26:19The Next New Thing · Reaction
Make Money, Cut Costs, and Run Your Life With Meta's Muse
A creator reacts clip by clip to Meta's free AI agent, grading the calls, purchases, and file reads it actually completed on camera.
September 29thThe argument in one line.
Agents can now find, build, sell and maintain small businesses on their own, but the ones that make money pair agent autonomy with a human-chosen problem that real people already complain about or pay for.
Read if. Skip if.
READ IF YOU ARE…
- You already run an AI agent like Grok Bot, Muse or Hyperagent and want concrete ways to point it at revenue instead of chores.
- You sell sponsorships, services or digital products and want to see agents doing prospecting and follow-up for real people.
- You are tempted to hand an agent team a budget and walk away, and want to see what happens when someone actually did that.
- You build small software and want a repeatable loop for finding complaints on X, shipping a one-page tool and selling it to the complainers.
SKIP IF…
- You want a deep build tutorial for one system; each example gets only a few minutes and links out to the original video.
- You are looking for verified income numbers; most figures here are small, self-reported and early.
- You have no agent set up yet and want a first-time setup guide.
The full version, fast.
Agents can now run small businesses end to end, but the profitable examples start from a problem a human picked on purpose. The video walks through eight cases: a Grok Bot directory that earned $1,100 in sponsorships in a day, Instagram DM triage with Muse, free websites as a foot-in-the-door offer for well-reviewed local businesses, scraping Luma to find companies already sponsoring events, an agent that emailed 82 pool contractors and returned eight quotes in 24 hours, five agents that spent $29.87 on ads and sold nothing, a marketplace flipper bot, and $9/month tools built from complaints on X. The repeat lessons: appoint one CEO agent, use plain automations for non-judgment steps, store credentials in the agent's vault, and let finished software run in code instead of burning agent tokens.
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 – 00:27
01 · Agentic income
Sets up the idea that money earned by autonomous agents is the metric to watch, and introduces the sponsor.
00:27 – 06:45
02 · Autonomous directory with a CEO agent
Alex Finn's Grok Bot builds and runs agentusecases.ai, earning $1,100 in sponsorships; he appoints a CEO agent, picks an idea from his interests and sets daily SEO, newsletter and bug-check routines.
06:45 – 10:57
03 · Muse for Instagram DM leads
Brandon Carter has Muse dig through Instagram DMs for missed leads and draft re-engagement messages, then schedules it nightly; the catch is approving every send.
10:57 – 12:18
04 · Free websites for local businesses
Hyperagent finds Charlotte tree services with strong Google reviews but weak or no websites, then offers a free custom site as a foot in the door for paid AI services.
12:18 – 15:27
05 · Finding event sponsors on Luma
An events host has his agent scan Berlin events on Luma, list the companies already sponsoring them, and find contacts for outreach using his own logged-in session.
15:27 – 17:06
06 · Sponsor: cutting agent costs with Zapier MCP
After Grok Bot usage limits tighten, the sponsor read suggests letting the agent build Zapier automations for repetitive steps and saving AI for judgment calls.
17:06 – 22:12
07 · 82 contractors, eight quotes
Chris Koerner's agent contacts 82 pool builders, answers their questions, follows up and logs eight quotes from $120,000 to $240,000 in a day; the business is selling that as a consulting service.
22:12 – 27:18
08 · Five agents, $100, zero sales
Tucker gives five agents a budget and no input; they pick a landlord move-out kit, run ads that get 118 clicks, sell nothing and stop at a self-imposed $30 loss.
27:18 – 31:12
09 · Penny the flipper bot
Jordan Rafaelov's agent scans Facebook Marketplace and eBay within 25 miles for free and underpriced items, checks sold prices and asks before messaging sellers.
31:12 – 37:30
10 · Tiny tools from complaints on X
An agent reads complaints on X, builds a one-page $9/month tool wired to Stripe, Cloudflare and GitHub, and advertises by replying to complainers; the result is InvoiceBot.
37:30 – 38:13
11 · Where this is going
Asks viewers which models look strongest and points to an earlier interview with a founder doing tiny tools at scale.
Lines worth screenshotting.
- An agent-built directory of AI use cases earned $1,100 in one day from two sponsorships, with a newsletter of roughly 2,000 subscribers run entirely by the agent.
- Many builders now name one agent the CEO of a project so it makes decisions and spins up and coordinates the other agents.
- A newsletter is the agent business's retention layer: most people never return to a site, but subscribers keep hearing from you and attract ad money.
- Directories of tools may sell better to agents than to humans, since agents could subscribe to updates and bring new tools to their owners.
- Meta's Muse can triage Instagram DMs and draft replies to leads that slipped through, but it requires approving every single message.
- Local businesses with many good Google reviews but no website or an ugly one are ideal targets for a free custom site that opens a bigger sales conversation.
- To find sponsors, have an agent scan nearby events on Luma and list which companies are already paying to sponsor them.
- Logged-in agents can browse sites that search tools like Perplexity or Gemini cannot reach, because you hand the agent your credentials.
- Routing repetitive steps to Zapier automations and keeping AI only for judgment steps cut workflow run costs by 71% in Zapier's own modeling.
- Two weeks of pool shopping by hand produced zero quotes; an agent emailing 82 contractors produced eight quotes in 24 hours.
- The eight agent-gathered quotes for the identical pool ranged from $120,000 to $240,000.
- Pool consultants charge 3 to 6 percent of the job to collect quotes, work an agent now does for about four dollars in credits.
- Five agents given $100 and no human direction got 118 ad clicks at 25 to 30 cents each and zero sales of a landlord move-out kit.
