guarantee. Just mention it on the call when you close them, and they're gonna come back to you and say, well, if they don't want it, then they can get it refunded. But as soon as they commit and as soon as they see, okay, how good of a job that they've actually done, they're not gonna want the refund.
1:32:40They're gonna want to actually implement it. Great. So on the call, you only talk 20% of the time.
1:32:45They talk 80% of the time because you use the doctor approach. You use this structure right here where you start with questions, you position a pain, you then position that pain to your solution exactly, you handle any objections and you close the sale.
1:32:57You don't wanna just have a slideshow where you just explain your product, because you don't position a pain. They're gonna judge you on your price, not on your value, and they're not gonna feel like your solution is a fit to their exact problem. Then you always always book the next call on your current call, so they need to know when they're gonna hear from you again.
1:33:15You don't make the mistake of not taking payment on the call. You always take payment on the call.
1:33:20You get a payment information. You get the $1 Stripe link, and if you're struggling with closing deals, then use the proof of concept offer, because this removes all risk. You still get commitment upfront, but if they don't like the bill afterwards, you basically just have a satisfaction guarantee, where they can get a full refund.
1:33:34Awesome. After the call, then this shouldn't come as a surprise, but you need to do a good job. When you're just starting out, your reputation and building trust and building up a client base is absolutely everything.
1:33:44Make sure that you do a hell of a job, that you go all out for this client and do a hell of a job. Number two, a mistake that many beginners make as well, is that they don't continue the calls. Let's say that you have set up an AI system.
1:33:56A lot of agencies, they just set it up, and then they never communicate with the client again. That's not a good way to do it. I recommend at least biweekly calls.
1:34:05So hop on a call with them every two weeks, go over performance, go over things that could be improved. Why is that? Well, that's gonna decrease your churn a lot.
1:34:13Let's say you're charging $2,000 a month, doing continuous calls biweekly will probably decrease your churn to something like 5% or something, where many agencies have upwards of 20 to 30% churn. If they have 30% churn, it means that a client on average only stays for three months. There's another reason to why we do biweekly calls.
1:34:29This is two steps that almost every beginners they miss. They don't upsell. Let's say that you've just helped someone build a very, very cool website.
1:34:36You don't stop providing services there. Then you ask, well, do you also want an AI to pick up the phone calls from the people that call you now because now you have a good website? Or, well, just helped you build this website.
1:34:47Do you also want to send additional traffic with Google Ads to this site to get more customers? You always keep upselling, and if you don't have it already, then come up with ideas that could help their business further. Upselling is a way to increase your lifetime value, also called LTV, which is the total amount of money that your client pays you over their lifetime as client.
1:35:06Upselling is the best way to increase LTV without increasing the cost of acquiring a client. You've already acquired this client.
1:35:13It doesn't cost you anything more just to hop on a call with them and say, well, do you also want this? And then charge for that as well, of course. And it doesn't stop there.
1:35:20And, again, another reason to why we do biweekly calls. You also want to ask for referrals. Referrals is another way to get more customers without paying for anything.
1:35:28You're not paying for additional cold emails. You're not paying for additional connections on Upwork. This is free clients if you just ask for referrals.
1:35:35So many beginners when they start out don't ask for upsells, and they don't ask for referrals. That's also because that they don't do it the right way. Here's how you do it.
1:35:43You don't say, do you know anyone that could use this as well? Instead, you're very specific. You want to mention the exact people that you want to work with, like, do you know any plumbers in your area that could use this as well?
1:35:53Or do you know anyone in this industry or in this industry that could use this? Like, do you know any plumbers, electricians, people that own a landscaping company?
1:36:00You basically just name a list of potential clients that you want to work with. The reason that you do this is if you just ask, well, do you know anyone else that could use this service? They might come with someone, but usually they're just gonna say no because it's harder for the human beings to just come up with something from scratch.
1:36:14If you give them a specific list, like, you know any plumbers, electricians, landscapers, do you have any other friends in business?
1:36:20If you already give their brain that to think about, usually they're gonna come up with more referrals to you. That's rule number one when asking for referrals.
1:36:28Rule number two is that you always wanna give a commission. If someone refers you to a very, big company, you can give them 20% of your revenue. That is fine.
1:36:36You're gonna get way more referrals if you actually incentivize people to refer you, saying, well, you're gonna get 20% of whatever they pay me. I'm gonna pay that to you.
1:36:45Do that to all your clients, and you're gonna get way more referrals. I hope you see how all of these are connected. You do a hell of a good job.
1:36:52You need to make the client happy, and a part of making the client happy and treating them well is also to give them the time every two weeks. And only if you do a good job, and they also want to refer you to their friends.
1:37:03It all stems from actually being good at what you do and doing a good job. Great. I hope you found this section of the course valuable.
1:37:07I would honestly take all of these things, write them down so you have them and so you remember them, look over them daily, because these rules are literally gonna save you thousands of dollars, if not tens of thousands of dollars through your AI agency career. Alright. The next part of the course is also very important, because that's how you actually do the service delivery, and I have an insane trick that's gonna save you so much time prompting Claude and basically have the product delivery ready before you even leave the call with the client.
1:37:31It's insane. Let's get into that. What most agencies do when they are done after a call is that they get to work, and if you're using something like Claude Code, you're then taking the information that you yourself learned from the call and you're passing that to Claude Code, right?
1:37:47You're spending time giving it all the context that you gained on the call. But human mistakes happen. You might have missed something.
1:37:54You might have forgotten something. Sometimes you don't start on the project right after the call is finished, which means that you might lose out on some information. Let me show you both a smarter way and a way where you never miss information ever again, where you pass all the context effectively to Claude.
1:38:08To do this, we're gonna go back into the Claude app that we installed earlier. If you don't have it already, then install Claude for desktop. You go to this settings tab right here, then you go to connectors, and now they move that, so you're going to customize.
1:38:21That's going to show all your MCPs right here. Click this plus right here and click browse connectors, and then search for Fathom or the notetaker that you are using. On my calls, I'm using something called that's called Fathom, which basically records my calls, and it gives me a transcript.
1:38:36This makes it very, easy for us to just say, hey, Claude. I just finished the call. Go in and check the transcript and get all the context.
1:38:43I'll recommend that you click this from needs approvals always allow, and now the Fathom MCP is connected. So if we go down and restart Claude, because every time you can install a new MCP, you need to restart Claude, and we then write slash m c p, you can now see that the Fathom MCP has been connected, and if I click in on it and I click view tools, you can see we can get list teams, list meetings, find person, and these two are the probably the most important, get meeting summary and get meeting transcript.
1:39:10So let's test the connection. Let's go in and find this call that I had with the team on March 6 and say, I have this call.
1:39:19Please use Fathom MCP and give me a summary.
1:39:24There we go. You can now see it says composing. Now you can see it's calling Fathom right here.
1:39:29And there we go. Now we get a summary of the meeting. And we can get all of the context, and we can ask specific questions to pass by the meeting, how we build this, how we build that.
1:39:37This is extremely useful if you're just starting out, because then Claude have insights into everything and can help you build anything that you want that you say on the call that you can build. So if a client wants a website that can do some specific thing, you can have Claude read through this transcript, build the entire plan for you, and then build everything out from you directly based on the call.
1:39:56Very, very useful to know. You should use this for every single meeting that you finish. Alright.
1:40:00Let me show you how to do the service delivery so you can start building cool stuff with Claude. If you do a small Claude search, and we basically just ask how many businesses in The US are still missing a website, what we will find is approximately 30% of US small businesses don't have a website in 2026. Right now, there's approximately 36,000,000 small businesses in The US, which means that we have around 10,000,000 businesses without a website.
1:40:25And that's also why I previously showed you these three levels of things that you can build right off the bat, where website is level one. The reason for that is that what this entire business model is about is that you learn skills, you learn how to do something well, where businesses would rather pay you to do it than do it and find out how to do it themselves.
1:40:43And building websites is one of those things that still have a lot of perceived value. If people see a very nice website, they think, okay, that must have cost a lot. Even to this day, where AI is so good at building websites, there's a lot of perceived value in websites.
1:40:56And that is also why it's what we're gonna start off with building. Here's an example of the website that you're gonna learn how to build, and I'll even give you this template right here, so you can actually build exact websites like this from one prompt. You can see we have this nav bar right here that when we scroll, that becomes like a sticky nav bar at the top.
1:41:14I really like this design, the choice of colors, the choice of fonts. It looks very, very professional. And what you'll also be learning in this next part of the module is how to also make these websites actually high converting, because that's one of the places where a lot of website designers, they mess up.
1:41:29It's because they create websites that might look good, but they are not high converting. You can see there's a very bright button right here that you would want to click on, and you also have a number right here that you can call if you wanna call them straight away. You'll also learn how to create these very, very clean animations directly with code and plot code.
1:41:44You can see this little plumbing animations right here from the pipe, and this calendar animation right here. All of these, you'll learn how to create and you'll learn how to prompt plot to get a result that looks this good.
1:41:55You'll also learn how to create these type of animations, which are kinda like these cards where they stack on top of each other, which looks very professional. Then you'll learn how to split this into a dark mode that creates, like, a very nice contrast. And at the end you're gonna have this form right here, and I'm gonna show you how you can hook this website up to any CIM that you want.
1:42:11This is a very nice website, and usually web agencies can charge a lot. If I create a new chat and I ask Claude, how much does web agencies charge for a nice looking website, the result will probably surprise you.
1:42:24Look at this. Depends on a lot what nice looking means, but rough market ranges. For a solo freelancer on a template where it takes one to two weeks, you can charge 1 to 3 k.
1:42:33If it's a small boutique, then it's 3 to 10 k. If it's a proper agency, it's 10 to 30 k. If it's a mid tier agency, it's 30 to 80 k.
1:42:43And if it's a ward tier, like a big web agency where they really spend a lot of their time on the design, they can charge 80 to $250,000. And if it's some of the very well known brands, they can charge $250,000.
1:42:57And you can even see for my agency, I recommend 3 to $10,000 for websites. So you can see how much you can actually charge. And, again, to set your expectations, you're not gonna go out and close your first website to $10,000.
1:43:08But what you might go out is close a website for $500, which would still be a massive win when you're just starting out. And we can build all of this from inside of Cloud Code.
1:43:16I'm gonna show you how to build it, of course. I have a skill that creates these websites in literally just, five minutes. I'll show you how you hook this up to any CRM, so the information that is captured when people click get an offer is actually being sent to somewhere, and I'll show you how to make your website responsive so it looks good on both desktop and mobile.
1:43:34And I'll also show you how you can host and deploy your website for free up to a certain amount of traffic. But usually, if you're doing it for smaller companies, the hosting would be free. You're gonna learn a lot in this next part, so let's get into it.
1:43:45To build clean websites like this, you're gonna be using the build premium website skill. And, again, you can find all of my Claude skills inside of our completely free community. Just go inside of classroom, then learning hub, scroll down until you find my Claude skills right here.
1:43:59I'll I'll also leave a link to this right below this video. This drive folder includes all of my clause goals. Because after you've installed this skill and you write slash build premium website, you can see this skill builds a premium animated marketing web site with React, Vite, and Tailwind CSS for an industry.
1:44:18Use when the user has to build a website, blah blah blah. So we're gonna use it and write slash build premium website. I want to build a website for a landscaping company.
1:44:28Let's just call it Cityscape. Then we hit enter, because what this skill will now do is that it's going to force Claude to ask follow-up questions about exactly what side it is that we want, so the colors, etcetera.
1:44:39There we go. Now it's going to ask us what the theme should be. I think a bold modern theme looks good.
1:44:44The brand color direction, urban slate, mast, forest rust, olive brass. I think this dark green one, masked green, could look pretty cool, so let's go with that. Let's go with design, hardscaping, lawn, trees, lighting, commercial for the services.
1:44:57I think that's pretty good. And then I'm going to ask it to build it in this folder. We are currently in.
1:45:06I'm going to write submit answers, and now it's going to build out a plan for this website. You can see it's reading all of these files from inside of the skill that it uses as a reference for how it should design this. Building this out can take a couple of minutes.
1:45:18And there we go. That took eight minutes and forty one seconds, And now I can see site is live at localhost five one seven three. So I'm going to copy this, then I'm gonna go into our browser, paste this in, hit enter, and there we go.
1:45:34Now we have it. Cityscape, home services approach process contact, then we have request a quote right up here.
1:45:41Landscape architecture for the city, we design, build, and maintain, and now we have a clean website in literally just eight minutes that's based on this template. And you can see we have this animation right here with these stock images, which looks extremely good.
1:45:55We're missing one right here, so we need to fix that. Every discipline under one roof from a single tree pruned to half acre estate, plant from scratch. This looks extremely good.
1:46:05Request a quote. When they click on this, they scroll to this section right here where they can fill in this information. They can even attach photos if they want to, and then they can send the inquiry.
1:46:14Absolutely insane website. And you can see this website right here is built from the same template that we used for the plumbing company. I'm really happy with how this side turned out.
1:46:25We can also right click and click inspect, and then turn it to mobile right here, and then we'd see, okay, it also looks good on mobile, which is also very important for websites these days.
1:46:36Great. You've learned how to create insane looking websites, and you can sell this for probably around $500.
1:46:42But let me show you how to also deploy it. You deploy it by first going to GitHub. If you don't have GitHub already, then create a profile.
1:46:50Then click new, and I'm going to call this cityscape right here. You're gonna set it to public because you don't wanna give everyone your code to your website, and then you click create repository.
