Everything Alex Hormozi Said About Learning AI for 2027 — Transcript

Full transcript

1. 0:00On September 30th, Alex Hormozi told his
2. 0:02team they're not getting any more AI
3. 0:03budgets. They already spend 16 times
4. 0:05more on AI per employee than the average
5. 0:08tech forward company, and in his words,
6. 0:10"Company revenue is flat compared to AI
7. 0:12usage increases." And this is the same
8. 0:14guy who said this in March.
9. 0:16>> AI will never be worse than it is right
10. 0:19now. Learning how to use AI should
11. 0:21become your number one priority, your
12. 0:23number two priority, number three
13. 0:24priority, and your number 10 priority.
14. 0:26>> So what I did is I went through over 200
15. 0:27of his videos and posts to find out how
16. 0:29him and his team really use AI, the
17. 0:31$350,000 mistake that he keeps seeing,
18. 0:34and what he says about cutting people.
19. 0:36So let's get right into it. All right,
20. 0:37so first off, Hormozi is not anti-AI at
21. 0:40all. He's actually said that AI takes up
22. 0:42about half of what he thinks about for
23. 0:44his business. For their book launch, for
24. 0:46example, they spun up five AI agents to
25. 0:48handle support, and most of the tickets
26. 0:49were just people asking where to get
27. 0:51their free stuff.
28. 0:51>> We spun up five agents that handled I
29. 0:55think it was like it's somewhere in the
30. 0:57neighborhood of like 120,000 support
31. 0:59tickets, but we were able to resolve 90%
32. 1:02of all tickets without any human
33. 1:04intervention.
34. 1:05>> And then there's his media team.
35. 1:07>> Like I don't know if you've seen our
36. 1:08media output is like tripled in the last
37. 1:10month. We haven't increased our head
38. 1:11count at all. And we did that because we
39. 1:13have prompts that can do 30, 40, 50%
40. 1:16they're eating up more and more of the
41. 1:17workload, so that one person can now do
42. 1:19the work of three, and then it'll be the
43. 1:20work of five, and then the work of 10.
44. 1:22>> And the way he thinks about getting
45. 1:23there works for basically any role. He
46. 1:26uses video editing as the example here.
47. 1:27>> And so the old thinking, or the old
48. 1:29paradigm, is I need to hire an editor.
49. 1:32The new thinking or paradigm is, "What
50. 1:34are these five things an editor actually
51. 1:35does that creates
52. 1:37a video?" And each one of those things
53. 1:39should be a workflow.
54. 1:39>> So for his Hormozi Hotline clip, an AI
55. 1:41workflow reads the transcript, finds the
56. 1:43highest tension moments, and then starts
57. 1:45the clips right there. And then it
58. 1:46removes all the ums and ahs and cuts it
59. 1:48up, which is pretty much how my own
60. 1:50videos get edited right now. You know, I
61. 1:52record, I hand the footage to Claude,
62. 1:53and it cuts the retakes and the dead air
63. 1:54for me, and I'm building up these
64. 1:55different editing pipelines over time.
65. 1:57He also does the same thing with his
66. 1:58writing, but he's really, really clear
67. 2:00that AI isn't the one actually writing
68. 2:02it.
69. 2:02>> Tweets themselves used to take me about
70. 2:04an hour and a half a day. And so now it
71. 2:06takes me about 90 minutes a week. So
72. 2:08basically I'm seven times more
73. 2:09efficient. I haven't automated this.
74. 2:10It's not AI making my tweets. It's that
75. 2:13it will help me generate ideas from
76. 2:15which to iterate to make tweets that are
77. 2:17good.
78. 2:17>> And his content ideas also come from AI,
79. 2:19too. He records and he transcribes his
80. 2:21meetings, and then he has AI go through
81. 2:22them along with his calendar and pull
82. 2:24out the most interesting moments. So
83. 2:26instead of just making up a story for a
84. 2:28video, he's just talking about real
85. 2:29things that actually happened to him
86. 2:30that week.