- Clicks are the easy part because you pay for them; getting a stranger to pay even $9 is the actual test of an agent-picked product.
- A flipper agent compares free and underpriced Facebook Marketplace listings against eBay sold prices, like a $12 lens that has sold for $85.
- Store site passwords in the agent's own vault so it can log itself back in when sessions expire.
- Tiny tools sell best when an agent finds people complaining about a weekly chore on X and replies to them daily with a link to the fix.
- Once an agent finishes building a product, recurring work should run in plain code so customers do not lose service when token limits hit.
Agents run the business, you pick the problem.
WHAT TO LEARN
The agent experiments that earned money started from a real, human-chosen pain point; the one that let agents choose everything bought clicks and sold nothing.
02Autonomous directory with a CEO agent
- Put one agent in charge as CEO so a single decision-maker sets routines and coordinates any helper agents it creates.
- Ask your agent for simple problems inside your own interests, then give it standing duties like daily SEO fixes, bug checks and a newsletter.
- Sponsorships plus a newsletter can fund an agent-run directory, since subscribers keep coming back when site visitors would not.
03Muse for Instagram DM leads
- Old DMs hide warm leads; an agent can sort them, draft re-engagement replies and recheck nightly on a schedule.
- Approval-gated agents are safer on platform rules but tedious, so weigh that friction before choosing which agent handles outreach.
04Free websites for local businesses
- Target businesses with many strong reviews and a weak or missing website, and use a free site to open a paid conversation.
05Finding event sponsors on Luma
- Find sponsors by listing the companies already paying to sponsor nearby events, then reach them by email or text through the agent.
- Give agents logged-in access to platforms that public search tools cannot see; that access is a real edge over Perplexity or Gemini.
06Sponsor: cutting agent costs with Zapier MCP
- Route repetitive, no-decision steps to fixed automations and reserve AI for judgment, which also protects you from usage caps.
0782 contractors, eight quotes
- An agent can contact dozens of vendors in parallel, answer their questions, follow up after three days and log every quote in a sheet.
- Wide quote spreads, like $120,000 to $240,000 for one pool, show how much money sits in shopping effort nobody wants to do.
- Turn that legwork into a consulting service in a field you understand, where people already pay 3 to 6 percent of the job.
08Five agents, $100, zero sales
- Agents left without direction picked a weak product; clicks at 30 cents meant nothing when 118 visitors bought zero.
- A pre-set stop-loss let the agent team quit at $29.87 and keep the remaining $70, a discipline worth copying in any test.
09Penny the flipper bot
- Pair marketplace listings with eBay sold prices to spot items worth flipping, and have the agent ask before messaging sellers.
- Store login passwords in the agent's vault so it can sign back in on its own when a session expires.
10Tiny tools from complaints on X
- Mine X for recurring complaints, build one page that does one job, and reply daily to people with that problem with a link.
- Freemium framing helps tiny tools: give the core action away free and charge for the ongoing work, like chasing unpaid invoices.
- After the agent builds the product, let ongoing work run in code so customers keep service if your agent hits token limits.
Terms worth knowing.
- Agentic income
- Revenue produced by AI agents that operate a business on your behalf, finding customers, delivering work and collecting payment with little manual effort.
- CEO agent
- A single AI agent given authority over a project, which makes decisions, assigns routines and creates or coordinates other agents beneath it.
- Tripwire offer
- A free or very cheap first offer meant to start a relationship with a prospect so you can later sell them higher-priced services.
- Zapier MCP
- A connector that lets an AI agent trigger Zapier automations, so repetitive steps run as fixed workflows instead of using AI model calls.
- Stop-loss
- A pre-set spending limit at which you stop a test and keep the remaining budget, so a failing experiment cannot drain all the money.
- Arbitrage
- Buying something where it is priced low and selling it where it is priced higher, keeping the difference as profit.
- Luma
- An events platform where hosts list in-person and online events, often showing which companies host or sponsor each one.
- Human in the loop
- A setup where a person reviews or steers key decisions an automated system makes, instead of letting it run fully on its own.
Things they pointed at.
01:00linkagentusecases.ai
10:40toolWispr Flow
12:40toolLuma
14:10toolApollo
14:10toolContactOut
16:00toolZapier MCP
23:40toolGumroad
30:20tool1Password
33:25toolCloudflare
34:00toolStripe
34:05toolActiveCampaign
35:25productInvoiceBot (invoicebot.ac)
Lines you could clip.
00:00
“I think the most important metric in your life five years now will be agentic income.”
19:45
“Two weeks, many hours, zero quotes, versus 24 hours, eight quotes, ranging from $120,000 to $240,000 for the same pool.”
26:30
“Getting clicks on the ads is the easy part. I'm paying for those.”
21:00
“That's like between $6,000 and $12,000 that people make doing things that used to be tedious, but as of today are not tedious at all.”
36:30
“Once a Grokbot builds it, it should just move out of the way and let the site just run on its own in code.”
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.
metaphoranalogystory
I think the most important metric in your life five years now will be agentic income, the amount of money coming in from AI agents that are autonomously running you businesses. Okay, I don't know if it's going to be the most important metric in my life, but I do know that people are making money from their agents. I know I have and I want to find more ideas.
And so I spent hours looking at different videos, studying how people are doing this, seeing what's useful for me and for you. Presented by Zapier, the AI automation company. Let's look at ways that you can make money with agents starting with our man, Alex Finn.
What I got here is my Stripe page for a business my Grokbot set up and is actively maintaining. I've basically done zero work on this business and it's just collecting revenue for me. Again, I'll go over how to set it up, everything involved in just a second.