1:46:59There we go. And now you can take this URL right here, and you can pass it to Claude code, and you can say push this code for the website only to this repo.
1:47:10Then you don't wanna have it in auto mode, because it won't be able to push to repo in auto mode, so you're just gonna set it like this, hit enter, and now Claude is going to push all of this code into the cloud into GitHub. And you can see as first checking, hit enter. When you do this the first time, it pop up with an authentication where you have to authenticate in order to do this.
1:47:29It'll basically just open a window on your GitHub where you just have to click accept. But because I've done this plenty of times, it already has a connection. Now it is creating a commit, and you can see pushed code is live at this GitHub URL right here.
1:47:43If we then hit enter on GitHub repo, then we can see our code from this website. And hosting it is actually extremely easy when we have it on GitHub. I'm gonna use the platform that's called vercel.com, which is a hosting platform.
1:47:58Then if you don't have it already, create a user. It's free. Go to the top right corner, click add new, and then project, and then you can paste in this GitHub repo right here when you have your GitHub account connected to your Vasell account.
1:48:12This makes it very, easy because Versal can now see automatically, okay, this is a wide project. It's called Cityscape.
1:48:21And all we have to do is just click deploy once, and there we go. Now you can see it's being deployed.
1:48:26This should take usually around something like thirty seconds before the website is deployed. This actually only took nine seconds this time. You can see it says, congratulations, you've now deployed a new project.
1:48:36We can then click continue to dashboard, and here we have our project. We actually already have a domain that's given by Versal. If we copy this domain right here and we open it in a new tab, this is the domain that we have been given, then we actually have a preview of what the site will look like when it's live.
1:48:51And it's actually live right now. We can go to this website from anywhere, and you can see it looks extremely good. It looks exactly like it did on development.
1:48:59If you wanted to add a domain, you would come down here onto domains on the left hand side, and you can see we have our standard domain right here. You would just click add existing, and here you can add any domain that you want from GoDaddy, Namecheap, from Cloudflare.
1:49:13It doesn't matter. It would take you through the review process of setting up the DNS, and when you've done that, then you have deployed your website to the web. If you want to sell these kind of websites, I would create five different templates, and then create a portfolio website where you show these websites that you have built.
1:49:28That makes it much easier for a potential client to make a decision, okay, should I work with this company or not? And if you're using the method that I showed previous in this course, which is signing up for Upwork, then you can find plenty of work. For example, this one right here, posted yesterday a website for a construction company, logo and website design specialist, website design and launch, looking to make a website.
1:49:49There are plenty of jobs in doing this, and this is one of the best ways to land your first couple of clients. The great thing about this is that when you have built up a portfolio and when you build a bunch of skills like this, I'll show you later in this course how you can create something and then create your own skill you can replicate the process much much easier.
1:50:07Your work becomes almost automated. Cloud can do 95% of your product delivery.
1:50:12That's the entire point of this business model. It's to build stuff with Cloud, and then creating skills from it to replicate the process and making the service delivery so much easier. Alright.
1:50:22That is websites complete. The next thing you're going to learn is how do you actually automate things. If a client comes to you and says, well, I want my invoicing process automated, or I want my onboarding process automated, how do you actually go about doing that?
1:50:36How do you create automations with Cloud Code? Usually, you would use a platform like make.com or like nnn, but with Cloud Code, there's a much better way of doing it now.
1:50:46We can use the platform that's called trigger.dev. Trigger is an open source platform for building and hosting automations. You can see that it's open source right here.
1:50:55It has 15,000 GitHub stars, and we can see the entire code base right here. Why is it smart to use trigger dot dev over something like n n n or make?
1:51:04There's a very simple reason. The reason is that trigger dot dev is built with code, and Cloud Code is really good at coding.
1:51:11We actually already have a skill that's called trigger dot dev, and this skill is going to be very important when we build out those workflows. What I'm going to do is that I'm going to clear the session inside of Claude code, and let's build out a couple of automations with trigger dot dev so you can see how it works. I'm going to write I want to automate my invoicing process.
1:51:30I want to be able to fill out a form with a bit of information and then it should build the invoice, export as PDF, and then send via Gmail to the client.
1:51:46Use two skills and you will have access to both these skills as well. First one is the composeo skill. I want to use composeo for authentication.
1:51:59Skill two is trigger dot dev. I want to build and host this with trigger dot dev.
1:52:09I'm going to give it an ultra think, put it in plan mode, and let's build this out. Remember what I showed you before with the process of connecting Claude directly with your AI notetaker like Fathom via the MTP? Right now I explained what I wanted.
1:52:22Right? But if you're talking to a client, I want this and this and this automated, you can take that directly, that context directly from the call into Cloud Code, and then you don't even have to explain it like I did here. What fields should the invoicing form capture?
1:52:35It should capture it all, so business information, contact information, line items, tax and discounts, everything that we usually have on an invoice.
1:52:55It should create its own invoicing number and it should take the date automatically, then we hit enter, and there we go.
1:53:06It's now going to plan out the process of automating this process of creating invoices. For how we trigger the automation, I think we should use a simple web form from Next. Js.
1:53:15So I'm going to choose simple web form and hit enter. For the environmental value, it's very important that they're not hardcoded, but in a dot env file, please.
1:53:23Let's use react for the PDF, and for the invoice number, let's do a date based, and then we hit enter. If you know how to build with Cloud Code, it almost doesn't matter what it is that you're building.
1:53:33If you know the process of first planning it out, making sure that you actually give it the right context so it knows exactly what you want, then the opportunity is literally endless. You can build basically anything that you want if you know the actual formatting and the method of building stuff. Here's asking about the invoice count.
1:53:47I think it's a bit overkill to set up the invoice counter in a database, so I'm just gonna say let's skip the invoice counter for now and just do it date based, the naming.
1:54:03For the invoicing, I think both attaching the PDF and storing it in Google Drive is going to be best. Who writes the subject, fixed template with placeholders. Let's actually do it AI generated, so the email is AI generated, and then we hit enter.
1:54:17Now it's going to plan out exactly how we can do this. And there we have it. We have an entire plan for how we should build this out.
1:54:24What's worth noticing is that we are using Composure. If you don't know what Composure is, it's a tool for authentication, we so can authenticate to a bunch of different things very, very easily.
1:54:34I made a full guide for Composio on my channel, so if you want to, you can go and watch that later. And then it's using trigger dot dev for building and hosting this automation. The great thing about trigger dot dev is that it's open source, and when it's open source, we can host this anywhere we want.
1:54:50The easiest thing is probably just to host it with trigger dot dev, and that's how they make money as well. But if you wanted to, or if your client wants this on their own servers, then you can literally just build these workflows and host them wherever you want.
1:55:03That's the power of open source. So now we have a plan. Let's take yes and use auto mode, and let's build out this automation.
1:55:11There we go. Now we can see that it is built. Let me open another terminal.
1:55:15The first thing we do is that we need to fill out this environmental file right here. I'm gonna click new file, add add a dot in file, copy this example over, and then we need to fill out all of this information.
1:55:29So the first thing we need is a trigger secret key. If you don't have it already, then make sure to create a user on trigger and log in.
1:55:36I'm going to create a new project. So in the top left hand, I'm gonna create new project, Going to call this one invoicing and just click create. Now we can head on the API keys and then copy this secret key right here.
1:55:48So we paste that in where it says secret key. And then we also need to go in and find the project ref right here.
1:55:56Copy this one as well. Now we need the Composure API key, so we go to Composure, and Composure is basically what manages our authentication.
1:56:04If you don't have a Composure account, then sign up for one. It is free up to a certain amount of phone calls. Copy the composure API key and insert that as well.
1:56:14Now we need the Anthropic API key. This is used for the AI generation of the email. So we're going to platform.plot.com, go into API keys, click create key, Call this invoicing.
1:56:29Click add. Copy this key. Paste that in as well.
1:56:34We need the Google Drive folder ID. We get that by going to drive. We just create a new folder.
1:56:40I'm gonna call this invoices. If we click in on that, we get this ID right here in in the header. So I'm gonna copy this one.
1:56:47Then we write all of this information, which is our business info. So let's just try and see if it works. I'm going to write Test Street 123 Building 4 V business phone number.
1:56:59Just write in some random phone number. Business logo URL. We can go to our shiny website.
1:57:06We can copy this image URL right here. That way we don't need to host it, which is smart. Business bank details.
1:57:13Let me just write a random E band. Default currency, USD. Default tax rate, zero, default payment terms, net 14, default due days.
1:57:23This looks good. Awesome. Then we close down this dot environmental file, and we say, great.
1:57:28I have filled out the event file. What now? How do we auth?
1:57:37Now run the one time composure auth script. It'll print out two auth URLs. Get Gmail drive.
1:57:42You click through both, and it waits until each connection flips active. Me Let just create a new API key. Maybe it was an old one that's already rotated.
1:57:49I'm going to write try again. We still hit a bug. This is a new bug, so Cloud can just fix that.
1:57:54There we go. Now you can see it says Gmail OAuth URL. So it's going to give us this URL right here that I can copy and put into our browser.
1:58:02And this is how Composeo works. This means that we don't have to set up the Google credentials ourselves. We can use Composeo that just gives us this link right here, and then click continue, and you can see now Composure is connected to Gmail.
1:58:14I can now say done. Come with all URLs, please, so we can connect.
1:58:21And for each service that we connect, it's going to give us a new URL. So now we have the Google Drive URL right here. Paste that in once again.
1:58:28Log in with our Google account. There we go. And now you can see then open localhost three thousand and one.
1:58:34Server is still not running, so I'm going to ask, please run the dev server for me. And there we go.
1:58:41Now the two servers are started, both the back end, which is trigger dot dev and our front end. We can go to local host three thousand one, and now we have send an invoice, generates a PDF drafts with email with Claude and ships it via Gmail. From default currency USD, tax 0%, terms net 14, due in fourteen days.
1:58:57So I can write the client's name is let's do Albert testing, and let's do the company as let's just call it DovaAgent, and the address, we maybe don't need to I can do, like, King Street 420.
1:59:12Line items, we can do consulting services. One quantity. Price, let's do 1,500.
1:59:19We need to add another, which is upfront fee. Let's do one. Let's call that 500.
1:59:24Currency, money, and date. Currency with speed will be USD. Tax rate, zero.
1:59:28Discount type, none. Discount value, none. Issue date, let's do that the eighteenth.
1:59:33Due date, we can just leave that. Payment terms, net 14. Any notes?
1:59:39No. And then we can click generate and send. Of course, we need an email right here.
1:59:43Let me just write in my own email just as a test, and then click generate and send. You can see it says Qt run, run ID, and then this run right here. Let's see if it works.
1:59:52Let's ask Cloud now to check the logs. Did my test work? You can see it checked the logs and found a couple of issues.
1:59:59So let's let Cloud fix those. Keep doing runs, and you can see Cloud fixed one issue.
2:00:05Let's try and check the logs again. Ran into another bug. We'll fix that.
2:00:09Try and submit it once again. Check now. And we just keep doing that until Cloud has fixed all of the bugs.
2:00:15And you can see compose here won't use the ID. That's depending on the tool. Submit again.
2:00:21Let's do that. Generate and send. Just now.
2:00:24There we go. Finished success. Let's see how it looks.
2:00:28If we go to my Gmail and now hit a refresh, you can see we just received an invoice from me. Hi, Albert. I hope things are going well at Build My Agent.
2:00:36Please find attached invoice for $2,000 due by payment net terms of 14. So let us know if you have any questions, need anything from my end.
2:00:43Thanks so much for continued trust in working with Shiny. It really means a lot. And here we have an invoice that is generated 100% by our workflow.
2:00:50It has all of our information, invoice number, blah blah blah, like this and this and this. Net 14 payment details. Boom.
2:00:58If a company now comes to you and say, well, we spent a lot of time sending out invoices, you know how to fix it. We've just successfully built a trigger dot dev workflow without even being inside of a workflow builder or anything.
2:01:09Claude did it all. But let's say we want to deploy this. Right right now it's running from our computer.
2:01:14It's doing all of this from our computer. We can then write, let's deploy this to production on trigger dev.
2:01:22Thank you. The way that trigger dot dev works in the cloud is that you have these projects right here, and inside of here, have different environments. Like, right now, I'm in development, but we want to be on production.
2:01:31So now you can see a cloud is deploying this. It's pushing it into the trigger dot dev cloud. What we have to do is just set the environmental values.
2:01:39We do that by going into environmental values right here, and then clicking add new, then making sure that we have it on production right here. A quick tip is that you can go inside of environmental value, copy this, then you can go back and paste all of this in.
2:01:54We're just going to show all of the environmental values the right way. We can click save, and there we go.
2:02:00All of the environmental values has now been added. And there we go. Claude has now pushed our task.
2:02:05You can see our task inside of here now. Send invoice has now been added. We don't have any activity yet because we just pushed it to production, but we can run it again if you want.
2:02:14So let's test it. Before we test it, we need to take the API key from inside of production because that's going to be a different secret key. We need to go in and fill out this new information.
2:02:23So the trigger secret key, change that to prod, Same with the project reference right here. That's actually the same, so we don't need to do anything there. Let's test it now by running npm run dev.
2:02:34I'm going to write let's test it in localhost v thousand first form. I changed the environmental values.
2:02:44Let's try and test it now. So I'm going to write Albert testing two. Gonna send it to my own email once again.
2:02:49Going to call it build my agent two. And address we don't want. I'm just gonna call it testing for, let's say, 1,500 USD, blah blah blah, all of this we don't need.
