The argument in one line.
AI only pays off when you point it at the one real constraint in your business and keep the judgment, risk and real-world proof for yourself.
Read if. Skip if.
READ IF YOU ARE…
- An agency owner or consultant building AI automations for clients who wants a sanity check on which projects actually move revenue.
- A business owner whose AI or token bill keeps climbing while views, leads and money stay flat.
- An employee worried about layoffs who wants a concrete plan for using AI to become the person who stays.
- A creator who uses AI for scripts and posts and wants to know what makes the output read as slop.
SKIP IF…
- You want a tutorial on building a specific agent or workflow; this is a synthesis of opinions and examples, not a build guide.
- You already follow Hormozi closely, since most of this is quotes and clips from material he has put out publicly.
The full version, fast.
Hormozi is not anti-AI. His team resolved 90% of roughly 120,000 support tickets with agents and tripled media output without hiring. But in a September memo he refused more AI budget because revenue is flat while usage climbs, which means people are using it on work that doesn't matter. His rule is to find your one constraint, ask whether a task should exist before automating it, and keep decisions, risk and real-world proof human. He expects raised bars rather than firings, and says to build your own agents before someone else replaces you.
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 – 00:45
01 · The memo that capped AI spend
Hormozi's September memo refuses more AI budget because revenue is flat, set against his earlier call to make AI priority number one.
00:45 – 03:27
02 · How his team actually uses AI
Support agents resolving 90% of tickets, tripled media output, editing as five workflows, tweets cut from 90 minutes a day to 90 a week, and content ideas mined from meetings.
03:27 – 05:37
03 · AI sales, and the $350,000 mistake
An AI text agent sells 200 units a month while humans still win on the phone, then the story of 11 VAs replaced at huge cost for a non-constraint process.
05:37 – 06:45
04 · The token bill test
Hormozi's tests for whether AI use is paying off, and the host's three questions about views, leads and money.
06:45 – 08:43
05 · Jobs, hiring and bring your own agent
He hasn't fired anyone over AI, advises raising the bar, and tells employees to train agents on their own work and replace themselves first.
08:43 – 10:42
06 · Judgment, risk and reality as the moat
Never outsource big decisions, stakes and risk are what humans get paid for, and real experience beats AI-generated advice.
10:42 – 11:47
07 · AI slop, selling AI, and the boiled-down rules
His list of AI writing patterns, the 12 rules plus 16 samples fix, advice to sell the customer's desire, and a closing recap.
Lines worth screenshotting.
- If you use AI all day and aren't making more money, you have the same problem you had before AI: you're doing things that don't matter, now with a token bill.
- One company spent $350,000 on an AI system to replace 11 virtual assistants costing about $11,000 a month, on a process that wasn't even their constraint.
- Ask whether a task should exist before you automate it. Deciding something isn't a priority is cheaper than automating it.
- Instead of hiring an editor, list the five things an editor does and turn each one into a workflow.
- AI providers bill you whether what you built was useful or not.
- Hormozi's team went from about 90 minutes of tweeting a day to 90 minutes a week by using AI for idea generation, not for writing the tweets.
- One AI text agent sells around 200 units a month, but he still hires humans for phone sales because they beat the best voice AI.
- Delegate your decisions to AI and you get dumber. Outsource the typing, not the thinking.
- In a world of infinite AI labor, the last valuable thing a human gets paid for is taking risk, because someone has to own the decision.
- Two people post the same six sales-team tips, but the one who built 10 real sales teams wins every time. Reality is the moat.
- To make AI write like you, give it about 12 rules it can never break and 16 samples of your own writing.
- Start with four hours a week, and in each block automate or amplify one specific task you do.
- Hormozi found no case of him firing someone because of AI; his stance is to raise the bar and let people who meet it stay.
- Starting an AI company around a tool you built is probably a terrible idea; advertise what the customer already wants.
Aim AI at the constraint, keep the stakes
WHAT TO LEARN
AI spend only pays when it targets your single growth constraint, while decisions, risk and real-world proof stay human.
02How his team actually uses AI
- Break a role into the handful of tasks it really does and turn each into a workflow, instead of hiring or automating the whole job.
- Use AI to generate ideas and mine your real meetings for stories, then write the final piece yourself so it stays yours.
03AI sales, and the $350,000 mistake
- AI works well for support triage and text-based sales, but humans still beat it on live phone sales, so match the tool to where it's proven.
- Ask whether a task should exist at all before building automation for it, since a $350,000 build on a non-constraint is pure waste.