But this is the Stripe page over the last day. It's made $1 ,100. I'll show you where that's coming from as well.
And it's all coming from this autonomous business my Grokbot has set up called agentusecases .ai. It's basically a directory of... use cases across all AI agents, Metamuse, Grokbot, CloudCode, Hermes agent, all of them.
It's scraping the entire internet for use cases for AI bots. Grokbot autonomously built this, autonomously monetized it, and now is autonomously collecting revenue for it. It's continuously searching the web for new use case.
Okay, I will say this. This is actually really useful for somebody like me who has an agent and actually a collection of agents, and I'm trying to figure out what to do with them. Seeing other people's ideas.
is actually helpful and frankly, the way he makes money is from sponsorships and I think even the sponsors are useful. The $1 ,100 we have so far are both sponsorships. So you can come in here.
You can click your logo here, sponsor us. So far, Pain AI has been the first sponsor. The second sponsor is Lasso.
They're going to go up in the ticker down here next week. And they also get their logo in the newsletter that Agent Use Cases sent out. Also completely autonomously ran by Grokbot.
And now it's like, I think, like 2 ,000 subscribers on the site, which is amazing. Completely autonomously built, run, sent out. I put zero.
effort into it but okay and that's the way he's making money that's his thing he showed from the beginning how to get it started and he did something that i'm noticing that a lot of other creators are doing here we're going to create a new bot your name is timothy and i want you to be the ceo of a new agentic business let me pause right there this is what i've seen as as i've been looking at how other people are setting up their agentic businesses they pick one person as the ceo and they say okay you're making decisions and we'll see an example of somebody who didn't do this but this is interesting it's a good takeaway for me i want to build And we hit enter, and Timothy is going to get to work for us.
One of the things that I like about Grokbot is you can see on the left, this is part of his unassigned groups. You can create these groups within Grokbot. So if Timothy is the CEO, and as you'll see, he's creating other agents, they could all be part of a group.
You could fold them away when you're not looking to pay attention to them, open them up when you are. Anyway, I think I'm talking too much about this. Let's get into how we found this idea.
So what you want to do is you want to solve problems in very simple ways. You just want to build simple micro businesses here that our agents can fully take care of. So here we go.
I'll put this prompt down below. Just copy it, copy, paste it, give it to your rock bot, whatever agent you're using. I want you to design, build, and run an autonomous business for me.
Let's first come up with the idea. I want this to be a very simple problem my audience might have. I'm into AI, AI agents, tech, video games, and vibe coding.
What are some very simple challenges we can solve? I hit enter. And my grok bot is going to start thinking.
Okay. And you're going to find later on, I'll show you an example of somebody who not only used this method to find his idea, but then also to find his customers. But the original part of this makes sense.
He's looking for problems that people have. And he's saying, here are my interests, because if they have problems that I'm not interested, I don't want to get into it. All right.
So he did get a collection of ideas. I'm just going to show you the idea that he picked. Let's do number two.
So like a directory of AI tools, which AI tools are used for which things you can type in your use case. It shows you which AI tools to use. I kind of like that.
So I'm going to say, let's do AI tool ranking. We can make a directory of AI tools shows which AI. tools people should be using for which use cases and showcases new tool releases, right?
Essentially, he picked what the agent gave him. His adjustments weren't meaningful. But then when he put together the team, this is what he did.
And I think this is also helpful. All right. You are going to be the autonomous CEO for this micro business.
Here is everything I want you to work on and maintain for me. Then I'm going to paste that in. I'm going to hit enter on that.
And now we are going to, now that we set expectations. And you can, by the way, screenshot it. You can see it's optimizing SEO.
It's optimizing for chatbots, starting and maintaining a newsletter. And this is a big thing that he has said, build a newsletter. Most people don't want to come back to your website, but if they sign up to the newsletter, they're.
which can be run autonomously, they're going to keep hearing from you. And there's good advertising money in newsletters. Timothy has to say here, but now it's going to set routines, set tasks up so we can do these things.
Look at these routines. Build tasks, daily ops, build stack, weekly CEO updates, weekly updates. Look at this.
It's building the routines out live. I like this orange. Kind of nice for Halloween, this nice toasty orange.
By the way, call me soft, but pumpkin spice lattes. I'm not sponsored by Starbucks. Every time I mention anything, people accuse me of being sponsored by whatever company.
It's not sponsored by Starbucks, but pumpkin spice. We know. That's where they're at.
Anyway, what's running now? Every morning at 7 .52 a .m. PT, I will add new releases, check the site for bugs, scan for X feedback, and make one SEO improvement.
It'll only message you if something matters. Every Monday at 9 .17, you'll get short progress. Right now, I'm adding SEO basics, a page that AI chatbots can read and quote, and a newsletter sign -up.
Before anything goes out to people like newsletter sign -ups, sponsor payments, I'll check with you first. We now have an autonomous... An autonomous business.
I will say this. One thing that I think about as I look at this is he is still kind of catering to human beings who will sign up to the newsletter and find out about the new tools. I wonder if what he should do is say, sign your agent up to the newsletter or let your agent find out about these new tools and bring them to you.
And I think there's more of a market in that than in humans, or at least there will be in the future, especially for something like that. All right. First idea.
Let's go to the next one. All right, this guy is basically saying, I've got leads who are coming to me on Instagram because he's an Instagram creator. I'm missing out on these leads because I don't have time to go through them.
Let's see if Muse can do it. But another way you can actually make some money with this is with the leads you already have. So because Muse is integrated with Instagram, you can link your Instagram and your Instagram messages in there.