2:02:58Generate and send. Now you can see it says sending. Queued run.
2:03:03If you now go into trigger dot dev and to our task, you can now see one run was just queued right here. We can also see run is now being executed.
2:03:13That's one of the good things about trigger dot dev as well, is that all of this dashboard right here where it's executing all of the runs, we have a very clean dashboard where we can see what's going on. And now we can see right here the run succeeded in thirteen seconds. If we go to our invoice and give it a refresh, we now have a new invoice with this new information, and our invoice generator is now deployed in the cloud.
2:03:35So that is the back end. Right? The back end is now deployed on trigger.dev.
2:03:40If you wanted to deploy the front end as well, we can also do that. We head to Versal, and then we write, great.
2:03:47Let's push the front end to Versal. First GitHub though. We're gonna go to GitHub, and just like we did before, we're gonna create a new repo.
2:03:58I'm gonna call it invoice front end. We can choose an owner. We're gonna choose ourselves.
2:04:02We can click private, and then we can create the repository just like this. Paste this into a cloud. Hit enter, and now it's going to push our front end.
2:04:11Our front end is this page right here. Right? The invoice generator.
2:04:15It's going to push that so we have it online. The crazy thing about this is that you don't have to remember, okay, how exactly did Albert do this? You just have to remember, you can use Composeo for authentication, trigger dot dev for the actual automation.
2:04:27When you tell that to Claude, Claude can figure out the rest. There we go. It is now pushed.
2:04:31We can now, just like we did before, to Vessel, click add new project, paste this in, click deploy, choose next, and just click deploy again, and that will deploy our project to the web.
2:04:43There we go. It is now deployed. We can click continue to dashboard, click on the domain right here, and now you can see the invoice generator is live on the web.
2:04:50The last thing we need to do is just set the environmental values. You do that by going in environmental variables right here, click add environmental variables, and just take the same one that we used before, paste that into the cell like this, click save.
2:05:02You're gonna have to redeploy, and now the invoice generator is live on the web. What's important to know is that right now we have hooked our Anthropic API key up to this invoice generator. It generates this text right here with AI when it sends the email.
2:05:16The issue with that is that if someone got access to this invoice generator, they could send maybe a thousand emails, which will burn a lot of credits on Claude. So you want to add some authentication.
2:05:27You don't wanna just deploy it like this, and in the next part of the course, I'm going to show you exactly how you do that, because you're usually not only gonna give them one of these, like, automations. Usually, you automate something within a business, you're automating maybe five, ten, 20 different processes that adds up to a bunch of time.
2:05:44This is just one of them. In the next part of the course, I'm going to show you how you can add authentication as well, so only the actual company can access this, and I'm gonna show you how you can build almost sort of a mini app with a bunch of individual small tools like this one, where all of the automations are in one place that you can then give to the client, which makes you look very, very professional, and a collection of these is something that you can charge $10.15, $20,000 for if you're doing it for big companies.
2:06:08There we go. Deployment has been created. If we go to it, now you can see it has all of our information from before, and now it will work from the web.
2:06:15Throughout this course, we have built a lot of things using skills. For example, the website that we created was heavily carried by this build premium website skill. This is probably one of the most important parts about this entire course.
2:06:26And the reason for that is that one of the most important skills that you want to learn if you wanna do this one person AI cloud business is to create a process, spend a lot of time building something out, like this invoicing, for example, and then building a skill so you can cut a lot of time if you had to build it again.
2:06:41So learning how to build generic skills that builds out a process exactly like how you want it will save you a bunch of time in the long run. And the way you want to do this is that you wanna use my skill that's called create skill. This skill is also inside of our drives.
2:06:55This one right here called create skill, which instructs Claude in how to create a good skill. I'm going to say, nice. Well done.
2:07:04This was a success. I want to create a skill for creating simple automations like this, where I have some kind of form and that is on Next.
2:07:24Js, that's what we used, and the back end is on trigger.
2:07:30Authentication, if possible, should be on Composeo.
2:07:34I want you to take what you learned from this session and add that information into a skill, so we can replicate it in in the future.
2:07:47Very important, use the create skill skill that you have access to. The user should in the start answer a couple of questions about what it is that they want to build.
2:08:04The skill should be generic for any mini automation slash tools that they want to build. So it shouldn't be invoicing specifically.
2:08:17In the skill, include that Claude should research before building.
2:08:26Then you want to hit it with an ultra think, you want to hit it with plan mode, and sometimes something I also like to do is to use sub agents to go into our current app as the example and get all context you need.
2:08:43This is something that I like to write this use sub agents because Claude is then going to speed up the process of finding context. Hit enter. And now Claude is going to build out a skill for how to create automations like this.
2:08:53This will take a couple of minutes. There we go. Now it's asking some question, like, what should the skill be named?
2:08:58I think mini automation fits pretty well. Should the skill bundle start a template directory? Yes.
2:09:03Where should it live? Let's do it in the personal skills, and then click submit answers.
2:09:09Because we use plan mode, it's now asking us if we want to accept this plan. I think it sounds good. So let's accept it, and Claud is now going to build up this skill force.
2:09:18And there we go. After a couple of minutes, the skill is now built. Something that you always want to ask is, is this a standalone skill that doesn't reference the invoice app?
2:09:31Answer this, please. The reason for that is that sometimes Claude likes to reference files, but if you want it to be its own skill, we don't want it to reference this folder.
2:09:40We want it to include all the information in that one skill without needing any other context. But you can see it says, yes, fully standalone, no path references to any other skill.
2:09:51Awesome. Now we have created a standalone skill when we write a control c and then click Claude right here, and we now write mini automation.
2:10:01You can see we have a skill that's called slash mini automation. Build a mini automation in Next. Js form on the front end, a trigger dot dev background task on the back end with composure of a third party authentications, etcetera, blah blah blah blah.
2:10:14And now you can replicate the automation that we just created with that front end form for anything that you want, and you wanna get a habit out of doing this. So anytime you build something that you think, well, I might build this for another client in the future, please make a skill for it.
2:10:28It's going to save you so much time, and especially if you're building out websites and you want, like, a couple of different templates to choose from, building out a custom website and then making it into a skill is going to be so valuable for you. I personally believe that in the future with these cloud businesses, the value, the IP is going to be the SOPs and processes that cloud knows.
2:10:48Just like in a regular business now, what's really valuable is all of the documents that businesses have that explains how they do things. The value in a cloud code company like this will be the processes, the skills that Claude has access to, the information that you have given it. So please make sure to make those skills, they're gonna become valuable and save you a bunch of time in the future.
2:11:07Make a habit out of it, it's one of the most important parts of running a Claude code business like this. Alright. Now we have gone over how you build websites for clients, how you also build automations, and how you can include AI into that to build agents.
2:11:19The third thing we're gonna cover is kinda like the next level to that, which is how do you then build out full AI systems? How do you take a business and then automate as many processes that you can, and then deliver that to the client? That's what we're gonna cover now.
2:11:31A full AI system is basically just a bunch of automations combined. So just like the invoicing automation that we just created, when you provide a full system, you usually automate maybe, like, five to 10 to 25 depending on depending on how big the project is. So how do you go about delivering this?
2:11:48This varies from agency to agency how you actually do this. Some agencies just deliver a bunch of in it inflows, where if the client wants, they can look at the executions of the in it inflows, but the reality is that the client is never going to do that. So the client just cannot trust the agency that it's set up correctly, but that is how most agencies used to do it.
2:12:07When we are building with Claude code, we want to give Claude access to everything as well. So we'll do this a little differently. We're gonna be building a dashboard for the client so they can see exactly what's going on, exactly how many executions, maybe even exactly how much time these automations have saved them so far.
2:12:23That also allows us to provide them with tools, like we just created with the invoicing tool, where if they have some process that requires, like, something to trigger it, like a form, for example, like we just showed, you can put those tools inside of this system as well. This also allows us to have a front end where they can see all of this, and then also have a separate back end where everything is running.
2:12:46And depending on the size of the project and the client that you're working with, sometimes the client wants to run everything on their own servers. And this system is great for that, because the front end, the actual visuals, and the back end is all going to be running as code.
2:13:01And specific to the back end, we are using trigger dot dev, exactly like we did in the previous automation, and because trigger dot dev is open source, it means that we can host it anywhere. We can host it on a virtual machine, like a machine that's just in the cloud.
2:13:15We can host it on the servers that they are already running if they're using something like Azure, which is the most normal one that is like a cloud provider, it's called, or if they're using AWS, Amazon Web Services, or if they're using Google Cloud.
2:13:30Having a system like this means that you can basically deploy it anywhere you want with a very nice user experience. Let me be clear from the start. Sometimes it's overkill building out a full AI system.
2:13:41Let's say that you're delivering a project for a client, and all they need is just, one automation done. What you could do is literally just provide the tool like we just did, connect it to something like a Typeform where you don't have your own front end.
2:13:52This is for bigger projects where you are auditing the entire business and then say, okay. These things are what we can improve, and I'm going to build this out like this and this and this, and you're basically get your custom dashboard, custom app where you have access to all of this.
2:14:03Alright. Let's get into how we actually build this. The first thing we're going to build out is the structure of what is really going to be a full app.
2:14:11We're gonna have the front end, and this is going to be very similar to what we already build. It's going to be Next. Js as the framework.
2:14:17It's gonna be Tailwind CSS for the design. And then I like to use something that's called Next. For the authentication that works well with Next.
2:14:25Js. And then we're gonna use the reset together with NextAuth, and reset was basically the service that we can use to send emails like magic links, and this is the most secure and easiest way to set up a login. That way you don't have to manage passwords, you don't have to manage usernames.
2:14:39All you're doing is just sending a magic link every time someone wants to log in. They click on that link, and they are in. So it's basically like forcing a two factor authentication every single time, which is the most secure thing that you can do.
2:14:49Then we have the back end. This is where all of our automations are running, and we're gonna be building that in trigger.dev. We're gonna start by hosting it on trigger.dev, but just know later on if you wanted to host the back end yourself, if you wanted it on something like a virtual machine, or hosting the back end on the client's actual servers, then trigger.
2:15:07Dev is the perfect thing to build with, because it allows you to just take the code and just host it somewhere else as well. And then because we're building out kind of like a full app, we also need a database. For the database, I like to launch fast, and the fastest way that I've found to launch is by using mongo which is a database provider.
2:15:27And the database is where we are basically gonna be storing all of this data, so all of the invoices, for example, if we have an invoice generator, all of the client's information, like the email and user information. And what's worth knowing as well is that we also gonna be hosting MongoDB in the cloud just to launch fast.
2:15:43If you are at one point wanting to build this entirely on the client's platform, then you can also do that because MongoDB also has a community edition that allows you to self host it just like you can with trigger dot dev.
2:15:54And while it's not entirely open source, community that edition is fine for what you're building as well. So that is kind of the stack that we are gonna be using, and don't worry if you don't know what all these things are yet. After watching this part of the course, you'll know exactly how to set up full AI systems like this.
2:16:07Alright. Let's get into it. The first thing we're going to do is that I'm going to build a new folder right here.
2:16:12I'm going to call this one let's just call it AI system, and then I'm going to go out of Claude and write CD AI system, which is basically just gonna get me into that folder, and then write Claude.
2:16:25The reason I do this is because I don't wanna give it the context of all the other things that we have built as well. We want it to be completely focused on this right here. What I'll then do is that I'm going to write, I want to build out a full app for a client that includes a bunch of small automations.
2:16:45And then I'm going to write the only domain that should be allowed into this app should be add dot shiny AI domains.
2:17:00The reason we do this is because if we deliver this for a client, want to add some additional security measures that only emails with their own domain should be allowed into it. This is basically just an additional layer of security. Then I'm going to write the stack.
2:17:13Next. Js. We want to use next auth for authentication with no Google login, because we don't need that.
2:17:21We want to be using the magic links, but, again, this is something that you could add later if you wanted to. Then we want to use tailwind CSS, recent for magic links.
2:17:31For the back end, I'm not going to write yet. I'm going to write back end. Let's not create this yet, but I want to use trigger dot dev at one point.
2:17:43So just create the structure forwards.
2:17:49And then for the database, we want to use Mongo. I'm gonna give it some additional information. This app will be a dashboard for the automations they have running.
2:18:02So keep as much as possible in the front end, only automations at the back end.
2:18:11Only automations and processes at the back end. For database, we're gonna use MongoDB, and then I don't want it to build anything out yet.
2:18:20I don't want you to build out any automations yet.
2:18:27I want you to create the structure of the app. I want a dashboard that they land inside, a simple sidebar where they will have all of the automations and a settings tab where they can log out, etcetera.
2:18:51Please build this out. Then we're gonna use UltraFinq. I'm gonna put it inside of plan mode to force it to create the full plan before it starts building, and then we are gonna hit enter.
2:19:03It's gonna take a little longer now the first time that we are building it out, but just like we have done previously with the automations, what we'll later do is that we're gonna create a skill out of this, and I'm gonna give that skill to you as well, and then we can basically build out full AI systems like this in a single prompt.
2:19:16Which NextAuth version? Use the latest one. For components, we can use shared c n components.
2:19:22How should the Shiny AI domain restriction work? Only allow add Shiny dot AI domains. MongoDB persistence for auth.
2:19:29Let's use the JWT session tokens and submit answers. There we go.
2:19:33Now it's going to plan out this entire thing. And there we go. Claude has now written a b plan.
2:19:38I'm also going to tell it to please use the super powers skill and I approve of plan.