04The token bill test
- Check whether AI is raising views, leads or money. If the token bill climbs and those don't, you're automating things that don't matter.
05Jobs, hiring and bring your own agent
- Hormozi hasn't fired anyone over AI, but he advises raising the bar and training agents on your own work so you replace yourself first.
- Start small with four hours a week, automating or amplifying one specific task per block, and let the gains compound.
06Judgment, risk and reality as the moat
- Don't outsource big decisions to AI, because delegating judgment makes you dumber; hand off the typing, not the thinking.
- Humans get paid for taking risk and owning outcomes, and content from someone who has done the real thing beats AI-generated advice.
07AI slop, selling AI, and the boiled-down rules
- To avoid AI slop, give the model about 12 hard rules and 16 samples of your own writing, and ban the 'not X, it's Y' patterns.
Terms worth knowing.
- AI usage cap memo
- An internal note Hormozi posted on September 30, 2026 saying requests for bigger AI budgets would not be approved because spend per employee was already high and revenue was flat.
- Bring your own agent
- Hormozi's name for a future job market where you train agents to do your work and offer a business the output of a department instead of one employee.
- Constraint
- The single thing capping a business's growth right now, usually leads, sales or delivery. AI spend only pays off when aimed here.
- AI SDR
- An AI sales development rep that handles outreach and appointment setting, here as text-based agents rather than voice calls.
- Forced negation
- A common AI writing pattern of the form "it's not X, it's Y" or "not A, not B, but C", which Hormozi treats as the clearest tell of AI-written content.
Things they pointed at.
02:30linkFree school community resource guide with prompt and quotes
03:37channelDiary of a CEO (Steven Bartlett interview with Hormozi)
Lines you could clip.
04:56
“If you're using AI all day and you're not making more money, it means you have the same problem you had before the AI. Except this time, it's costing you tokens.”
08:35
“You can outsource your thinking but never your understanding.”
09:47
“Reality is the moat.”
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.
On September 30th, Alex Ramosi told his team they're not getting any more AI budget. They already spend 16 times more on AI per employee than the average tech forward company. And in his words, company revenue is flat compared to AI usage increases.
And this is the same guy who said this in March. AI will never be worse than it is right now. Learning how to use AI should become your number one priority, your number two priority, number three priority, and your number 10 priority.
So what I did is I went through over 200 of his videos and posts to find out how him and his team really use AI, the $350 ,000 mistake that he keeps seeing, and what he says about cutting people. So let's get right into it. All right, so first off, Hermosi is not anti -AI at all.
He's actually said that AI takes up about half of what he thinks about for his business. For their book launch, for example, they spun up five AI agents to handle support. And most of the tickets were just people asking where to get their free stuff.
We spun up five agents that handled, I think it was like somewhere in the neighborhood of like 120 ,000 support tickets. But we were able to resolve 90 % of all tickets without any human intervention. And then there's his media team.
Like, I don't know if you've seen our media output is like tripled in the last month and we haven't increased our headcount at all. And we did that because we have prompts that can do 30, 40, 50%. They're eating up more and more of the workload so that one person can now do the work of three and then it'll be the work of five and the work of 10.
And the way he thinks about getting there works for basically any role. He uses video editing as the example here. And so the old thinking or the old paradigm is I need to hire an editor.
The new thinking or paradigm is what are these five things an editor actually does that creates a video and each one of those things should be a workflow so for his hermosi hotline clip an ai workflow reads the transcript finds the highest tension moments and then starts the clips right there and then it removes all the ums and ahs and it up which is pretty much how my own videos get edited right now you know i record i hand the footage to claude and it cuts the retakes and the dead air for me and i'm building up these different editing pipelines over time He also does the same thing with his writing, but he's really, really clear that AI isn't the one actually writing it.
Tweets themselves used to take me about an hour and a half a day. And so now it takes me about 90 minutes a week. So basically I'm seven times more efficient.
I haven't automated this. It's not AI making my tweets. It's that it will help me generate ideas from which to iterate to make tweets that are good.
And his content ideas also come from AI too. He records and he transcribes his meetings, and then he has AI go through them along with his calendar. and pull out the most interesting moments.
So instead of just making up a story for a video, he's just talking about real things that actually happened to him that week. And stuff like this is super easy to build for yourself, especially if your calls are already recorded. So I put that prompt plus every Hermosi quote from this video with the link and the timestamp and a free resource guide, which you can find inside of my free school community.
The link for that is down in the description. Now let's talk about sales. Sales is the one that you can actually track over time because he's been predicting this out loud.