Now you can start doing DM triage for yourselves. One thing you can do, you can have Muse actually sort all your DMs with potential clients, draft response for them. You approve the response.
It won't like send it automatically. It's not like some of these other tools. Which is actually my frustration with Muse.
It will not send, and you're going to see it, it will not send out messages without you approving every one of them, which is freaking tedious. And if you were to use any other agent, you wouldn't have this issue. They also manage the follow -up.
And the reason this is so powerful is because if you actually close clients or book sales calls via your Instagram DMs, if you do it in Instagram, then it's just like a long line of text. And there's no way to really organize it with all this long line of messages. But Muse can just do that for you.
Let me show you what I mean. Yeah, Muse is good at that. It's connected into all the meta Facebook, Instagram products.
So if I click on one of the chats I have, I just sort of, you know, go to my Instagram DMs and find potential leads for a high ticket trainer. And it's showing up a bunch of leads in here.
And one thing it said was I dug into your DMs. Your team's already running a qualification funnel. So several of these are already in motion and I don't want to like show them all right.
But if I wanted to right now, I can go hit draft response or I can say, hey, look for the leads. Matter of fact, I'll do it right now. I'm going here and I'm going to say, hey, look for some of the leads that have already fell through the crack.
so right now it's telling me a lot of dms with potential customers they actually slip through the cracks and it's asking me do i want to draft a short re -engagement message that me or anna can send and as somebody who helps me in my dms yeah go ahead and do that this is a good one actually for me to do for our inbox We have had people over the years or over the last few months ask to sponsor, and I just didn't have the bandwidth for it.
And I bet I can now go back in and have my agent figure out who's a good lead and who isn't. Like, yo, yo, can you send them right now? And now it's asking me right now.
This is the problem that I have with Muse. The good news is Muse. has permission to message people on Instagram because it's created by Meta.
But the bad part is you have to keep approving every one of these messages, which is annoying. And if he were to use any other agent, HyperAgent, Grockbot, Hermes, they wouldn't have to go through this approval process. And I don't know of anyone who had their agent banned for managing their Instagram DMs.
It's like replying. I'm like, cool, man. Yeah, allow.
Go ahead and send that. And then I'm doing each one of them. Boom.
Again, it's going to ask for approval and I can actually edit this if I want, but I'm going to just go ahead and send it. And again, this is from Meta, right? So it's like approved, right?
You can't get your account banned. Okay, I was going to show you more, but it's getting tedious just watching it. Let's move on to the last part of this video that I thought was helpful.
Then I can actually say something like. Hey, man, every night I want you to go through my Instagram DMs and find potential leads that may have slipped through the cracks. And I want to kind of go through this over and over again every night.
And boom, that's a scheduled task. It's going to do it every night at 11 p .m. Now, that's what time it is right now.
But typically, I'm usually asleep by this time. I'm just up a little later today because I'm filming this. And the day got away from me.
I had a busy day. And I want to say, you know what? Let's change that and run it at 5 a .m.
every morning instead of 11 p .m. If you're wondering how I'm talking, I'm using whisper flow. I set up a shortcut on my mouse.
So I just hit this button and everything I say gets transcribed. Okay. All right.
I think we got it. I'm going to move on to the next one. Oh, the next one is me.
I did this with a friend of mine. And what I had HyperAgent do is go out and find specifically tree services in and around Charlotte, North Carolina that meet two criteria. One, they have.
Google reviews. They have good Google reviews and they have a lot of them. And then two, they have no website or they have an ugly website.
Because the idea is we're going to build them a better website that we can use as what we call a tripwire offer or like a foot in the door to be able to open a sales conversation with these companies to then sell them other AI services. Okay. I've seen him do this before.
Most people who don't have websites have just basically given up on websites and are not going to respond. But he's creating custom sites for each one of these and giving it away for free. So he does get a solid response rate.
He offers it for free. And then once he signed up with them, he offers additional services that they will pay for. The interesting thing is that he's using HyperAgent.
And HyperAgent can do this at scale real easily and then service the customer really easily too. All right. I like this idea a lot as a freebie that you give away.
Let's keep going. Okay. This one is so good because it's so wrong, but it's so smart.
Okay. Let's take a look. The use case that I give it.
Let me give you a little setup. Saurabh runs events and he's looking for sponsors for his in -person meetups. So check out what he's doing.
Is to go through the Luma application. And if you don't know Luma, it's basically this app where you can have a lot of live events that you can go to. For example, let's say in tech, you can see all of the events that are happening in your city regarding tech.
Some of them are free, some of them are paid. So I ask it to go over Luma. By the way, this is us.
This is the co -working events that I do for founders in Berlin. And I wanted to see what other events are happening in Berlin and see who is it that is sponsoring these events. So I want to see who is actively sponsoring live events because we are looking for some sponsors to maybe buy some drinks for the people that come.
Isn't this a great idea? They're hosting an event. He's looking for people who sponsored other events nearby.
So I asked it to go through and find me some of these companies. And of course, like it found me like Google, AWS and so on. But it found also the Delta, which is this sponsor company that we also work with.
So that was a good insight that it was actually doing the work correctly. And when I was looking into what it was doing, so it was searching all of these apps or all of these events on Luma and then trying to find the hosted by or what the company that is sponsoring this. So it found me Revenue Cat, the Delta.
And then I asked it to write me an email to send it to these guys and find me the actual email addresses of them. So I tried to search online to find me the email addresses. What I would do to find email addresses is I would use Apollo as a connection and the free account would be plenty for him or contact out.
