--- title: "How Much Should You Spend on AI? Hormozi's Budget Lesson" url: "https://goremy.ai/blog/articles/hormozi-ai-budget-token-spend-lesson" author: Luis Chavez-Mattos editor: Luis Chavez-Mattos editor_title: Director, Product published: "2026-11-13T17:40:10.838Z" updated: "2026-11-13T17:40:17.554Z" topic: AI Pricing & Costs description: "Hormozi froze AI budgets when spend rose and revenue stayed flat. Here is his $350,000 VA story and three questions to audit your own AI spend." --- # How Much Should You Spend on AI? Hormozi's Budget Lesson > Hormozi froze AI budgets when spend rose and revenue stayed flat. Here is his $350,000 VA story and three questions to audit your own AI spend. Edited by Luis Chavez-Mattos, Director, Product ยท Published 2026-11-13 ## How do you know if your AI spending is paying off? Check whether a number you care about moved. Alex Hormozi's test is three questions: since you started using AI, are you getting more views, more leads, or more money? If your AI bill keeps climbing and none of those three are, you are probably using AI on work that doesn't matter. The fix is to find the real bottleneck in the business and point AI at that, instead of buying more usage. ## TL;DR - On September 30, Hormozi told his team they would get no more AI budget, because they already spent **16 times more on AI per employee** than the average tech-forward company while company revenue stayed flat. - AI vendors bill you **whether or not what you built was useful**, so the bill proves nothing about value. - His cautionary story: a business spent **$350,000 on an AI system** to replace 11 virtual assistants costing about $11,000 a month, and the task wasn't even what limited the business. - The audit is simple: **views, leads, money**. If AI use goes up and those don't, find your real constraint. - Before automating a task, ask whether it should **exist at all**. Deciding it isn't a priority is cheaper than automating it. - Hormozi isn't anti-AI. His team uses it heavily where it ties to output, like support tickets and content, and he tells employees to show **how little they can use for how much they get**. ## What did Hormozi's September 30 memo actually say? According to a summary of his public remarks, Hormozi told his team that no more AI budgets would be approved. His reasoning was a comparison: the company already spends 16 times more on AI per employee than the average tech-forward company. His summary of the results was blunt: "Company revenue is flat compared to AI usage increases." His reading of that pattern was that people were probably spending AI resources on projects that don't matter. Instead of more money, the team gets training on two things: what is worth working on, and how to use fewer tokens (the units AI providers charge for) to get the same type of output. The memo's closing note is useful for anyone on a team: show how little you can use and how much you can get from it. This came a few months after he said, in March, that learning to use AI should be your number one priority. The two positions fit together. Use it a lot, but only where it moves something. ## Why did a $350,000 AI project replace a $11,000-a-month process that worked? Hormozi tells a story about a business he looked at. It had 11 virtual assistants (VAs) doing data cleaning work. They cost roughly $11,000 a month, did a good job, and everything worked. The company then spent $350,000 to build an AI system to replace them. Two things went wrong: 1. **The math was bad.** At $11,000 a month, the VAs cost about $132,000 a year. The AI build cost nearly three years of their pay before saving a cent. 2. **It wasn't the constraint.** The data cleaning wasn't what was holding the business back. The money and attention went to a process that was fine, while the actual limit on growth sat untouched. Hormozi says every business has one thing capping its growth at any given time, usually leads, sales, or delivery. Automating anything else is spending without a payoff, however clever the system is. ## Does Hormozi think AI doesn't work? No. The memo is about aim, not about whether the technology is useful. The examples he has shared show what "aimed" looks like: | Area | What he reported | Why it counts | |---|---|---| | Book launch support | Five AI agents handled something in the neighborhood of 120,000 tickets, resolving 90% without a human | High volume, repetitive questions (mostly "where do I get my free stuff") | | Media output | Output roughly tripled in a month with no added headcount | Prompts take over 30 to 50% of the workload per person | | Writing tweets | From about 90 minutes a day to about 90 minutes a week | AI helps generate ideas; it doesn't write the final tweets | | Sales by text | One AI text agent sells around 200 units a month | Measured in sales, not activity | On phone sales, he said his team still hires humans because they beat the best AI voice agent they have, even after two years of work on it. That is a useful reminder that AI doesn't win everywhere, and that you find out by measuring. Each good example has a number attached: tickets resolved, units sold, hours saved, output produced. That is the difference between a project and a hobby. ## What are the three questions to audit your AI spend? Ask them in this order. Hormozi's version is "views, leads, money," which maps to whatever your team is paid to produce. 1. **Since we started using AI, are we getting more views, more leads, or more money?