2:19:46Again, if you don't know what the superpower skill is, superpowers is basically a plugin that includes a couple of skills for, like, writing plans, writing skills, executing plans. It's basically forcing Claude to think like a developer, which is very useful when it's building out a full app.
2:20:02So now it's going to build out and scaffold this entire project. That took thirteen minutes to build out. Now we want to click into the folder where we have created all of this.
2:20:10We want to go into the dot n global example, copy this, paste this, and then rename it and just call it dot env dot local like this. Cool.
2:20:20Now we need to fill out all of this information. The first is the auth secret. To create this, we can use this command right here that it was so nice to show, which is called open s s l rand and then base sixty four thirty two.
2:20:32Hit enter. That's going to give us a random string of letters. We can paste that instead of the auth secret right here, just like this one.
2:20:40Then we have the next auth URL. For now, we're just gonna keep that at localhost 3,000, and then we go on the API keys, create an API key.
2:20:48I'm going to call this one AI system. It should have full access. Click add.
2:20:52Copy it. Go back. Paste it in right here.
2:20:54And then I'm just going to write shiny portal. That is going to be the sender. And for the email, I'm going to use this email that I've already set up, which is just gonna be this one.
2:21:03So it's going to be noreply@mail.clip.dev. This is the domain that I'm just gonna use because I've already set it up once, and we're using the MongoDB URI. We get that by going to MongoDB.
2:21:14If you don't have one already, then create a user. Create a database and a cluster if you don't have one already. I'm now gonna click connect, click drivers, and this gives me this string right here.
2:21:24What we then also want to do is that we want to click done right here, then go into database and network access, click edit, and then we can edit the password, also generate a secure password, copy this, and paste it in right here where it says DB password, Just like this.
2:21:39And the trigger dot dev key is not set up yet. We're gonna set that up later when we set up the automations and the back end. Now we can click update user right here, and now it should work.
2:21:48The last thing we need to do is that we need to insert the name of the database. It's not called shiny right here. It is called new c fi right here.
2:21:56It's from a pre previous project, so I'm going to fill this out as well. Cool. Let's try and run it.
2:22:01So we need to go inside of AI system by using c d and AI system, And then we're gonna write n p m run dev to start the dev server. Hit enter.
2:22:11This is going to open it up, and you can see it starts on localhost 3,000. Let's see what it build out. We definitely need to work on the design of this.
2:22:19Let me sign in with alibot@shiny.ai. Send magic link. There you go.
2:22:23Check your inbox. And then go inside of our inbox, alibot@shiny.ai. Click sign in.
2:22:29And there we go. Now we have our dashboard right here that shows total automations, how many active, how many drafts. Recent activity, nothing yet.
2:22:37We'll come back. Your automation dashboard, blah blah blah, and then it says Shiny up here. Then we have an example automation right here, but we're gonna have all of the automations here on the left side.
2:22:45And then inside of settings, you can see sign in to shiny, email, and then you can sign out. Great. This is just the structure of the dashboard.
2:22:53We definitely need to make this look a lot better before we do anything else. The way we do that is that we use the front end design skill. I'm going to write.
2:23:01That looks good.
2:23:04I want to give our app a makeover.
2:23:12Use the front end design skill. I want it in a linear light mode type of vibe.
2:23:20We should have both light mode and dark mode. Let's build this out.
2:23:27Finks. Ultra think. Hit it with a plain mode, so it plans out how we should do it, and then hit enter.
2:23:33I like to build the structure first where it doesn't care at all about the design first, so it builds it out the right way, and then later, we give it a makeover so it doesn't look this bad. That will probably take a couple of minutes to fix. Great.
2:23:45Now we have something that looks more like this, which looks good. But I can see we already have an error right here, so I can click on this little icon to copy the error, go back into Claw, paste it in, and say, please fix this.
2:23:59Thank you. And now it's going to fix this error right here. It's running a lint go.
2:24:06Let's see if it'll there we go. There it disappeared. If you write at Albert@Shiny.ai, you can see it says check your inbox.
2:24:13We can then go to our inbox, click sign in. Now it says something went wrong, which is weird.
2:24:19That's because I clicked on the one that was twenty minutes ago. You gotta click on the newest one right here, zero minutes ago. And there we go.
2:24:26Welcome back. Now it looks like this. And the light mode, this looks a lot better.
2:24:31There's some weird thing going on with the gradient. What I'll do is that I'm going to take a screenshot of this and say this title and other text has a weird gradient.
2:24:42That is too much. Please fix it. Thanks.
2:24:47I like this theme a lot better. And now we have both a light mode and a dark mode. It fixed this Albert text right here, but it still does that weird thing.
2:24:56I'm going to just take a screenshot of it again and say, see how in the sides it's less bright than in the middle?
2:25:04You don't want that. Please fix. Also, this light mode button up here, we probably don't want that.
2:25:11So what I'm going to do is that I'm going to say, nice. The light mode button top right corner, please remove that so we only have it inside of settings.
2:25:25Looks a bit brighter, but it still has that weird shadow on the left hand side right here. Let's try and write on the sides. There's a weird shadow.
2:25:31Can we please remove that? Actually, think it's because these dots in the background, they probably would need this shadow.
2:25:37So I'm going to write they should only have the effect on the background dots, not on the headlines and text and components, etcetera.
2:25:52You can see it's doing it on every single page, kinda like dark to the sides, but then in the middle, it's like bright. We want it to be white everywhere. There we go.
2:26:00Now it fixed it. So the text is completely white. I think this looks a lot better.
2:26:04What we're going to do now is we're going to install our first So I'm going to write, let's install our first app, which will be this invoicing app that also uses trigger dot dev as back end.
2:26:21I want you to, one, install this front end form to our AI system folder, and two, create the trigger dot dev back end.
2:26:36We're gonna call this the AI system back end right here inside this folder that will include the back end for this automation.
2:26:48Also, make it so we later can add more tasks.
2:26:54That's what it's called inside of trigger dot dev. So more alterations later. For now, let's add this one.
2:27:03I'm going to set it to plan mode and write use sub agents to get all context you need and implement this automation.
2:27:16So now we're gonna take the invoice app that we just built previously, and we're gonna install it into our AI system. And there we go. Now you can see inside of our app, the send invoice automation has now been set up.
2:27:27And now they have the same form that's now in this new branding where they can generate and send an invoice much faster. And then when you're building out this AI system, it will consist of, let's say, five to 10 different automations that are automating something, and all the employees would need access to this dashboard right here that would allow them to use the automations.
2:27:45The next thing I want to do is to handle the authentication. If it's not off yet, I want to add a off button to this right here. For this automation, we need to handle the auth.
2:27:57So please add if we don't have authentication set up when we trigger it, that a modal pops up where they can authenticate.
2:28:13When we create this standalone automation, we handle the authentication using the link inside of the terminal via PlotCode. But, of course, our clients will not be able to do that, so we need to add a nice button where they can just authenticate with whatever they're using. Cool.
2:28:26I hope you see the idea of how you can build fully custom AI systems, how you can make it domain specific so only people from this company can use it, and how you now have the structure to build anything that you want inside of here. Your automations doesn't have to be a form where then something happens.
2:28:41You can also set up automations that doesn't require anyone to do anything, where it just runs on autopilot. Let me show you how to do that too.
2:28:47But before we do that, let's make sure that this actually works. So the first thing I'm going to do is that I'm going to go to trigger.dev. I'm going to create a new project that I'm going to call AI system.
2:28:58Then I want to push this straight to production, so I am going to get the API key right here, the secret key, paste that into environmental value right here, and we also need to do that right here. Then we need to go down, and we need to get inside of general, the project ref.
2:29:14So I'm going to insert that here too. Close this down like that. Just like before, we need to set the environmental variables.
2:29:19I'm gonna go inside of here, click add new, and then go inside of our environmental file, copy all of this, and then paste this into the production like this and click save. And now I'm gonna wait for it to set up the connection so everything works, and then we're test it.
2:29:34Cool. Let's try and push this to production. Going to give it the project ref right here.
2:29:39I'm gonna say we need to set up the back end. I have installed all the n one trigger dot dev.
2:29:49This is the project ref. Can you please set this up so it works with everything authentication, etcetera.
2:30:03Thank you. So now we're basically just gonna push this back end code right here, which is the AI system back end. We're gonna take these tasks that are the trigger dot dev tasks.
2:30:13We're gonna push it to this new trigger dot dev that we just set up, so we have everything in one place. You can see there's no tasks inside of here yet. And there we go.
2:30:22I now deployed it to production. Now I can see the task is inside of our AI system on trigger dot dev in the production. If I can then just make a test, we can call it Elbert test.
2:30:33Company name, BMA. Just write consulting. Set it to $1,500,100 dollars and click generate and send.
2:30:39You can see it says there we go. Queued run ID. Run like this.
2:30:44And now if you go inside of runs, you can see it's now executing, and you can see invoice from Shiny. Here we have it. Cool.
2:30:50So now our send invoice is working. What I'll do now, just like I explained before, is that now we have the start. Right?
2:30:57This is a good starting point for every new client that you get on as well. This overview dashboard, where you have a company name and one automation is the perfect starting point for every new client that you sign on as well. So we want to give cloud this information.
2:31:09What I'll do is I'm simply just gonna create a new session, then I'm going to use the create skill, and I'm going to drag these in, and I'm going to write I want to create a skill that creates this kind of front end and back end system to when I close a new client.
2:31:30It should be a generic skill. So first, it should ask what the client name is, what the domain is for their emails, etcetera.
2:31:45It should gather all of the context first, then I wanted to build out this template with the front end and back end first.
2:31:57After we have built out the structure, we should not have any automations yet.
2:32:05We will then later build automations, but the entire structure should be there, so it's very easy to do.
2:32:16This skill should work standalone, should not reference these folders, It should have everything included with references, etcetera, inside its own skill.
2:32:35I'm to give it an ultra think and hit enter. Actually, before I do that, I'm going to write use sub agents to gather all the context for how this works.
2:32:46Thank you. I'm also going to write include everything, like design, how it works, domain, the stack, etcetera.
2:33:00Use sub angels to gather all the context for how this works. Thank you. There we go.
2:33:05Bit of a long prompt, but we're gonna use the create skill now, and again, this skill is inside the Claude skills folder, which is inside the free community. So to get this skill, just go inside of classroom, go inside of learning hub, and inside of Claude skills right here, you have all of my Claude skills in this drive folder, and it's completely free to get.
2:33:22There we go. Now it's going to start up building out this, and it's calling it new client system skill. This will probably take some time because it needs to gather all of the context first and then build the skill.
2:33:30But you really wanna learn how to do this, because every time you build something new for a client that potentially could be replicated in the future, wanna you build a skill around it. If you don't do this, you're gonna start from scratch every time you land a new client. When you do this, you're gonna make your life much easier.
2:33:43And there we go. Now you can see the skill is ready. If I go into a new Claude session, let's restart Claude right quick, and we then write new client system like this.
2:33:52It says scaffold a new client's full stack next j s 16 front end dashboard trigger dot dev back end worker from the shiny automations template used when the user says new client, onboarded client scaffold a new client system, etcetera. And, of course, I'm also going to drop this skill together with the rest of the skills inside of the Google Drive.
2:34:09Now I know how to create full agent dashboards, and you can, of course, customize this to however you want. So if you want graphs right here showing you the task, you can ask Claude to build it. If you want to add a specific automation, you can ask Claude to build it.
2:34:21If you want to add something inside of settings where they can add the users themselves, you can tell Claude to build it. But what I want to do now is build some automations inside of this dashboard to show you how that would look as well. Let's get into it.
2:34:33Alright. Let me show you how to build out some more automations inside of our own automation system. What I'll write is I'm going to take this folder and this folder right here, which is basically all the files that this system includes, and I'm going to write read how this works, use sub agents, and then I'm going to say specifically look at how the automations are created, etcetera, then return to me.
2:34:59We just wanna give the context for how this AI system works before we then try and build out a new automation inside of this session. So it'll probably take a couple of minutes. Alright.
2:35:09So let's automate some more business processes. Let's say the onboarding, for example. A lot of businesses right now, they do manual onboarding, so they send out emails manually, they send out the contract manually, they do all of this stuff manually.
2:35:22We can create an automation for this as well and display it inside of here. So we can now write, I want to create a new automation that is called start onboarding.
2:35:33This should let's write up what it should do. Send over contract via Gmail using composure, of course.
2:35:41Send another email with a Calendly link where they can book an onboarding call, send an email with an onboarding survey, for example, send another email with Calendly link.
2:35:55So the connection we need is probably just Gmail. The user will provide onboarding link and Calendly link.
2:36:08Because it already now knows the structure of our system, this should be extremely easy to set up. I'm also going to write, please add this task to trigger dot dev, of course, like the other ones, straight to prod.
2:36:24And there we go. Now it's going to build out this automation just like the other one. There we go.
2:36:29That was extremely fast. We can actually see that building up this automation actually took only three minutes, and it was also a pretty similar automation, but we're gonna start here before we build something a little more advanced.
2:36:39But let's say that we just closed me. This is my email. This is my company.
2:36:45Then we can say send a contract URL. This could be something like a DocuSign, but we might wanna change this to an actual PDF. So maybe let's do that.
2:36:54Let's copy this and say, can we instead here upload a PDF on the new tool?
2:37:03Thank you. Then it's going to change the automations through that. Then we insert the onboarding survey and then the calendar URL.