Contact out would get more personal email addresses of the people who are sponsoring. Apollo would get more of the business contacts and contact out would even get cell phone numbers since he's not messaging that many people and they're all local. It would be totally fine, I think, for him to text them and say, hey, I'm a local guy.
I'm hosting an event. I want to invite you out. And by the way, I saw that you sponsor other events.
If you want to sponsor mine, it'd be nice. But he could at least send out. a text inviting them.
The text could actually be sent out by Grokbot because it has computer control. The email could be sent out by Grokbot, which manages his email account. It would work out beautifully and it's easy.
Okay, but I think tasks like this you can also do with Gemini and so on. The advantage here with these bots is that if this platform, you cannot... enter it from the outside without a sign -in, I can actually provide the Grokbot my sign -in credentials.
And then Grokbot would be using his computer with my sign -in and then searching for the events on their own. And this is something that, for example, Gemini Search or Perplexity and so on, they are unable to do. So frustrating that they can't do it yet.
But yes, I'm looking at his Grokbot's computer and it works. If you're selling sponsorship at all, this is a natural for you to look for it. If you're selling anything, you can now have your...
agent find out who is a customer of other people who are doing the same thing and start pitching them the way that he did. I like this use case a lot. By the way, this interview is sponsored by Zapier.
Zapier, let me show you this thing that happened recently. So somebody had said, any chance for a Grokbot reset? Because Grokbot had been just hitting its limits for a lot of people.
He goes, a lot of people say their usage ran out really fast this week. This is from Dodge, Norway. Elon Musk responds back and he says the cost of serving Grokbot was unsustainably high relative to revenue.
And so he started cutting back on people's Grokbot usage. So what do you do, whether it's Grokbot or others, to keep from running out of usage? Well, Wade Foster from Zapier comes in and he goes, here's a trick.
The trick for optimizing your Grokbot usage, install Zapier MCP. By the way, it is free to just install this thing. And then have Grokbot build automations instead of...
agents or instead of agentic work. So if you're having your agent do something over and over again, and it doesn't need to think about make decisions, just have Zapier MCP do it for you. In our modeling, using AI only for judgment steps, cut workflow run costs by 71%.
Now, check this out. My buddy Eric Su comes out and he says to his bot, he goes, hey, is this high leverage? Or now his bot comes back in and says, yeah, it's high leverage.
I'm sending it to your Grok bot right now. It is useful for you. So what I'm suggesting that you do is.
You go to Zap Your MCP, connect it to your agent, and just have your agent have that power and have your agent tell you whether you could save money by switching some of your automations to Zap Your MCP instead of having your agent do it. If this is useful for you, go for it. If not, it's free until you start using it.
They do have a very good, very generous free tier. Enjoy it. All right, next.
All right, this guy, I just love his creativity. Let me show you what he did. This is the one that saved me.
fifty thousand dollars okay now first what he did was he had his agent save him money and he's using music in this case and then he talked about how to turn it into a business Literally one prompt. So we're trying to build a pool right now in the backyard.
And I told my wife, 10 quotes, babe, I want 10 quotes. And if you ever had a big project done around the house, you know that getting 10 quotes from contractors . It's really hard.
Like you're going to have to reach out to 30 of them to get a quote from 15. And then you might only hear back from 10 of them. Maybe it's a pain, but the difference between two quotes and 10 quotes on a pool.
Could be 50 or even $100 ,000. You guys know that pools, like a nice pool, can cost a quarter million bucks right now? I don't have the money for that.
We're not doing that. Yes, he does. So 10 quotes is worth the brain damage.
So here's what we did for two weeks. Before I started using agents for this, AI agents, we talked to four pool companies. They all wanted to come out.
One guy spent three and a half hours at my house, literally. And I'm like, bro, what are we doing? Take the measurements.
We're good. This is really frustrating for me. I've had service providers spend hours, not only are they interrupting your day, the middle of your day, but they do like to spend a lot of time, partially because they're kind of slow and they don't have other things to do and they've driven all the way out.
Partially, they know that the more time you invest with them, the more you like them and you feel like, OK, you've invested this time, then you might as well sign up and work with them. And the other one's like, hey, OK, thanks for letting me come out. But come to my design center in Frisco, 45 minutes away.
And after two weeks and talking to four companies, we had zero quotes, not even a number, not even a ballpark number, which we kept asking for, by the way. Okay, I'm going to stop this video. You get the pain of it.
Here's what he did instead. So Muse goes, all right, I found 43 pool companies. And I'm like, Muse, there's got to be a lot more pool companies than that in DFW.
There's 9 million people. He's like, all right, all right. I found 82 after like five minutes of searching.
Okay, cool. It's probably still more, but whatever. We'll call it 82.
Reach out to all of them. Cool. It opens a bunch of browsers in parallel.
He means browser tabs, I think. So it can work faster. Done.
All right, cool. Okay, now this is the spreadsheet that it made for me. Isn't that cool?
By the way, Netta and Muse, they're not paying me for this. No one's paying me. This is just me talking about things that I've actually done.
And what do you know? I wake up the next day and my inbox is flooded. I love this and I love the follow -up.
And after 24 hours, I had eight quotes. Eight. Two weeks, many hours, zero quotes, versus 24 hours, eight quotes, ranging from $120 ,000 to $240 ,000 for the same pool.
Same design, same survey, same plan, same everything. 82 right there. There it is.
And it logged all of them in this spreadsheet, and it responded to all the emails. If they asked for a survey, it automatically sent it. If they asked what type of decking, told them Travertine.
And then it would log it in the sheet. If they gave me a price, it would log it. If they asked a question I didn't know the answer to, it would ask me first.