** Translate this to your function. For a support team it might be tickets resolved. For recruiting it might be qualified candidates per week. For finance it might be days to close. 2. **If the bill is rising and those numbers aren't, what is our real constraint?** Usually it is leads, sales, or delivery. Name the one thing that limits growth right now. 3. **Should this task exist at all?** Hormozi's view is that it is more efficient to decide something isn't a priority than to automate something that isn't a priority. If the answer to the first question is no, the answer is not to stop using AI. It is to move it to the constraint you named in the second question. ## How do you run this audit on your own team? You don't need a finance department. A one-page review each month works. Copy this template into a doc or spreadsheet: ``` AI spend review (monthly) For each AI project or tool: - Project / tool name: - Owner: - Monthly cost (subscription + usage): - The one number it is supposed to move: - That number 3 months ago: - That number now: - Is this task a priority, or should it stop existing? - Keep / change / cut: ``` A few rules make it work: - **One number per project.** If you can't name it, that is your finding. - **Compare before and after.** Without a starting figure, every project looks successful. - **Count your own time.** The $350,000 story is a reminder that build effort is a cost too, not only the monthly bill. - **Reward lower usage at equal output.** Hormozi's PS to his team is the standard to adopt: get the same results with fewer tokens. ## Why does the bill not tell you if AI is useful? Hormozi put it plainly: if Anthropic, OpenAI, or any other provider is billing you, they bill you whether what you built was useful or not. Usage-based pricing charges for activity. It has no idea whether that activity produced a lead. This matters most with tools that make it easy to run all day. If a person is using AI constantly and isn't producing more, Hormozi's diagnosis is that they have the same problem they had before AI: they're doing stuff that doesn't matter. The difference is that now it costs tokens. ## What does this mean for you? **If you manage a budget:** Don't approve AI spend on enthusiasm. Ask each owner for the one number their project moves and the before-and-after figure. Pause projects that can't answer. Fund the ones attached to your actual bottleneck first. **If you are an individual on a team:** Treat efficiency as a skill. Hormozi's advice to employees is to show how little you can use and how much you can get from it. Keep a short log of tasks you sped up and by how much (his tweet example went from about 90 minutes a day to about 90 minutes a week). That log is your evidence in a review or a budget request. **If you're thinking about a big automation build:** Ask three things before you start. Is this task the constraint? Does it already work fine? How many years of the current cost does the build equal? If the build costs nearly three years of what you pay now and the task isn't limiting growth, don't start. **If you want to start small:** Hormozi suggests blocking four hours a week and using each block to automate or speed up one specific task you do. That is small enough to measure and cheap enough to stop. Also note what he keeps with humans: big decisions. He says he has asked several AI models the same decision question and gotten very different answers, and that delegating decisions will make you dumber. Outsource the typing, not the thinking. ## Can you build your own AI spend tracker? A review template in a spreadsheet works for a few projects, then gets stale. If you want something your team actually keeps up to date, you could describe a small internal app to [Remy](https://goremy.ai), which builds and deploys full-stack apps from a plain-English description. For example: a page where each owner logs an AI project, its monthly cost, the one number it should move, and the before-and-after values, with sign-in so only your team sees it. Remy's platform fee is flat, and AI tokens are billed at cost, so the tracker's own running cost stays visible too. ## Frequently Asked Questions ### What did Hormozi say about his company's AI spending? On September 30 he told his team there would be no more AI budgets. He said they already spend 16 times more per employee than the average tech-forward company, while company revenue is flat compared to rising AI usage. His team gets training on what is worth working on and how to use fewer tokens. ### Is Hormozi against using AI at work? No. His team uses AI for support, content, and sales texting, and the memo is about aim, not about whether the technology is useful. His objection is to AI use that doesn't connect to views, leads, or money. ### What was the $350,000 mistake? A business paid about $11,000 a month for 11 VAs doing data cleaning that worked well. It then spent $350,000 on an AI system to replace them, even though that work wasn't what limited the business. The lesson is to automate your real constraint, and to ask whether a task should exist before automating it. ### What should I measure to judge my AI ROI? Pick the one number each AI project is meant to move, such as leads, tickets resolved, sales, or time saved, and record it before and after. Hormozi's shorthand is views, leads, and money. If cost rises and none of them improve, change the target. ### Does this mean I should cut AI use? Not necessarily. It means you should cut AI use that doesn't connect to a result and redirect it. Hormozi's own approach is heavy use where output is measurable, plus a push to get the same output with fewer tokens.