2:37:08So the calendar URL could be something like this. The onboarding survey URL be something like this. And then we need the contract URL, but that's going to change.
2:37:17Now we can instead upload a PDF, but I actually want to generate the PDF instead. Can we actually instead generate that PDF? For now, we just generate a dummy contract.
2:37:30But when we generate, we insert the actual name and business information, etcetera.
2:37:38Please update. And, again, if the client comes back and says, well, I want this changed, so can we do this instead? Then you have everything in code.
2:37:44Vault can literally change anything inside of here. So it's fully customizable. There we go.
2:37:49Now we don't have a field to upload the contract, but now it's basically just gonna send out a dummy one. So let's test if it works. Oh, but a shiny AI BMA?
2:37:59Insert our onboarding survey. Insert our Calendly form. Send this in.
2:38:06Click start onboarding. There we go. You can see it queued this run right here.
2:38:10If you go into trigger.dev and look at runs, we can now see that it's now executing this run. There we go.
2:38:17It took twelve seconds. Let's see if I received something in my email. Now I get that book your onboarding call, onboarding survey, and your contract.
2:38:25Let's get started. This is all automatically sent. You can see we even generated this draft or placeholder contract that we can always change later to exactly what we want, but now it sends out this contract as well that includes my information.
2:38:37Absolutely insane. We just automated the onboarding process as well. And if the client says, well, we want a new lead updated inside of our CRM or we want something else done, they should just tell Cloud to build it, and it's going to build it up for you.
2:38:50But let's build out some more advanced automations. Let's get into it. Alright.
2:38:54Now you know the basics of creating these automations. But let's say that we want to build something a little more impressive, and this is usually also the things that you'll be sending to clients. For example, let's imagine that a client that you're working with, they have five support reps that are answering emails all day long, and you want to automate, let's say, our goal is 60% of that.
2:39:16So all of the questions that are the ones that are asked all the time, that an AI can literally just answer those questions almost instantly. We can also build a full AI agent system like this inside of our dashboards. The great thing about this is that it is custom code.
2:39:32It means that there are literally no boundaries. We can build whatever we want. I'm So going to go back into Claude code.
2:39:38I'm again going to tell it to read the structure of this full AI system, then return back to me.
2:39:49Use sub agents. Understand how it works, hit enter, it will gather all of the context that it needs, and then it's going to come right back.
2:39:57Great. It returns back after two minutes and has all the context, then we can write, I want to build another more advanced automation.
2:40:08This should be a full AI support email ticket response system that can respond to around 60% of all emails.
2:40:20In this automation, I want the front end to show the AI responses.
2:40:29I want it to show if any has been escalated. I want to be able to upload a full knowledge base, so it should use reg to search through it and answer questions based on this knowledge base.
2:40:47It's getting a little more advanced here, but luckily, do have cloud that can build up all of this for us. We will use Composio to answer the emails using Gmail.
2:40:58I want to trigger the trigger dot dev every ten minutes to check for new emails and then respond to every single one that has come in if the ticket requires human attention, we should be able to escalate it and see the escalated emails in the front end and be able to respond to them from within the front end as well.
2:41:35I also want to be able to see the AR responses inside the front end too. Let's see if there's anything else that we need.
2:41:44Let's plan this out. Please use the superpower skill and ask me if you have any questions.
2:41:55Let's fire away that prompt, and you can see it's gonna load the superpower skill and ask us any questions if it needs more context. We can all of a sudden start building some really cool advanced stuff based on the current structure that we have. How should the AI reply behave for the 60% it's confident about?
2:42:09It should also send immediately. Of course, if you're building this out for a client, you would, of course, have talked with the client and asked how it should do it, but you're gonna also send immediately. What format will the knowledge base come in?
2:42:20Let's do let's do PDF files. Let's actually just do text files. That is a lot easier.
2:42:26Well, we can actually turn on all of them to make the answers. Where should we store the vector embeddings for Rack? We could use something like Pinecone or Postgres, but we actually already have vector search inside of MongoDB, so let's use that.
2:42:39Let's do if any human asks for, like, low confidence from AI, angry tone and refund, and let's also do specific clarifiers like this.
2:42:49Submit. You can see it's asking all of the right questions that it needs in order to build out this plan. Which email account does this monitor reply from?
2:42:55One shared inbox. Yes. And there we go.
2:42:58Now I can see it's exploring the project to get context, and then it's going to ask a color clarifying questions, come with a color approaches, and then design the plan for building this out. How should we identify a ticket?
2:43:11What groups emails together? A Gmail thread is one ticket. That sounds good.
2:43:16When you escalate and reply from the front end, should the AI help draft the human apply? Yep. If we can get AI to predraft the reply but just not send, I think that could be pretty cool.
2:43:24Where do we store tickets, messages, KB chunks, and embeddings? Reuse existing MongoDB. Let's do that.
2:43:30Anything important I should add to the spec? Pig in and then apply. Pair ticket tags, labels, analytics view.
2:43:37Go to win analytics view and a tag. Submit this. Submit the answers.
2:43:43Got it. And after design, they'll be presented in sections. Confirm out of each.
2:43:46Section one. This is this. That sounds good.
2:43:49Then it's going to use trigger dot dev to poll every single ten minutes. This also looks good. There we go.
2:43:54That was all of the sections. Now it's saying writing out the spec docs now. That's probably going to take a couple of minutes.
2:44:00Now we have the full design spec, which looks good. So I'm going to tell it to continue. Now it's going to use this design spec to build out the entire plan so it can build this out.
2:44:09It's going to ask us if we want to use sub agent driven or inline execution. We definitely want to use sub agent to speed it up. This means that different agents are going to be working on different tasks.
2:44:20You can see I dispatch a fresh sub agent per task, review between task, fast iteration, usually the fastest and also what's recommended in this superpower skill. So now it's going to use the sub agent driven development, loaded this superpower skill, and now it's going to build out this entire thing for us based on our instructions.
2:44:39This will probably take a bit to build out because it is a pretty extensive AI system that we are building up. There we go. The next thing we need to do is that we need to go in and get our MongoDB secret, this one right here.
2:44:50Then we need to go to our production trigger dot dev, go inside of environmental variables, click add new, click on production, Make sure that we have added both the MongoDB URI and the DB name. Click save. There we go.
2:45:03Now I'm going to write we need this to update on prod trigger dev, so it will work. Can we please do that?
2:45:15Hit enter, and now it's going to push the changes it has made to our back end. It's going to push those trigger dot dev so everything will run smoothly. Alright.
2:45:22There we go. Find out spin up the server. You can see on localhost 3,000, we now have another automation that's called AI support.
2:45:30And here we have a couple of tabs for escalated support request, AI replied, human applied, and all. We also have this analytics tab right here that shows us how many tickets we have, and we have this knowledge base tab where we can either paste text into the knowledge base. We can add PDFs or URLs.
2:45:46Let's see if this actually works. I can take our shiny.ai website, for example, paste this in, click add to knowledge base. It is processing now.
2:45:55Let's see if it actually works. It says error though. So this didn't work for some reason.
2:46:00Let's try and see if we can fix this. I tried to add knowledge base.
2:46:07Ran into this issue. Can we check what happened? In the meantime, let's try and check the other things, like paste text.
2:46:15If I just, like, copy this entire page and paste it in right here, add to knowledge base, and call it basic info. Let's see if it wants to add this, but then give it a refresh again.
2:46:27Still says failed. So there's obviously something wrong with the knowledge base, but Talt should, of course, be able to fix this. Ah, now I see why it fails.
2:46:35We need an OpenAI API key. So I'm going to set this value as well inside our environmental value right here at the end. OpenAI API key, and then I'm going to go to platform.openai.com, log in with my Google account.
2:46:50Let's just call it AI system, create the secret key, copy it, paste it in right here. And then we also need to go into trigger dot dev inside of the environmental variables here. Click add new and edit here too.
2:47:02Let's do a refresh of this right here. Go to support. We're actually getting emails now that is escalated to us.
2:47:08No emails has been replied to yet. Just try and add the knowledge base once again. Probably our entire website right here.
2:47:14Paste it in. Call it basic info, add to knowledge base, now it is processing, refresh. Still processing now, so it didn't fail instantly.
2:47:21That's good. Let's delete these others while we are at it. And there we go.
2:47:25Now it says ready. So if we go back to inbox, we can see it has no subject line right now. So I'm going to copy this.
2:47:30I'm going to paste this in and say that seems to work. Issue now is that the subject lines don't show even though we do have subject lines.
2:47:42Can you fix this? Hit enter, and then hopefully, it can fix this subject line not showing.
2:47:48But this is an actual email from my inbox, which is kinda cool. Same with this one. This is also an actual email.
2:47:53So it's actually fetching the emails, which is kinda cool, and it is escalating them. Great. I think I also want to add a refresh button somewhere on this page.
2:48:02So I'm going to give it this page right here and say, can we add a refresh button here as well that triggers the polling, please.
2:48:13That way we can also do it manually just to get all the newest ones. And there we have it. Now we have a little poll inbox button right here.
2:48:20If we poll it, it says refreshing just like this, and then it shows us if we have any new emails in our inbox. So let's actually try it. Let me go on this email account right here, right to my personal email right here, basically write what does shiny yeah.
2:48:36I do. We can basically just ask, I saw your website.
2:48:41What is it that ShinyAI does? We can click send.
2:48:45And now I receive my email inside of my inbox. Let's see if we also receive it or if we hit a little polling right there. For some reason, it doesn't show.
2:48:53Maybe it's because it's under AI replied. Check the latest poll.
2:49:01I ran it after sending an email to myself to check it, but the email doesn't show up.
2:49:13Why is that? Check the latest logs. Let's see if I can figure out what happened.
2:49:17There we go. The issue was apparently that I clicked in on it, so it became read. Of course, we only want to take the unread emails.
2:49:24So, yeah, the system is actually working like it's intended. Cool. But it's still being escalated, so we need to fix some prompting, I think.
2:49:32I can write, see this email. It's a simple question, and we have it in our knowledge base.
2:49:43Why did it get escalated? Hit enter and see what it says. Alright.
2:49:49This was another one of these issue where it wasn't really an issue. I just had to refresh. But now I've added auto refreshes every thirty seconds inside of this tab.
2:49:57And if we go to AI replied and we click in on this, we can actually see, hey. Saw your website blah blah blah. And then the AI actually replied with this.
2:50:04Hi. Thanks for reaching out. China AI hills businesses boost sales.
2:50:07It's using the information that it got from the website, and if we click why this reply, we can even see the confidence score of how smart it thinks it was. We can see the citations that it used from our knowledge base.
2:50:20This is extremely cool, and it works. So now we basically have an AI inside of our support email that can reply to questions. When this confidence score isn't as high, it's going to put it in the escalator tab so we can reply to it.
2:50:32I hope you start to see the power of these insane automations and agent systems that you can all build inside of your own kind of dashboard right here and provide to a client. And the great thing is that as you build these systems out, you could then sell this support inbox to another client, and it wouldn't really take you that much work to set it up with a new knowledge base on a new client.
2:50:53And the crazy thing is the amount that you could sell this for. If you sold this to a big company, you could probably charge anywhere from 2 to $5,000 a month, depending on how big the team is.
2:51:03Because think about it. Right now, their support rep is probably being paid maybe like $2,000 a month.
2:51:08If it's in The US, then probably 3 to $4,000 a month. So if a team has, let's say, 10 support reps, and all of a sudden they only need five, you're saving them five times, let's say, thousand dollars, that's $15,000 a month you're saving them.
2:51:23If you're charging $5,000, you're still saving them $10,000 a month.
2:51:28And this is just one of the automation systems that you can build out. You can automate all processes inside of a business, smack all of these automations here on the left hand side, and you have a killer product, and you can manage all of this with Cloud Code. What I'll do is that I'll take the code for this entire dashboard right here with the trigger.dev back end, so these folders right here, AI system and AI system back end.
2:51:49I'm also going to take the cityscape website, so you have that as well as reference. I'm going to put the GitHub repost right below this video. The last thing I want to do is that I want to host the front end.
2:51:59Right now, we are still on localhost 3,000. The back end is hosted on the cloud.trigger.dev, but the front end isn't hosted yet.
2:52:07And it's honestly quite easy to do. It's the exact same thing we did with the other website. We just open vizel.com and click add new project, then we want to push this to GitHub.
2:52:16So I'm going to say I want to push only the front end, so that only this AI system folder right here, that's the front end, not the back end, only the front end to GitHub. Then we're gonna open a new tab.
2:52:30We're gonna go to github.com. We're gonna create a new repo. I'm going to call it AI system frontend.
2:52:37We're gonna make this private, create repository, then we can copy the link right here.
2:52:43We can paste it in and say push to this repo. You won't be able to do this in auto mode, so I'm going to set it to just like the default mode. Hit enter, and then we will have to accept the commands.
2:52:53It's asking us what to push. I'm going to ask it, I want to push this to separate repo so I can host the front end.
2:53:01So only the AI system. Standard push, it's it's empty brand new, and then we push it. Something that I always like to do is that I write off calls, don't push any secrets, etcetera, just so we don't push any of these secrets to GitHub for security.
2:53:18Always good to remind Claude not to do that. Sometimes it gets set up itself. It is running all of these commands.
2:53:23We'll just let it do that. There we go. Created a little mini plan for itself.
2:53:27Now it's going to push this front end repo to GitHub. There we go. Now it is pushed.
2:53:32If we go on to GitHub and then hit enter, now you can see we have all of the files inside of here. Let's host that by going to Wizzle, copying this GitHub repo, pasting it in, clicking deploy.