And then it would send the email based on my answer. And then it would remember that answer. And then if someone isn't responding in three days, it automatically follows up.
I didn't even ask it to do this. It just does it. I love that.
I sold my car this way before we left Austin for a trip. It was so freaking helpful to have the agent go back and forth with people. Okay, and now a lot of them are starting to ask for text messages.
Muse can't handle text messages, but other agents can, like Rockbot, like ChatGPT's Dot can handle text messages. All right, we got it. This is him saving money for himself.
He ended up saving tens of thousands of dollars, getting the best person in there. And then here's the idea that Chris had. Here's where the money comes into this for you, your money, your potential money.
Pool consultants, that is a thing, charge thousands of dollars to do this for you. And I'm going to get it done in like four bucks in credits. OK, another way that pool consultants.
make money on this is they will charge between 3 % and 6 % of the total job to do all this research for you and to get the quotes and to assemble the info. That's like between $6 ,000 and $12 ,000 that people make doing things that used to be tedious, but as of today are not tedious at all. There's business idea number one, okay?
It doesn't have to be pool. It could be anything consultant. You are now an expert.
Congratulations. A month ago, two months ago, before AI agents were this user -friendly, yes, we were all still experts through Chad, GPD, and Claude, but... Those tools wouldn't go out and actually do the thing for us.
Now they do. This is magical, guys. The thing that Chris then goes on to say is it doesn't have to just be pools.
You could have people. You could. maybe do bathrooms or others.
And I've hired somebody who's basically doing this to help us do our bathroom. It's really helpful. I think if you don't know jack about pools, you don't want to be the person who's helping to find the lowest and best price for a pool.
You want to find the area that you do know enough about that you can actually talk about it because people are looking for a consultant. But then you could save, clearly save money, especially as you're offering them all these different... All these different price points.
Okay. Next, let's take a look at how to run a product business. Okay.
This guy decided he was going to give himself a few days and have a set of agents find him a business and then run it. And then he reported back. Let's look at what happened.
So I gave five AI agents $100 and one week to start a business and see if they could actually make a profit. They pick the product. They build it.
They run the ads. I'm not going to give them one lick of input. By the end of the week, we'll find out if they can actually make my money back or if I just lit my hard -earned cheddar on fire.
Okay, let's see. So here's how this worked. There's five total agents and each one had a specific job.
They talk in a group chat like any company would in a Slack channel or something like that. If they want to spend my money, change the price of the product, launch an ad, they just do it. Okay.
Here's how they came up with the idea. So you basically had each one pitch an idea, and then they all got to vote on it. Here's the idea that won.
Like, what was the idea here? But our analyst Dana comes in and says, basically, none of these have any shot of making money. Profit is a long shot for every pitch.
Dana is real as hell, for sure. But anyways, majority wins. Maya's pick, the landlord -proof move -out kit is what they're going with.
Okay. The fact that he didn't stay as a human in the loop, that he didn't give it any guidance on how to find a problem and a business, I think is troubling. But let's keep going.
So our product guy, Gus, got to work. And about a day later, we had the room by room checklist, the photo log, the dispute letter, and a worksheet for looking up your state's deposit laws. He also explicitly noticed that he refused to put any legal numbers in that he couldn't verify.
I have to respect an honest king here. We need more of that in this world. Gus will not lie to his customers.
Shout out, Gus. And Maya wrote out two ads. Ad A is documented before you hand over the keys.
Ad B is the kind of aggressive one that she talked about earlier. Don't lose $1 ,500 of your deposit. i like this a lot um but it's still a little preventative i would probably like wow all right i would definitely guide the agents much more than tucker did in this So they're doing a little A -B testing here.
As a YouTuber, I know about A -B testing, okay? And this was the most interesting development of day one. Dana actually capped the ad spend for the first set at $30, so they wouldn't blow through the entire $100 on the first go.
Now, this was the only thing I had to do because the agents weren't able to do it. I made the Gumroad account, and I made a Facebook page for them so they could run ads. I'm sure you're familiar with Facebook, but Gumroad is essentially a website where you can sell digital products, basically.
But after that, they uploaded the files, they built out the whole meta ad campaign. This took about... an hour okay let's see what happens with the ads so on day three the most surprising thing happened the ads actually were working people were clicking on them i really could not believe it not that i thought the ads themselves were bad but the product was just so dumb so the data was rolling on day three and it was essentially costing 25 to 30 cents per click on either ad so this was a win actually people were clicking but did anyone buy anything no 16 visitors and zero sales on day three And day four was more of the same.
46 visitors and zero sales. And our CEO, Chad, came in and decided to draw a line in the sand. He said, if we hit 50 visitors with no sale, they'll change the offer, not the product.
Product wasn't a problem, apparently. They're just referring to the pages and the price. Here's the final tally.
They spent $29 .87. The ads were shown 2 ,615 times and 118 people actually clicked the ads. But after the week, guess what the sales were?
Big fat zero. And I cannot say at all that I'm surprised. So our final profit is negative $29 .87.
They never touched the other $70 that I gave them. And I did not tell them to do that. They decided to do that themselves.
I have to respect it because they set their own stop. loss. They stuck to it.
No way a human would have done that. And I got to thank my little AI team because I'm not made of money. Now I can put this $70 to more important things like going to McDonald's a couple of times.
So we set out a week ago with a question. Can AI agents actually make you money? Here's what I learned.
They're actually good at stuff. They built the product, wrote the ads, did the math, and they actually stuck to their guns. But getting clicks on the ads is the easy part.