In January of 2025, he said an AI appointment setter was basically like three months off. And then this in April. We have an AI SDR right now who's who's matching our human team.
But in a clip his team posted just a couple of weeks ago, he's a lot more specific. SDRs, we're still hiring humans because human is still better than the best AI that we do have an AI text agent that sells around 200 units a month. Just one agent.
So one AI agent is already selling over text, but on the phone, humans are still winning even after two years of work on voice AI. And if you guys build this stuff for clients, then that's a really useful benchmark just to keep in mind. But for every business he sees doing this well, he sees a lot more using AI in a completely different way.
He told the story on Diary of a CEO about a business that he just looked at. And it's basically his whole take on AI in about 30 seconds. They had 11 VAs that were doing some sort of like data cleaning work.
And I think it was something in the neighborhood of like $11 ,000 a month is what they were paying for these 11 VAs. And they were doing a good job and it all worked. They then spent $350 ,000 to try and...
create an AI system that would replace what these 11 virtual assistants were doing. And it wasn't even the constraint of the business. So that's three years of VA salaries spent on a process that was working just fine.
Every business has one thing that's actually capping its growth right now. It's usually leads, sales, or delivery. And that money went somewhere else.
His fix here is to ask whether the task should even exist before you automate it. It is more efficient for me. to determine that something is not a priority than for me to automate something that is not a priority.
That was good. That was pretty good. And his own company ran into the same problem, which is what that memo is talking about.
He wrote that flat revenue next to rising AI suggests that people are using it on projects that don't matter. So instead of more budget, his team gets training on what's worth working on and how to use fewer tokens for that same type of output. And he calls out Cloud by name in this next one, which, since I basically live inside Cloud Code, is pretty fair.
And so I will promise you that Claude or, you know, Anthropic or, you know, OpenAI, etc. If they're billing you, they're billing you whether what you built was useful or not. So here's his test, and it's from a post that he made the day before the memo.
If you're using AI all day and you're not making more money, it means you have the same problem you had before the AI. You're doing stuff that doesn't matter. And then the last line, except this time, it's costing you tokens.
So I'd have you guys ask yourself three questions. Since I started using AI, Am I getting more views, more leads, and or more money?
If your token bill keeps going up and those three aren't, then go find your real constraint and point AI at that instead. And this is the same exact mindset that I've been telling you guys for how to work with your clients as well. You don't just want to slap AI onto a business for no reason.
You have to have intent and you have to be trying to move a specific number. And the PS on his memo is great career advice for anyone on a team. show how little you can use and how much you can get from it.
AI is just a tool. It's not the answer. So what does all of that mean for the people on your team?
So who has Hermosi fired because of AI? Now, I looked pretty hard and I couldn't find a single time he said he let someone go because of it. But what he does say is that they're still hiring.
We're very heavy in terms of using AI across acquisition .com, how we make ads, how we make content. And so what's that allowed us to do right now, transparently, is that we're still hiring people. And to give a little bit of a proof point around this, Look up how many employees OpenAI has.
Look up how many employees Claude has or Anthropic has. Thousands. But he is also really direct about what he'd tell any other owner.
I would encourage you to raise the bar for the whole company and the people who can meet that new bar get to stay and the people who don't, don't. And I'm sorry and I know that's ugly and that's harsh, but like this is reality. And he's actually given that advice on stage.
The owner of a $20 million website company told him AI was killing his business, and Hermosi told him to spend six months rebuilding his workflows and cut his headcount in half with AI, which would roughly double his margin. So if you're the one on the team, his advice is to get there first. he calls it bring your own agent the job market of the future is going to be byos which is bring your own software or bring your own agent and so what that means is you're going to train agents on doing what you do so that you can go to a business and say hey instead of hiring a department you can just bring me and i can do all of it the goal is that you replace yourself and don't let someone else do it and that lines up with what i found when i surveyed over a thousand people in my community the people who'd used ai to fix a real problem at work several times 70 of those people had seen a real career change, like a new role or new clients or a big raise.
And he himself has a great example of this too. In July, a woman named Tina was running cohort programs out of Telegram and her team kept answering the same questions over and over. And he told her to vibe code an agent that searches all the past answers and suggests one so her team can just check it and then hit send.
And a couple months later, she came back. So build an AI dashboard so the team works outside the messaging platform? Absolutely.
So I got that done. I'm out of Telegram completely. Because inspired by what you told me to do, I ended up building nine other tools for my business.
His version of how to start is to block off four hours a week. And in each block, automate or amplify one specific task that you do. And you just get faster every time.