2:53:45It's going to see that it's a Next. Js application right here. Click deploy again, and let's see if we have any build errors.
2:53:51See, we already have a build error. So what I'll do is that I'm going to copy this. Better thing to do actually is to go inside of here where we are running it.
2:54:00And then instead of NPM run dev, we can write NPM run build, which is going to show us all of the build errors inside of this folder. You can see it compiles successfully, running TypeScript.
2:54:11This actually built successfully, so I'm not sure what this error is. So let's copy this, paste it in, and say when I try and host on the cell, I get this.
2:54:24Why? The front end installed MongoDB seven, but Pianza, if we should install both, the cell doesn't pass. Like, it's these by default.
2:54:31So clean fixes. Agree. Docker in MongoDB six.
2:54:33Let's do what it recommends. There we go. It pushed the changes to GitHub, so let's try and deploy it again.
2:54:40We can just click deploy once more, and now we should not get this error again. You can see it passed the first ten seconds, which was where we had an error before, and there we go. Congratulations.
2:54:50You now deployed a project. Let's continue to dashboard. We get this domain right here that we can go to, and you can see this will now be the login screen right here, which will only allow us to sign in with a Shiny domain.
2:55:01So I'm going to write my email, click send magic link, and we see something went wrong, which is because we don't have our environmental variable variables yet. So go inside of here, and then we need to fill out all of the environmental variables.
2:55:13Click add, pasting in all of these. Click save, and do a little redeploy like this.
2:55:19This will take another minute or so, and then it will deploy it. Let's click right here, view deployment, look at logs right here, build logs to see if it deploys like it should. There we go.
2:55:31Build completed. If we then give it a refresh and write in our email, now it should work. I can see right here that it's going to local host, and that's because I forgot to change one of the environmental variables.
2:55:42So let's go to next auth URL right here. You can see we're gonna be using this URL instead. Paste this in, click save, and click redeploy.
2:55:50Right now, we're using this domain right here, which is the domain that Versil gives us. But you probably want to add your own domain when you set up an AI system like this. The way you do that is that you just go inside of Versil again, go down to domains, and then just click add existing, and you can write in whatever domain you want and connect it using DNS.
2:56:07If have any issues with that, then you can just ask Claude. After you've changed the domain, then it's very important that you go inside of the environmental variables, just like I forgot to do now, and change this next auth URL to be the URL where your app is running. If you don't do that, you're gonna run into the same issue that we just did before.
2:56:24Now it's deployed once again. Let's hit enter. Write in Albert@shiny.ai.
2:56:30Now it is sending the magic of URL. And now we are in, and we have our AI support. We have our AI send invoice.
2:56:37We have our start onboarding. We have all of the processes that we built just before. There we go.
2:56:41Now you have deployed an AI system that could be for a client. Congratulations for reaching this far into the course. Great.
2:56:49Now you know how to build in these three levels. Websites as the easiest thing to build, then individual automations to automate stuff with trigger dot dev, and then how you can build a full app and dashboard around it that has multiple different automations in order to deliver a full project.
2:57:04What I'll show you next is my structure and my way of basically being able to build anything. Because now you know a couple of things you can build and deliver to a client. But if a client has some specific request and and you want to build something that you don't have a tutorial on, that doesn't exist in YouTube, how do you go about that?
2:57:22How do you build anything? I usually break it down into four different steps. The first thing before I even start building anything, we need to create a really, really good build plan.
2:57:33This includes things like the tech stack that we are using, how it will work, what database, what frameworks, what exactly is it that we are building.
2:57:43It also includes things like the design. How should it look? What should the feel of the thing we are building?
2:57:48What should that be like? And that is actually the most important step of building anything. And the reason for that is that there are so many things that goes into this step of figuring out, okay, what are we building and how do we build it?
2:58:00Specifically with the tech stack. Because if you're building something like an app or an automation, the tech stack is what determines what the cost of running it will be, and how fast it's going to be as well, and how well does it scale. And if we at one point later on wants to add more things, is that possible?
2:58:19Can we add more features? Can we actually do the things that we want?
2:58:23Can we give it the functionality that we need and want in our app automation or whatever it is that we are building. This build plan, what most beginners do when they're building something is that they just write, I want to build this and this and this, hit enter, and then take it from there. But what happens most of the time is that they end up with a bad tech stack.
2:58:42They don't go back and forth with the large language model in order to actually get the best tech stack possible for what they're trying to do, and they end up launching something that is mediocre, that makes them run into issues later down the line when it comes to scaling the cost, how fast there is functionality, etcetera.
2:58:59So before we start building anything, we want to have the tech stack dialed. We want to keep going back and forth with Claude until we get the best tech stack possible. And then we also want to nail the design so we know exactly what the feel should be of the user experience before we even start building anything.
2:59:16It's much harder down the line to change something from bad to good rather than just building it from good in the start. After we have done that, after we have spent a lot of time, and I'm literally talking hours here if you're building something complex, figuring out the tech stack, figuring out the design, and making sure that all of these will be good.
2:59:33Now it's time to build out the implementation plan. What people that are a bit more intermediate, but still beginners, what they do is that they maybe just make one plan of, okay, this is what I want to build, give it to Claude, and then have that build it out.
2:59:45And Fodo actually do this next step itself, which is building out an implementation plan, but you're gonna get much better results if you actually verify that you have a good implementation plan. An implementation plan is basically a plan for how do you actually build the thing that you have written out in the build plan.
3:00:01And the way I like to do them is split them up in steps. And for each step, I want a bunch of different check boxes. So to give you an example, step one might be to set up file structure.
3:00:12And in the file structure, there's a bunch of different sub steps in setting up the file structure, and you want to write all of this out in your implementation plan. And if you're building something very complex, you might have something like 40 steps in your implementation plan in order to build the thing out that you want.
3:00:28And before you even start building anything, you want to make sure that you have this implementation plan dialed. In step two, the next thing we might want to build out is maybe the database structure, and then maybe authentication. Plot knows the best sequence of how to build out anything, but you want to force it to make a good implementation plan.
3:00:46You want to force the AI to really make a decision, make a good decision for how to build out the thing that you want. And then after you've made the build plan, you've made a good implementation plan, that is when you start building.
3:00:58And the way you do that now is that you give the large language model this implementation plan, of course, and then you tell it, please start with step one. And the large language model is now going to start with this sub step, then do this sub step, and then do this sub step, and come back to you and say, well, step one has now been complete.
3:01:14Do you want me to start with step two? And then it's going to start step two, it's going to do the sub steps inside of step two, and then it's going to finish. This is much better than just giving it an entire plan or just giving it a build plan and letting it control everything itself.
3:01:29By this, you force it to do it in the best sequence possible, and you've planned everything out before you even start building anything. I hope you start to see the power of building out both a build plan and an implementation plan. And, actually, when it comes to, like, the time that you're building, step one and two will be where you spend 80% of your time.
3:01:48Step three when building something out, and actually building it will only take around 20% of the time. A lot of people in the AI space get this wrong, and they spend 80% of the time just letting the LLM build, and only 20% of the time planning it out, if they even do plan it out at all. So when you've made a build plan, you've made an implementation plan, the AI has gone through every single step all the way down to step 40 or however long implementation plan is, and the AI is now done building it, then it's finished, and you'll have to test and refine it a lot.
3:02:16You'll definitely run into bugs. It almost always happens, and you basically just go through them, paste the bug to Claude, and have it finish everything. And it's basically how you go about building something.
3:02:26In all of this, you want Claude to be almost a consultant that can explain to you what is the best tech stack, what should I use for building this, what gives me the best results. And a bonus tip that gives you even better results is if you use two large language models. It could either be two Claude code sessions, could also be one Claude and then one codex.
3:02:47But if you have one AI, it's kinda like the consultant that decides what to use, and then you take that decision and you give it to another AI model and you say, is this true? They're then gonna bounce ideas and make sure you get the best result possible.
3:03:00And by doing that, you can start with absolutely no technical ability, and you can figure everything out on the fly. Awesome. Let me actually demonstrate exactly how you do this by building something that I have no idea how to build, and I'll show you exactly how you use this four step framework right here to do the exact thing that we just talked about.
3:03:17Let's get into it. To show you that this actually works and to show you how to actually do it, let's build something that I haven't built before just to show you that it actually works and that you can build something really, really good without knowing how it works first. I want to build a lead enricher.
3:03:32What that basically means is that I want to be able to insert a name or an email or a LinkedIn URL, and from that I want to search up a database of some kind of sort, and from that I want a software that can go out and get all information on that person. So if I wanted to find someone's email, I could just paste in their LinkedIn URL, that would give them their email.
3:03:54I have a slight idea of how to build this. I can probably use some API from like Apollo or something, but I wanted to show you how you can know none of that and get all the answers from Claude in order to build it out.
3:04:06So I'm not going to help it and tell it what to do. Claude is going to make the decisions for us. I have a blank project right here.
3:04:12I'm just gonna head under the terminal and click new terminal, and then open up a Claude session right here and click, yes, I trust the author. Then going to drag it on top right here, and I'm going to write, I want to build a lead in Redshaw.
3:04:28First, I want to build a build plan.
3:04:33Ask me some questions about how it should work so we can create a build plan as a dot m p file.
3:04:45We just set it to auto mode and hit enter. And now Claude will ask us a couple of questions to basically build the thing that you want to build.
3:04:52And you can actually see it's even using the superpower skill, which kind of forces this entire process. This superpower skill first creates the build plan of what we want to build, which is basically what it calls a design spec, then and it actually also builds out an implementation plan.
3:05:05This is what's going on under the hook for the superpower scope. You can see it says what's the input to the lead enricher? What do you start with?
3:05:13I want to be able to start with anything, and from that so either email, company domain, name plus company, LinkedIn URL, and from that, get all of the information.
3:05:38So I should be able to insert what I have.
3:05:45Hit enter. What output fields do you want the enricher to return for each lead? Let's do person basics and contact info and company info, and then yeah.
3:05:55Let's get socials as well. Submit. Hit enter.
3:05:59Which enrichment data source do you want to use? This is why we want to make sure that we actually make a good decision. Choosing an enricher is the biggest cost of running this.
3:06:08And right here, it's giving an option to have multiple and use each as fallback. I like that idea, so let's do that. What do you want to interact with in richer?
3:06:16Let's do a small web app like this. How should leads be input into the web app? Both single bulk, one form from a single lead.
3:06:25Let's just do a single lead only for now. Do you want to store enriched leads or is it lookup only, store history?
3:06:32And if you reach a question where you don't know the answer, you can literally just scroll down if you're using the superpowers plugin. You can also just go out of it and then discuss it with Claude, but we can write chat about this and then say, what would you recommend?
3:06:48And then it says, I recommend store history in a database, reasons, saving on cost, blah blah blah. The only reason to skip storage is we have a strict privacy policy compliance reason.
3:06:57Let me just say for now that we don't need to store it. Now you can see it says, got it. Lookup only.
3:07:02How should the provider fallback chain work when reaching a lead? Sequential with merge? Try Apollo first, then Hunter.
3:07:08Let's try and just say let's just try Hunter first, and I want it to stop when it finds the email because that is the main thing.
3:07:22If there's something that you don't understand, something that you don't have an answer to that you think is best, then you just ask Todd what it would recommend. Here you can see it says, what order should the providers be tried in.
3:07:34It gives us a couple of options. I'm just going to say hunter only, but you can literally just go back and forth with Claude like this. What's the tech stack?
3:07:42Should I match the shiny automation template? It knows this from memory. Let's do next JS only and no trigger dot dev for now.
3:07:51I have enough to draft the design. Here's the proposal, architecture, blah blah blah.
3:07:55That's all of this. And now it's creating what's called the design spec, and the design spec is basically what I called the build plan in my previous explanation of how to build anything.
3:08:05You can see it has two questions, two things I like you to decide before I write the spec. How should I handle the LinkedIn in URL case? Let's for now skip skip URL, and let's do a drop down.
3:08:17The skill I'm using right here, which is also very valuable for you to know about, is called superpowers. And what superpowers basically does is that it forces Claude to think like an engineer. All you have to do is literally just search github superpowers.
3:08:29You're going to find this GitHub repo right here. It's called Ober Superpowers. Click on this and you can see it has 208,000 stars.
3:08:37All you do is that you literally just copy this right here and you go to Claude and you say, please install these skills, search up the documentation for installing global Claude skills before you do it.
3:08:53Did you just send this like this, and it's going to install the skills for you? After you've done that, you might have to restart claw just like this before it's ready. But I can see spec written and saved to docs superpowers, and then it has the spec right here and the design.
3:09:09There we go. So if you click this, you can now see it build out a full plan for us, and we want to read this carefully to make sure that aligns with exactly what we want to build. So you can see it says lead enricher, design spec, it shows the date, it shows the status, which is draft.
3:09:22Goal, a personal web tool for enriching a single lead at a time. The user enters whatever they have, email, company name, or domain name, and domain name, company, and the app returns the leads contact information, company info, and social presence.
3:09:34There we go. None goals. We don't need a CSV enrichment.
3:09:37This is not what we want. It shows the stack right here. It shows the user flow.
3:09:42It shows the architecture. Shows the component responsibilities, what each page should do.
3:09:48You can basically see we have a full plan already for exactly what it is we want to build. If we find something inside of here we that we don't like, we can tell Claude to fix it in the implementation plan. It even shows how the enriched lead should be structured.