87. 2:31>> And stuff like this is super easy to
88. 2:33build for yourself, especially if your
89. 2:34calls are already recorded. So I put
90. 2:36that prompt plus every Hermosi quote
91. 2:38from this video with the link and the
92. 2:39timestamp [music] in a free resource
93. 2:41guide, which you can find inside of my
94. 2:42free school community. The link for that
95. 2:43is down in the description. Now let's
96. 2:45talk about sales. Sales is the one that
97. 2:47you can actually track over time because
98. 2:48he's been predicting this out loud. In
99. 2:50January of 2025, he said an AI
100. 2:52appointment setter was basically like
101. 2:54three months off. And then this in
102. 2:57April.
103. 2:57>> We have an AI SDR right now who's who's
104. 3:01matching our our human team.
105. 3:03>> But in a clip his team posted just a
106. 3:04couple weeks ago, he's a lot more
107. 3:06specific.
108. 3:06>> SDRs, we're still hiring humans because
109. 3:10human is still better than the best AI
110. 3:12that we do have an AI text agent that
111. 3:15sells around 200 units a month.
112. 3:18Just one one agent.
113. 3:20>> So one AI agent is already selling over
114. 3:22text, but on the phone humans are still
115. 3:24winning even after two years of work on
116. 3:26voice AI. And if you guys build this
117. 3:28stuff for clients, then that's a really
118. 3:29useful benchmark just to keep in mind.
119. 3:31But for every business he sees doing
120. 3:32this well, he sees a lot more using AI
121. 3:35in a completely different way. He told
122. 3:37the story on Diary of a CEO about a
123. 3:39business that he just looked at, and
124. 3:41it's basically his whole take on AI in
125. 3:42about 30 seconds.
126. 3:44>> They had 11 VAs that were doing some
127. 3:46sort of like data cleaning work, and I
128. 3:48think it was something in the
129. 3:49neighborhood of like $11,000 a month is
130. 3:51what [music] they were paying for these
131. 3:5211 VAs. And they were doing a good job
132. 3:53and it all worked. They then spent
133. 3:55$350,000
134. 3:56to try and create an AI system that
135. 3:59would replace what these 11 virtual
136. 4:02assistants were doing. And [music] it
137. 4:04wasn't even the constraint of the
138. 4:05business.
139. 4:05>> So, that's 3 years of VA salaries spent
140. 4:07on a process that was working just fine.
141. 4:10Every business has one thing that's
142. 4:11actually capping its growth right now.
143. 4:13It's usually leads, sales, or delivery.
144. 4:15And that money went somewhere else. His
145. 4:18fix here is to ask whether the task
146. 4:19should even exist before you automate
147. 4:21it.
148. 4:21>> It is more efficient for me to determine
149. 4:24that something is not a priority than
150. 4:26for me to automate something that is not
151. 4:28a priority. That was good.
152. 4:30That's pretty good.
153. 4:31>> And his own company ran into the same
154. 4:32problem, which is what that memo was
155. 4:33talking about. He wrote that flat
156. 4:35revenue next to rising AI suggests that
157. 4:37people are using it on projects that
158. 4:39don't matter. So, instead of more
159. 4:41budget, his team gets training on what's
160. 4:43worth working on and how to use fewer
161. 4:45tokens for that same type of output. And
162. 4:47he calls out Claude by name in this next
163. 4:49one, which since I basically live inside
164. 4:50Claude code, is pretty fair.
165. 4:52>> And so, I will promise you that Claude
166. 4:54or, you know, Anthropic or, you know,
167. 4:56OpenAI, etc. If they're billing you,
168. 4:58they're billing you whether what you
169. 5:00built was useful or not.
170. 5:01>> So, here's his test, and it's from a
171. 5:03post that he made the day before the
172. 5:04memo.
173. 5:05If you're using AI all day and you're
174. 5:07not making more money, it means you have
175. 5:08the same problem you had before the AI.
176. 5:11You're doing stuff that doesn't matter.