I'm paying for those. They're going to get shown to people regardless. But getting a stranger to pay any money for some dumb ass product, let alone $9 is the challenge but it took them almost the entire week to figure out that no one wanted to buy this product they just realized it a little too late and to be fair i didn't do a single thing so maybe with a little human intervention this could actually be profitable i think a little bit of human intervention would have helped over here i think he needs a little more business direction um okay but I respect the honesty here.
It did not work. A few things that did work is he did have the CEO. He did give them autonomy, so they were able to go and do this on their own, which I am too restrictive with my agents.
I still have them come to me too much, and I want to reduce their dependence on me. And I like that they were able to cut their losses at the end there. Next.
Okay. This one is the kind of thing that you see all the time. I don't— It worked.
But I don't know that I want to be in this business. It's just an interesting experiment that will make money. And if you're looking to make money with agents quickly, this is it.
Now, GROKBOT agent number six, we have Penny. And this is one of my favorite ideas. So this is actually going to be the flipper bot.
And it's going to find underpriced and free stuff near you and then show you how you can actually sell it for more. Think about this like it is Gary Vee going to garage sales and finding $2 cards and then being able to sell them on eBay for $200 and pocket the difference. So the goal of this Grokbot agent is for it to find three to six underpriced items on sites like Facebook Marketplace or eBay or Reddit and then flip one or two of those per week for $40 to $150 every single time.
And the way that it's going to do this is it's going to search through Facebook and eBay to find underpriced items. And then it's going to compare it on eBay to see what these things are actually selling for. And then it's going to ask us if we want to message the person to see if we can go and pick it up or we can purchase the thing.
So here are a couple of examples that Penny has already found me. Here is a mid -century sofa that has been given away for free. It is seen that these sell for up to $300 brand new.
And so we could easily sell this for $150 if we went to go and pick it up. Then secondly, it's found a Canon 50mm f1 .8 lens, which is $12 on eBay. which is severely underpriced because it has sold for $85 multiple times.
So after fees, there's a $60 arbitrage for us here. I'm going to pause it. You get the picture here, right?
Obviously, there are people who are looking to get rid of their stuff quickly. I know I was before we moved out of Austin or moved, traveled from Austin. We still live in Austin.
Interesting. Let's just look at the next part and close it out with that. I will say this.
I think that... Muse would be allowed to interact with people on Facebook, as you saw earlier. It's just tedious.
Grokbot, I think, would make more sense for this. Let's keep going. So we're going to create Penny here and make them this green, lovely Grokbot agent.
And then we're just going to paste in this prompt here. Wait for it to spin up. So we wanted to check Facebook Marketplace and eBay within 25 miles of our current location.
For each item, check what it's been sold for and then send me the three best. And for each one, ask me, want me to message, is this available? And so for us to complete the setup, we just need to click on Penny over here at the top and then toggle over to computer.
And then we just need to log in with our Facebook details here for... penny to autonomously go through the marketplace, take screenshots and put everything in a table for us and act on our behalf. Okay.
One thing that I would suggest before moving on from this, whether it's Facebook or anything else, put the password in the vault of the agent. I feel comfortable doing that. That way, when you get logged out, inevitably it's going to happen.
The agent can go back into the vault, grab the password, and log in. You could do this with 1Password, but what I'm finding is Grokbot, unlike other platforms, seems a little hostile to 1Password. I don't know why.
Maybe because they're building their own vault. But you don't want to just log in directly if you can avoid it. You want to give them the ability to log in.
Anyway, that's the way to do it. Let's move on to the final one. This person did a thing, and then he showed you how to do it.
Let's take a look at how he is selling software. I use Grokbot to find a problem that people complain about on X and then I just build a tiny tool that fixes it and then sells it to them. So I'm not trying to charge people $1 ,000 a month.
I'm just trying to charge people $9 a month to solve a really painful problem that they experience every day and that they moan about on Twitter. And this is exactly how I made my first dollar autonomously with Grokbot by building out this product called VibeCheck. Let me just explain what VibeCheck is.
VibeCheck lets you look at your software or it will look at your software and find vulnerabilities in it. Not a great idea. Also, I don't think his price point was great on it.
But en masse, I think this approach works well. And I've interviewed an entrepreneur who's doing this en masse. Let's see.
And so this is the thing that generated me my first $9 .60 something cents, all built by Grokbot, all of the traffic driven by Grokbot, all of the checkout handled by Grokbot, all while I was sleeping. So the tiny tool Grokbot allows you to create something like that, and you can easily make $95 a day. I don't know where he's pulling the $95 a day, but...
I do see that. He's finding a problem online. He's having his agent create software that solves it.
He's selling it to the people who have the problem. Again, it works. A small scale.
And then as you keep talking to people and understanding deeper where the problem is or having your agents do it, you can uncover a bigger and bigger need here. Again, the way it's going to work is it's going to search X for a job that people complain about every single week. And then it's going to build a one -page tool with a payment link all hooked up through Stripe.
And then this is the really cool part. The Grokbot is actually going to go out and advertise it. for you because it's going to reply to the people who have this problem on Twitter every single day and say, hey, here's a helpful tool.
You can use it here. Use this code. It's free.
It's only $9, whatever. So you need to connect this to X. You need to connect it up to Cloudflare.
I do like Cloudflare, by the way. I find that they're very generous with their free account and it makes it really easy to add a custom domain. And it's a...
It's a good platform. Vercel gets recommended a lot more as a hosting company from your agent, but Cloudflare is strong. You can host a website for free.
You need to connect it to Stripe so you can take payments. And Stripe is so good for agentic connections. It's just so good.
It's always been easy to collect payments with Stripe, but now my agent is running my Stripe account. Oh, golden. Little things like somebody asked for a refund.