3:10:01It shows error handling. It shows file layout, and it shows out of the scope, maybe for the future like this.
3:10:07Nice. Now we have our full build plan, and right now we're building a fairly simple tool. It's just like a personal tool.
3:10:13If you were to build out a full software, for example, then this would definitely be an extremely long build plan. You would have to include pricing and authentication and databases and maybe you're running some virtual machine that's like hosting the back end.
3:10:27All of this would be inside of the build plan or what superpowers call it, the design spec. And since that spec written and saved, I'll say I have read it now.
3:10:37Please continue. Now you can see it's using a different skill also coming from superpowers that's called writing plans. And this is where we go to the next step, which is building out the implementation plan with these sub steps.
3:10:49And building out this implementation plan is actually the thing that usually takes the longest. I'll let it do that, and then I'll come right back. And there we go.
3:10:56Now you can see the plan is complete and saved inside of plans, and this is what the implementation plan is. Right? If we click on plans, I'll click on this, you can see we have the entire file structure.
3:11:08We have task one, do this and this and this and this, set all of these things, step five, step six.
3:11:15So it's basically split up in tasks and then steps inside of these tasks exactly like this. So now we have two options.
3:11:24Do we want sub agent driven by each thing are built where a fresh sub agent is doing each task? This speeds it up a lot, or do we want to do everything in line? I'm going to do one sub agent, start the build, and only now after we've made a spec and after we've made a plan is when we start building it out.
3:11:43So now it's going to build out this little mini app for us, which is a lead enricher. And there we have it.
3:11:49Now Claude has built out this entire thing by following the implementation plan, and this took thirteen minutes. The next thing we want to do is testing and refining, and we are probably gonna run into some bugs. You can see we have our environmental values right here.
3:12:03Let me copy that, paste it, and then rename it to just local. And let's then update this hunter API key.
3:12:11Go to hunter.io and go to API. Copy this API key.
3:12:15Go back. Paste it in right here. Close it down, and let's try and run the server.
3:12:20So we can now write n p m run dev. That's going to give us localhost 3,000. Paste it in, and let's see if our app works.
3:12:26So it says lead enricher right here. What do you have? Let's say we have a name and company, and the name we can write this.
3:12:35Let's test it on myself first. Company name, shiny dot a I. Let's see if we can find it using this.
3:12:43Click enrich. Now we run into this issue right here. So I'm gonna copy this.
3:12:47This was as expected. Paste it in and say we run into this issue. Hit enter, and then we should be able to fix It was something to do with how it's passing the information.
3:12:57So let's see if it works now. So you can see it's making a bunch of updates now in the code. Now it wants to test it.
3:13:02We can just stop it here. Let's try and restart the server. Write name and company email again.
3:13:08Click enrich, and there we go. Now we have a bunch of information, and it could even, just from my name and the company name, it could find my email. Let's see if I just write shiny dot a I and click enrich, see what data it then finds, then you can see it actually finds Oliver, my cofounder, and his email.
3:13:23And there we go. The lead enricher works. It's not a very impressive app because it literally just uses the Hunter API, but it was also show you this flow right here of building something that you've never built before by using these four steps, creating the build plan, creating the implementation plan, and then having the LLM build for you, spending a lot of time on the build plan and the implementation plan.
3:13:42And then when you have all of that, when you have sharpened your axe, that's when you start building it. And I guess we can do the next step, which is testing it and refining it. We have already done a bit of testing, but let's refine it as well.
3:13:54I can give it the front end skill now and say, it look clean and modern like a modern SaaS.
3:14:04Hit enter. And now it's basically just going to give our entire software a full makeover of how it looks. That's a part of the refining step, actually making it look good.
3:14:13There we go. Now we have given it a bit of a makeover now. Find anyone's work email, and now we can choose either email domain name company.
3:14:20Works exactly like it did before, but now it just looks a little better. Awesome. The reason I wanted to build this out is just to showcase how you basically build anything.
3:14:29So when a client comes to you and tells you, I want to build out this and this and this. I wanna try and automate this. Is this possible?
3:14:36You can literally just ask Claude about how would I build this? How would I go about this? Use this framework right here for first creating a build plan, creating the implementation, then having it build out, and then basically allows you to build anything you really want or would never need.
3:14:50If you're already technical, then you might be able to skip some of these stages. Like, you might already say that I want to use this and this and this tech stack, because you already know that this tech stack would be good for the task. But if you're a beginner and you're just starting out and you don't have that much technical expertise, don't skip the step of just chatting with Claude and finding out what the best tech stack would be for things like scaling, cost, how fast it is, the functionality, etcetera.
3:15:14Go back and forth with Claude a lot about that, and you're gonna get the best results. The next part of the course, we're gonna talk about pricing, because now you know how to create a bunch of stuff. But how do you actually price it?
3:15:24How much can you charge? What pricing model should you use? That is what we're gonna cover now.
3:15:28Now you know how to build a bunch of different things, and you also know the framework for basically being able to build anything using cloud code. But how do you actually price your services? How much should you be charging?
3:15:41This is a question I get a lot, and I always use the golden rule that when it comes to developing an AI project, you always want to give your clients a five x return on investment. So let's say that you save your client $5,000 in payroll or in whatever solution that you implemented, you would be able to charge a thousand dollars from that for your service.
3:16:08If you keep this five x return on investment, your client will always be happy, and a happy client means low churn, which basically means that they stay with you for a very long time. If they're paying you monthly, you of course want them to stay as long as possible, and it increases referrals. If you did a good job for them, then they're gonna refer you to their friends that are also business owners that might also have the issue that you're solving, and this is where you potentially can achieve negative churn, which basically means that your agency grows without you doing anything.
3:16:39Just from serving your clients right now without paying anything else in marketing, your agency would still increase because the churn is so low and clients are referring you to other clients in a higher rate than the churn you have. And that all starts by giving your clients a good return on investment.
3:16:57That's the entire point of this business model, and in any service business, really. You need to go in and do a hell of a job, give them a return on investment in order to actually deserve the thousand dollars. But how do you actually calculate the return on investment for your services?
3:17:12It is basic math. Let's just use an example that you are installing voice agents, and right now you're working with a client that right now has 10 people in customer support, and they are taking calls all day.
3:17:24They're answering the same questions over and over again. So now you install your voice agent system, and this now means that they go from 10 customer support reps to, let's say, five customer support reps. This means that they have cut five customer support reps.
3:17:38Each rep might be paid something like $2,000 a month. So when you have cut from 10 down to five, that is five reps, that basically means that you're now saving this company $10,000 a month.
3:17:53So using the five x return on investment rule, that mean that from that, in order to keep the client very happy, you'll be able to charge $2,000 a month for that. And if you do it like this, then they have no reason to churn.
3:18:06You're literally saving them still $8,000 every month just from using your system. Of course, you need to do some work.
3:18:12You need to maintain it. You need to make sure that voice agents actually do a good job, but that is also what you're paid to do. That is how you can calculate a return on investment.
3:18:20You might also have a system that directly generates revenue. Let's say you have like a cold email AI system that you install for clients, and that system is generating a business, let's say a $100,000 a month in new revenue.
3:18:37When you do this, you need to calculate in gross profit, because this is revenue. Gross profit is what you have left after you remove the fixed cost. So things like payroll for taking care of the client, how much is the company actually profiting from that $10,000 worth of revenue.
3:18:52That might be, let's say, if they have good margins, like $40,000. So that is, in theory, what you're making them every single month. You can charge one fifth of that, which means if you're running this system right here, you'll be paid $8,000 a month.
3:19:06So you always want to give them at least a five x return of their money. And, course, this $100,000 a month in revenue that you're giving them needs to be new revenue, right, not their existing revenue already.
3:19:18So if you can, by installing your system, give them a five x return on investment, then they're gonna be very happy and they're gonna stay with you forever. But what pricing structure do you then use? If you want to make it really easy and if you wanna follow the the golden rule when it's available, when result based, when it makes sense, it's a very very good offer.
3:19:39If you literally just say to them upfront that you're going to be taking 20% of the gross profit that you make a company, that is such a no brainer, because there's no risk in it for them.
3:19:51You remove all of the risk. Now there's only an upside. And it also shows that you're very confident in what you're doing.
3:19:57The negatives are, of course, is that you need to be able to track all of this. So this usually works with larger clients where you can actually go in and control and see, okay, how much money am I making them, that you can follow each lead in the process, that they have a good CRM, etcetera.
3:20:13That's one of the really good offers, result based. And when you are good at what you do, result based will also usually pay you more than something like a regular retainer. Let me give you an example.
3:20:23Shiny.ai, which is our AI agency, does lead reactivations, and we always use result based.
3:20:30We had one client where we probably could have charged maybe, like, $1,000 a month on a retainer. But because we did it as result based, because we went after the golden rule, that client in total actually made us around $240,000 in LTV.
3:20:47So we probably at least 10 x the lifetime value just from doing it result based. Result based is one of those offers that you'd rather have a retainer if you're bad at what you do, but if you're really, really good at what you do, you would always rather have result based because you're paid for every result that you bring.
3:21:03If you can scale that up, your result also scales infinitely. I hope that kinda makes sense. But this offer right here is not always the best one to use because sometimes it can be hard to calculate how much something is worth.
3:21:15Let's say that you are implementing customer support widgets on websites. Right?
3:21:20So this little website widget in the bottom right corner. It's very hard to say how much a conversation is worth. You could still do result based.
3:21:27It will probably not be of gross profit. Instead, it might be per conversation, for example, that you're charging them a little something.
3:21:36Maybe you're charging them something like $1 per conversation. So this still definitely still add up if they have, like, 200, 300 conversations a month, but that's one way that you can do these kinda, support agents as result based.
3:21:48But you cannot really, like, calculate gross profit directly from, like, a support agent. So sometimes result based really doesn't make sense. Instead, what I recommend at that point is the upfront offer plus recurring, because sometimes it's simply, like, impossible to track, and it also becomes a headache if you have, like, 50 clients to track their performance every single month.
3:22:09So charging something upfront and then a recurring afterwards is sometimes the best option. But if you do that, I'll always recommend that you use the satisfaction guarantee. This has worked so well for us.
3:22:20Let's say that we want to charge a client $2,000 upfront and then let's say $500 a month. If we just give them this offer right here, there's a bunch of risk.
3:22:29Right? The client is taking a risk. Okay.
3:22:30Does this even work? And they're paying a pretty big amount right upfront without even seeing if it's going to work yet.
3:22:37Instead, what I want you to do is to give a satisfaction guarantee. So you still charge $2,000 upfront.
3:22:44You say, I'm going to build this out. It's going to cost $2,000 as a setup fee, but I'm so confident in my own ability that I can build this really well out for you.
3:22:54So after I build out the entire system, so if you're building a dashboard, you'll build out everything. If you're if you're building a website, you'll build out everything. You're basically going to say that I'm going to build out everything so you can see exactly how it looks, then we're gonna hop on a meeting, and if you don't like the result at that moment, then you can get a full refund.
3:23:12So then we have the satisfaction guarantee. The reason why this is so strong is because now all of a sudden there's no risk in it for them. If they don't like what you have built, then they can just get their money refunded right afterwards.
3:23:24But if you're actually good at what you're doing, which you hopefully are, then people are not gonna refund, because they're gonna see a website that they actually like, or they're gonna see a dashboard that actually automates the things that they want automated. It's very important sometimes for some projects that you get some money upfront.
3:23:39The reason for that is that it has happened countless of times for us that we have started the project without taking any money upfront, and then the client simply like stopped answering.
3:23:50Maybe he got cold feet, maybe something else came up, maybe he was just too busy, but the reason they could just like stop answering and don't care anymore is because they were not invested. You want people to be invested right from the start when you work with them.
3:24:04So this is like the perfect middle ground. You still get them invested. You still get money upfront, but you remove the risk by giving them a satisfaction guarantee.
3:24:11And if you're good at what you're actually doing, we have given this offer a bunch of times, and we have never had to give a refund because every single time we have shown someone a website, for example, or shown someone a system that we have built out for them, they have said, this looks good. I'm ready to try it. Awesome.
3:24:26Let's get started. And for a lot of the time, actually forget about the satisfaction guarantee when they see something that's really good. And then I'm a big fan of recurring businesses.
3:24:34I don't like to work on a project basis. I like having recurring revenue coming in every single month. So when you charge an upfront fee and then charge a lower recurring fee, this lower recurring fee is usually not a problem.
3:24:46And this really helps when you start to stack, like, 10 plus clients. All of a sudden, this recurring revenue is just stacking and compounding. What you basically justify is that this is to keep up support.
3:24:56It's to keep the system active, to fix anything if any issues come up. That is what this fee, basically like a service fee covers, right, which is a monthly recurring fee. Another thing that's very important to talk about is that you always, of course, want to make sure that you profit.
3:25:09You don't wanna build out a full system for them and then cover, let's say, the software cost yourself, and then realize that you that you're not going to profit. A way to overcome that is to use tiers, just like you would with the software where you have, like, tier one, tier two, tier three that becomes more and more expensive.
3:25:27You can do that as well with your AI services. The key here being that you always want to make sure that you profit. Let's say that you're building out the voice agents to use the example from before.
3:25:38Then you could have, let's say, a 100 calls on plan one, let's say a thousand calls on plan two, and let's say 5,000 plans on plan three, and then have maybe like a fourth plan that's like 5,000 plus, which is maybe per conversation or something that you can charge for.