177. 5:13And then the last line, "Except this
178. 5:15time, it's costing you tokens." So, I'd
179. 5:17have you guys ask yourself three
180. 5:18questions. "Since I started using AI, am
181. 5:21I getting more views, more leads, and or
182. 5:23more money?" If your token bill keeps
183. 5:25going up and those three aren't, then go
184. 5:28find your real constraint and point AI
185. 5:30at that instead. And this is the same
186. 5:31exact mindset that I've been telling you
187. 5:33guys for how to work with your clients
188. 5:35as well. You don't just want to slap AI
189. 5:36onto a business for no reason. You have
190. 5:38to have intent, and you have to be
191. 5:39trying to move a specific number. And
192. 5:41the PS on his memo is great career
193. 5:43advice for anyone on a team. "Show how
194. 5:46little you can use and how much you can
195. 5:48get from it."
196. 5:49>> AI is just a tool. It's not the answer.
197. 5:51>> So, what does all of that mean for the
198. 5:53people on your team? So, who has Hermozi
199. 5:55fired because of AI? Now, I looked
200. 5:57pretty hard and I couldn't find a single
201. 5:59time he said he let someone go because
202. 6:00of it, but what he does say is that
203. 6:02they're still hiring.
204. 6:03>> We're very heavy in terms of using AI
205. 6:05across acquisition.com,
206. 6:07um how we make ads, how we make content.
207. 6:08And so, what that allowed us to do right
208. 6:09now, transparently, is that we're still
209. 6:11hiring people. And to give a little bit
210. 6:13of a proof point around this, look up
211. 6:15how many employees OpenAI has.
212. 6:17Look up how many employees Claude has,
213. 6:19or Anthropic has. Thousands.
214. 6:21>> But he is also really direct about what
215. 6:23he'd tell any other owner.
216. 6:24>> I would encourage you to raise the bar
217. 6:26for the whole company, and the people
218. 6:28who who can meet that new bar get to
219. 6:30stay, and the people who don't don't.
220. 6:32And I'm sorry, and I know that's that's
221. 6:34ugly and that's harsh, but like this is
222. 6:35reality.
223. 6:36>> And he's actually given that advice on
224. 6:37stage. The owner of a $20 million
225. 6:40website company told him AI was killing
226. 6:41his business, and Hermozi told him to
227. 6:43spend 6 months rebuilding his workflows
228. 6:45and cut his head count in half with AI,
229. 6:47which would roughly double his margin.
230. 6:49So, if you're the one on the team, his
231. 6:52advice is to get there first. He calls
232. 6:54it bring your own agent.
233. 6:56>> The job market of the future is going to
234. 6:57be BYOS, which is bring your own
235. 6:59software or bring your own agent. And
236. 7:01so, what that means is you're going to
237. 7:03train agents on doing what you do, so
238. 7:06that you can go to a business and say,
239. 7:07"Hey, instead of hiring a department,
240. 7:09you can just bring me, and I can do all
241. 7:11of it."
242. 7:12>> The goal is that you replace yourself
243. 7:14and don't let someone else do it.
244. 7:15>> And that lines up with what I found when
245. 7:16I surveyed over a thousand people in my
246. 7:18community. The people who'd use AI to
247. 7:20fix a real problem at work multiple
248. 7:21times, 70% of those people had seen a
249. 7:24real career change, like a new role, or
250. 7:26new clients, or a big raise. And he
251. 7:28himself has a great example of this,
252. 7:29too. In July, a woman named Tina was
253. 7:32running cohort programs out of Telegram,
254. 7:34and her team kept answering the same
255. 7:36questions over and over. And he told her
256. 7:39to five-code an agent that searches all
257. 7:40the past answers and suggests one, so
258. 7:42her team can just check it and then hit
259. 7:44send. And a couple months later, she
260. 7:45came back.
261. 7:46>> So, build an AI dashboard so the team
262. 7:48works outside the messaging platform.
263. 7:50>> Absolutely. So, I got that done. I'm out
264. 7:52of Telegram completely, because inspired
265. 7:54by what you told me to do, I ended up
266. 7:56building nine other tools for my
267. 7:58business.
268. 7:59>> His version of how to start is to block
269. 8:01off 4 hours a week, and in each block,
270. 8:03automate or amplify one specific task
271. 8:05that you do, [music] and you just get
272. 8:07faster every time. But, there's one job
273. 8:08Hermosi says you should never hand to
274. 8:10AI, which is the big decisions.