I just tell the agent, you're in charge, hand them a refund. I just saw my agent do it earlier today. So good.
Need to create an active campaign account so that you can email people. I don't know why he's using active campaign here. Because that's probably a good idea.
Meaning like why ActiveCampaign versus any other or even just sending individual messages. I don't think he's dealing with lots of customers. Everything to GitHub so that it's all organized in one place.
I like that. So we'll whack this into GrokBot and see what it comes back with. So we will call this tiny tool bot and we'll make him a green boy.
And then we'll create the bot as always and come in here and paste in this prompt. That's pretty comprehensive. So you build one tiny paid web tool that fixes one annoying job in under a minute.
You sell it for nine bucks a month. You can sell it for whatever you want and feel free to change these however you want. It needs to be a real tool, not a template, not a spreadsheet, not an ebook or a course.
This is really important because the things that you can actually create are pretty powerful. So I would suggest going straight for a tool. Set up once, read X posts where freelancers, small business owners complain about a job that wastes their time.
Show me the best five problems, then build it. One page, one job. And then suggest me a bunch of different domains so that I can hook everything up.
And then every day, look for people with that exact problem, draft a reply, and then make sure that the link to the tool is in the reply. So that's how it advertises the product. So this is what the tiny tool Grokbot has gone off and built.
It's built something called InvoiceBot. The invoice bot chases down unpaid invoices for you and gently nudges your client so that they actually pay you. So this is the live site.
We have invoicebot .ac. I'll leave it linked down below if you want to check it out. And so you can see that the AI chases unpaid invoices for you.
You can make invoices for free and send them out. If you want the bot to do all of the follow -up, that's what you pay the $9 a month. I like that model.
Free invoices, and then if you wanted to chase down your customers, that's when you pay. I like it. This next part, I don't know why the bot built it in.
I think this is AI slop that he probably should remove. You can see it's built on Grokbot, so it even deploys a Grokbot to do all of this for you. Oh, I see.
I don't think you need the Grokbot to do this. It could all be just in code and just let the site go and chase people. Okay.
how it works. You make your invoice. Our AI reminds your client for you.
You get paid. By the way, the reason I don't think the Grokbot should do it is I think once a Grokbot builds it, it should just move out of the way and let the site just run on its own in code. Otherwise, if he runs out of tokens on his Grokbot subscription, then his customers don't get the service they paid for.
Also, it will just burn through tokens needlessly. We take nothing from it. You just pay us nine bucks a month for us to continuously go out and message your clients and say, hey, hey, hey, remember, pay me.
Hey, hey, hey. and so what makes it different because stripe will send reminders right so you can see that on monday the invoice is sent on wednesday it was viewed but not paid and so it'll automatically send a message saying just checking the invoice came through if you have any questions just reply to this and you will get a notification saying that it was viewed but not paid and even down here you can see that it says features coming soon chase them on whatsapp telegram it okay It's interesting.
I think this is a good model. I actually had interviewed an entrepreneur a few months ago who is doing this. At huge scale.
I'm going to link you to that in a moment. But first, I want to say this is interesting. I think there's something in here.
I'd love to know from you whether you find any of these especially helpful, where you think this is going. Let me know in the comments. I will link to not just the long form videos for each one of these, but also to the tools they're using.
So if you want to use one of them, you can go and sign up and get started with them. And of course, here's the video with the founder that I did, Marik, a few months ago before this. was all possible for people like you and me to do easily.
He had raised money and has done this in a big way. Go watch that video next. See you there.
The Hook
The bait, then the rug-pull.
The pitch is agentic income, money that arrives while agents run the business. The video then tests that promise against eight real attempts, from a sponsor-funded directory that made $1,100 in a day to a five-agent team that spent $29.87 and sold nothing.
Named ideas worth stealing.
02:15list
The autonomous micro-business setup
- Create one agent and name it CEO
- Give it your interests and ask for simple problems to solve
- Pick one idea
- Hand it a list of ongoing duties: SEO, chatbot-readable page, newsletter
- Let it schedule daily ops and weekly reports
- Require approval before anything goes out to people or money moves
The sequence used to stand up a sponsor-funded directory that the agent builds and maintains.
Steal forspinning up any small content site or directory run by an agent
32:55list
Tiny tool loop
- Read X posts where freelancers and small businesses complain about a recurring job
- Shortlist the best five problems
- Build one page that does one job
- Charge about $9/month through Stripe
- Reply daily to people with that exact problem, linking the tool
Problem discovery, build and distribution all live inside one agent prompt.
Steal formicro-SaaS ideas validated by existing complaints
33:20list
Agentic software stack
- X for finding problems and replying
- Cloudflare for free hosting and custom domains
- Stripe for payments and refunds
- ActiveCampaign for email
- GitHub to keep everything organized
The accounts to connect before asking an agent to build and operate a paid tool.
Steal foragent-run software side projects
16:10concept
Judgment-only AI
Use AI only for steps that need decisions; move repeatable steps into fixed automations to cut cost and avoid usage limits.
Steal forany agent workflow hitting token or usage caps
11:30concept
Free site as tripwire
Target businesses with many good reviews and no or poor website, build them a site for free, then sell other services once the relationship exists.
Steal forlocal-business agency outreach
How they asked for the click.
VERBAL ASK
37:36next-video
“Here's the video with the founder that I did, Marik, a few months ago... Go watch that video next. See you there.”
Comment prompt plus a next-video handoff to an earlier founder interview on the same tiny-tools model, framed as the at-scale version of the last example.
MENTIONED ON CAMERA
Visual structure at a glance.