3:25:55By doing that, by having different tiers, you always make sure that you are in the green. If you're just charging, like, $500 a month and all of a sudden you realize that this company is massive and they're gonna have, like, 10,000 calls a month, you might get ruined. So you want to make sure that you're dialed in on this and that you are always in the profit.
3:26:13A question I get a lot as well is, should I include software in my price? Like, let's say that you're building out the voice voice agents as an example. It costs maybe, like, a $100 a month for you to run it.
3:26:25Should you tell that to the client and get them to pay a $100 extra, or should you include in your price? So if before, if you're charging $500, then you would just charge $600 without telling them.
3:26:35And from my experience, what I've seen is that a confused person doesn't buy. If someone is confused about what they're going to pay, what they're going to get, they're not going to purchase anything.
3:26:46So you want to make your offer as simple as possible. Instead of saying, well, it's like $500 a month for the service, and then a $100 a month for the actual software, we almost always include it all in our price and just have a very simple offer that is easy to understand.
3:27:02There's one scenario, however, where we don't do this, but where we actually do this, and this is if the client wants to, like, own the system completely or, like, run it on their own servers. This is usually with, like, bigger clients that wants to run everything on their own servers. The price tag for that is, of course, going to be much higher as we now need to install it on their servers.
3:27:20And when we're doing that, they are, of course, also paying for, like, the hosting and the software cost. So I'd say when you're just starting out, like, 99% of the time, you would just include the cost inside of your price. When you start going upmarket and working with larger clients, that's sometimes when you have to sell them, well, it's gonna cost this and this on your own servers, and you basically take the price out of your service fee.
3:27:42I mentioned it a bit earlier, but please, this is just like a reminder, you never say the price of the service before you have shown the value.
3:27:52It's what we talked about. Right? You don't for example, if you're typing with someone in the DMs over email, and they're gonna ask how much is it gonna cost, You never say the price before you show what you're actually delivering.
3:28:03You also want to say on the call that you're trying to give them a five x return on investment. They need to see the value that they are going to get to compare it to the price that they're going to pay. You wanna be judged on the value, not on the price.
3:28:15And the only way that you can do that is that you hop on a call with them, you talk about the issues that they currently have, and then show them how your solution could be a solution to their problem. If you fuck this up, then you're never gonna get them on a call because they're just gonna say, well, it's $2,000. That's way too much in debt.
3:28:30But if they realize that it's actually going to save them $10,000, then paying that $2,000 is a steal. And the other thing is that a lot of people, when they start selling in the AI space, they are selling AI.
3:28:42They're focusing on the features like we are using the newest models. We're using the newest technology. We are running it smooth so it's as fast as possible.
3:28:50The reality is that most companies you're gonna work with doesn't really care that it's AI at all. All they care about is the outcome that they're going to get.
3:29:00Will they be able to save $10,000 a month? Yes or no?
3:29:03They don't care about if it's AI or what it is doing it. They just care about the outcome. So when you are on calls and you're explaining the value, that you focus on the ping that they have right now, the amount that they're paying extra right now, which they don't have to, and then from that, show them the outcome, show them what you can help with.
3:29:19Don't focus on the features. Don't focus on it being AI. Hope that clarified some things in terms of pricing.
3:29:26Now you have all of the information that you really need in order to go out and actually start selling AI services. But in what sequence do you actually do it and what does your roadmap look like now? That is what we're going to talk about in the next part of the course.
3:29:40Now you have literally everything you need in order to go out and land your first clients. Don't let this just be another YouTube video that you watch, forget about, and where you never actually take action. Go out and actually use this information.
3:29:53To make it extremely easy for you, let me give you a road map that you can do right now. The first thing is what you've actually already completed. It is to learn a skill.
3:30:02Just from watching this video alone and understanding how to build with tools like Cloud Code, you already now know a valuable skill that you can go out and sell. You should of course still keep learning as much as humanly possible. The AI space is always moving, so you need to stay up to date, but don't get in the trap of overlearning and and never actually going out and utilizing this information that you're learning to anything.
3:30:25So while you're still learning and while you're still keeping up to date with the AI space, now it's time to go out and land clients. And these clients are not going to be the one that pays you the most.
3:30:37Start by working for free. Reach out to people on cold email, like I've shown you how to do this in calls and offer to design a website for them for free. Become their internet partner or get started on Upwork, like you also know how to do now, and charge something low like $15 an hour just to get your first projects done and to build that Upwork profile that can become a very, very valuable asset.
3:31:00One of the biggest mistakes beginners make in this space is that they think, well, I'm starting an AI agency, so I spend most of my time actually learning AI. But that is actually not the truth. The main thing you should be learning is how do you market it, so how do you do marketing for your services, and how do you sell.
3:31:17Those are the first things that you need to figure out, and those are usually the things that are the hardest for most people in this space. People massively underestimate how much time they should actually be spending on marketing and selling.
3:31:30If you're just starting a business, try and spend four hours every single day doing marketing and doing selling. So spend four hours outreaching to everyone you know, sending out Upwork proposals, building code email campaigns.
3:31:43Do everything you can in order to land that first client. And again, the first step is not going to be to make a bunch of money. You're still building the skill, which is the actual valuable part.
3:31:53The next step after landing a couple of projects and landing a couple of clients is literally just to sustain yourself. When you talk about making it a business, right, the success criteria of making it in business is to stay in business. This is the point that you need to get to.
3:32:09Because if you can just sustain yourself, if you can just pay yourself a small salary every single month that covers your rent so you don't have to worry about it, that literally means that you can go all in and that you will never give up. People say that nine out of 10 businesses fail, which is true, but the majority fail right here as step one and two, and that's because they never reach the point, so at one point they say, fuck it, Let me try something else, and they give up.
3:32:32Or they might just be false to, well, this month I can't pay rent. I need to go out and get a job, so I'm going to clip this or hold off for a bit, and they never get started again. See if you can get to this point right here.
3:32:42This should be your main goal right now. Can you just sustain yourself at 2 to $3,000 a month?
3:32:48Get to that point because then you never quit again. And then after that, you reach step four, which is that you start charging more. At this point, you should have a consistent stream of clients coming in either through Upwork, either through cold email, and when you have that, that's when you charge more.
3:33:03You choose the clients that can pay you more and you say no to the irritating clients that are broke and that can't pay. So you work with fewer clients that pay you more. Step number five.
3:33:13Now you need to find bottlenecks. Scaling a business is basically just the process of finding out the bottleneck, solving that bottleneck, and then going to the next bottleneck.
3:33:23Let's say that this is a funnel that looks like this. Let's say that we're getting a bunch of leads from cold email. Right?
3:33:29They all come through here. But then we have a sales process right now, where it's only us taking sales calls, and this only means that we have time for some calls, and our calendar is completely booked up. The bottleneck of this business right here is to hire more salespeople, all by figuring out how can we only get the most qualified through this funnel.
3:33:46Maybe you open this up and all of a sudden you get way more calls and way more closes. Now the bottleneck might be that you only have yourself as a tech person, so then you need to hire more tech people, or you need to automate more with cloud in order to make the onboarding process more smooth, better, etcetera.
3:34:01But that is the process of scaling a business, and that's step five. And honestly, you're gonna be at step five forever. That is the process.
3:34:08Now you're just finding bottlenecks, solving those bottlenecks, and this basically just keeps going on forever as your business grows. This course is about building a one person business with Claude Code, but this doesn't mean that you should limit yourself to just being one person.
3:34:22You start out by being one person in the business, and Claude Code can definitely carry a lot of that weight and doing and do a lot of work for you, and you can probably get to ten, fifteen clients yourself, where Claude does most of the work. However, don't get stuck in this mindset of, okay, I shouldn't hire anyone, because sometimes the best solution is not AI, and that's even coming from me.
3:34:41Sometimes the best solution is to just find a good salesperson that can convert more calls and close more deals. Or sometimes the best solution is literally just to find another tech person that can onboard more clients and make you more money.
3:34:54And the mistake that I see a lot of people make is that they try to over automate, where the easiest solution that they could have done in a couple of days is just hiring someone, and instead they spend weeks building out and automating a solution that then turns out mediocre.
3:35:06But that is the end game. The reason why this course has built a one person business is because that's how you start, and that's also how you can get to make a really good living for yourself, just working by yourself with Cloud Code. I hope that this roadmap is valuable for you.
3:35:19Write it down, and let you just start outreaching. Start setting up your cold email campaign, start setting up Upwork, buy connections, send proposal requests, actually spend a bunch of time doing these Loom videos, and just keep going.
3:35:32When I started my business, I was considering myself smart, but it still took me four months of your straight grind outreaching every single day before I landed my first client that paid me $400. These expectations of being able to make tens of thousands of dollars your first month is just completely unrealistic because you haven't learned the skill that makes you worth that much.
3:35:52In order to do that, you actually need to build something up that can take months, if not years. And the best thing that you can do right now is to set your expectations right from the start, say to yourself that it's okay if I put my heart and soul into this and that it takes one year. Don't expect it to take a month.
3:36:07And when it comes to growth in a business, it's usually exponential. You're gonna have a pain in the start. You're be hovering around a couple $100 a month probably when you're just starting out.
3:36:15But then all of a sudden, something clicks, and that's when the graph goes like this. And just like the stock market, it doesn't go in a nice curve. It goes like this.
3:36:23Right? It goes up and down, up and down, and then you're gonna have good months, and then you're have terrible months, and you're gonna have even better months, and then you're gonna have even bigger falls, and then it goes like this. Right?
3:36:32That is how it always goes. Take it from me. Keep grinding, and it's going to be so worth it when it actually clicks.
3:36:38The last thing I want to say is that if you're already in a job or already in a position right now, don't let that stop you. It's a myth that you have to quit everything that you're doing right now in order to go all in on something, because for a lot of people that actually keeps them from starting in the first place.
3:36:53Because what if you cannot afford to quit your job? That is a big risk if you don't make this work the first couple of months. Instead, what you do, and this is by no means easy, is that you build your business from five to nine, then you are at work from, let's say, nine to five, and then you might be able to get a couple of hours in at the evening before you then go to sleep.
3:37:11Right? That still gives you four hours right here and two hours right here, and usually, you are the most aware right when you wake up. So spend this time working on your business, And then probably in weekends, you probably have ten hours that you can spend on your business.
3:37:24So this, you have to go all in when you're just starting out. It's a myth. You can start getting your first couple of clients, making your first money, setting up your Upwork profile, and do all of that while you're still working a job.
3:37:34Don't quit your job. Don't think that you need to quit your job. Do it on the side until you can actually afford to quit.
3:37:39I really appreciate that you watched this full free course, and I really hope that you found it valuable. If you want my help, then you should check out our community, the 1% in AI. Inside of here, we have the thirty day challenge and the ninety day challenge.
3:37:52And the great thing about these challenges is that when you complete them, then you actually get your first month completely refunded. The way it works is that you get this tracker right here that shows you what to do every single day, what modules to watch inside of the 1%, what tasks to complete, and it basically holds your hand throughout the entire process of learning the skills that you need to learn and setting up your AI agency.
3:38:10So if that sounds interesting, then I'll leave it as the top link right below. Video. Thank you guys so much for watching.
3:38:15I hope you have a wonderful rest of your day. I hope that you're actually going to use this for something. Don't let this be another YouTube video that you've just watched and never actually took action.
3:38:24I am rooting for you. Good luck.
The Hook
The bait, then the rug-pull.
Albert Olgaard opens with borrowed credibility — Sam Altman and a journalist on screen arguing that one person with AI can now build a billion-dollar company — then cuts to himself in a bedroom to say he has done over a million dollars in revenue without a computer science degree, and asks how.
Named ideas worth stealing.
2:36:48model
Explore-Plan-Build
- Explore (gather context, no code)
- Plan (architecture document)
- Build (execute against the plan)
Three-phase Claude Code workflow that prevents costly rebuilds from misunderstood scope.
Steal forAny Claude Code project where requirements are fuzzy or the system is complex
3:29:38list
5-Step Business Roadmap
- 1. Learn the skill
- 2. Land first clients (free or $15/hr)
- 3. Sustain yourself ($2-3K/month)
- 4. Charge more / fire bad clients
- 5. Find and fix bottlenecks (forever)
Zero-to-scaled agency roadmap emphasizing survival at step 3 as the real inflection point.
Steal forPositioning a new service business to an audience of beginners
57:00list
Cold Email Infrastructure Stack
- Custom domain (not main domain)
- Separate sending mailboxes
- Warmup period before sending
- Apollo.io lead list
- Instantly AI sequence automation
- Loom personalization for high-value prospects
Full cold email system that can be built and configured entirely inside Claude Code.
Steal forAny outbound client acquisition system
2:59:08model
AI Services Pricing Model
- Tier 1: $500-1,000 (small projects)
- Tier 2: $2,000 upfront (satisfaction guarantee) + $500/month recurring
- Never compete on hourly rate
Satisfaction guarantee on upfront fee removes client risk, enabling premium pricing on new relationships.
Steal forAny service business moving from hourly to retainer pricing
How they asked for the click.
VERBAL ASK
3:37:30product
“If you want my help, then you should check out our community, the 1% in AI. Inside of here, we have the 30-day challenge and the 90-day challenge. And the great thing about these challenges is that when you complete them, you get your first month completely refunded.”
Soft sell after 218 minutes of free content. The refund mechanic is the main hook — de-risks the paid community purchase. Also links free community (150K members) as lower-commitment alternative.
MENTIONED ON CAMERA
FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
AFFILIATECommission earned if you click.
OTHER LINKSAlso linked in the description.
Visual structure at a glance.