But there's one job Hermosi says you should never hand to AI, which is the big decisions. And so every time I'm like, you know, I think he has gotten a little better. I'll just give it like a really obvious decision to make.
I'll go, okay, let's open up a claw. Let's open up, open AI. Let's like, let's open all of them up and I'll ask the same question to all three.
And they're just all over the place. If you do delegate your decision -making to it, you get dumber. So his rule, don't outsource your thinking, only the typing.
And that's really similar to something that I say a ton. Now it's not my original quote, but I do say it a lot, which is you can outsource your thinking, but never your understanding. And if judgment is the thing that AI can't do for you, then what else can't it copy?
And his answer to that is risk. He posted earlier this year that in a world of infinite AI labor and intelligence, the last valuable thing a human will get paid to do will be to take risk. And when Stephen Bartlett asked him where the value goes once intelligence is cheap, he went right back to it.
I think stakes. For example, AI can give all of the recommendations in the world, but someone has to own the decision. And to take this into different domains from the media perspective, imagine Jimmy said, Mr.
Beast has a video and we're like, Mr. Beast is going to replace by AI. He's not going to replace by AI because if he had a video that said $5 million and the $5 million wasn't real and the Lamborghini wasn't real, there'd be no stakes.
So humans want stakes. And so there's the compensation side of it, which is that somebody's going to be responsible. Someone has to risk the money.
Somebody paid for the token. Somebody owns the LLC. Someone owns this.
He applies the same idea to content as well. If two people post the exact same six tips for scaling a sales team, and one of them just asked ChatGPT while the other one actually built 10 real sales teams, the second person is going to win every time. Reality is the moat.
When reality is the mode, it means that your brand is the mode. It means your reputation is the mode, which can only happen in reality. And he has zero patience for AI written content.
He even posted the patterns, you know, it's not X, it's Y, not A, not B, not C, but D. I hate that so much of the world has almost immediately become AI slop. And the biggest thing you can detect with it is forced negation and staccato sentences, meaning it's not this or this or this, but that.
And yes, I, of course, keep a list of those exact. what I call AI kill phrases that I won't let into my own scripts or content as well. And I hope my team doesn't either.
His fix is if you write with AI is to give it something like 12 rules that it can never break and 16 samples of your own writing. And he says that this makes the output about five times better, which I would agree. He's just as blunt about selling it.
He says that starting an AI company around a tool you built is probably a terrible idea and you should just advertise what the customer wants. The customer hasn't changed. They still want to lose weight.
They still want things fast. They still want things easy. They don't care that you have an AI consultant that analyzes the transcript to tell them why their husband's cheating on them.
Just say that you'll help them fix their marriage. So what would he do if superintelligence really was only a few years away, which it really might be?
Stephen asked him exactly that. From an entrepreneurship perspective, I would continue to build trust and distribution. From a personal perspective, I would think really hard about what it means to be human and why I'm here to begin with.
So if you boil down everything that he said about AI, use it everywhere in your business. Aim it at your one real constraint. Automate your own job before someone else does.
keep the big decisions for yourself, and spend the time it saves building proof that AI can't fake. And if you want to see what that looks like when you're starting from zero, you'll really like this video where Anthropic's CEO breaks down how to build a one -person business. With Claude, I'll tag that right up here, and it picks up right where this one leaves off.
But anyways, that is going to do it for this one. So if you guys enjoyed, you learned something new, please give it a like. It definitely helps me out a ton.
And as always, I appreciate you guys making it to the end of the video, and I'll see you on the next one. Thanks, everyone.
The Hook
The bait, then the rug-pull.
Alex Hormozi told his team no more AI budget, even though he is the person who said AI should be your number one priority. This video goes through 200+ of his videos and posts to square those two positions.
Named ideas worth stealing.
05:21list
Three questions for your token bill
- Am I getting more views?
- Am I getting more leads?
- Am I getting more money?
If the AI bill keeps rising and none of the three move, find the real constraint and point AI there instead.
Steal forany AI project review
01:37model
Break a job into workflows
Replace 'hire an editor' with a list of the five things an editor does, then make each one a workflow.
Steal forscoping any role before hiring or automating
07:52model
Four hours a week
Block four hours a week and in each block automate or amplify one specific task you do.
Steal forpersonal AI adoption
How they asked for the click.
VERBAL ASK
11:26next-video
“If you want to see what that looks like when you're starting from zero, you'll really like this video where Anthropic's CEO breaks down how to build a one-person business.”
Soft hand-off to a related video, followed by a like-and-subscribe ask.
Visual structure at a glance.