275. 8:12>> And so, every time I'm like, you know, I
276. 8:14think AI's gotten a little better. I'll
277. 8:14just give it a like a really obvious
278. 8:16decision to make. I'll go, "Okay, let's
279. 8:17open up Cloud, let's open up Open AI,
280. 8:19let's let's open all them up." And I'll
281. 8:21ask the same question to all three, and
282. 8:24they're just all over the place. If you
283. 8:25do delegate your decision-making to it,
284. 8:28you'll get dumber.
285. 8:28>> So, his rule, don't outsource your
286. 8:30thinking, only the [music] typing. And
287. 8:32that's really similar to something that
288. 8:33I say a ton. Now, it's not my original
289. 8:34quote, but I do say it a lot, which is
290. 8:36you can outsource your thinking, but
291. 8:38never your understanding. And if
292. 8:39judgment is the thing that AI can't do
293. 8:41for you, then what else can't it copy?
294. 8:43And his answer to that is risk. He
295. 8:45posted earlier this year that in a world
296. 8:47of infinite AI labor and intelligence,
297. 8:49the last valuable thing a human will get
298. 8:52paid to do will be to take risk. And
299. 8:55when Steven Bartlett asked him where the
300. 8:56value goes once intelligence is cheap,
301. 8:58he went right back to it.
302. 8:59>> I think stakes. For example, AI can give
303. 9:02all of the recommendations in the world,
304. 9:03but someone has to own the decision. And
305. 9:05to take this into different domains,
306. 9:06from the media perspective, imagine
307. 9:08Jimmy, say Mr. Beast, has a video, and
308. 9:10they're like, "Mr. Beast is getting
309. 9:12replaced by AI." He's not getting
310. 9:13replaced by AI, because
311. 9:15if he had a video that said $5 million
312. 9:17and the $5 million wasn't real, and the
313. 9:19Lamborghini wasn't real,
314. 9:21there'd be no stakes. So, humans want
315. 9:24stakes. And so, there's the compensation
316. 9:26side of it, which is that somebody's
317. 9:27going to be responsible, someone has to
318. 9:28risk the money, somebody paid for the
319. 9:29tokens, someone owns the LLC, someone is
320. 9:31is owns this.
321. 9:33>> He applies the same idea to content as
322. 9:34well. If two people post the exact same
323. 9:37six tips for scaling a sales team, and
324. 9:39one of them just asked ChatGPT, while
325. 9:41the other one actually built 10 real
326. 9:43sales teams, the second person's going
327. 9:45to win every time.
328. 9:46>> Reality is the mode. When reality is the
329. 9:48mode, it means that your brand is the
330. 9:50mode. It means your reputation is the
331. 9:51mode, which can only happen in reality.
332. 9:53>> And he has zero patience for AI-written
333. 9:54content. He even posted the patterns,
334. 9:56you know, it's not X, it's Y, not A, not
335. 9:58B, not C, but D.
336. 10:00>> I hate that so much of the world has
337. 10:02almost immediately become AI slop. And
338. 10:03the biggest thing [music] you can detect
339. 10:05with it is forced negation and staccato
340. 10:07sentences, meaning it's not this or this
341. 10:10or this, but that.
342. 10:11>> And yes, I of course keep a list of
343. 10:13those exact what I call AI kill phrases
344. 10:16that I won't let into my own scripts or
345. 10:17content as well, and I hope my team
346. 10:19doesn't either. His fix is if you write
347. 10:21with AI is to give it something like 12
348. 10:23rules that it can never break and 16
349. 10:25samples of your own writing. And he says
350. 10:27that this makes the output about five
351. 10:28times better, which I would agree. He's
352. 10:30just as blunt about selling it. He says
353. 10:32that starting an AI company around a
354. 10:33tool you built is probably a terrible
355. 10:35idea and you should just advertise what
356. 10:37the customer wants.
357. 10:38>> The customer hasn't changed. They still
358. 10:39want to lose weight, they still want
359. 10:41things fast, they still want things
360. 10:42easy. They don't care that you have an
361. 10:43AI consultant that analyzes the
362. 10:45transcript to tell them why their
363. 10:47husband's cheating on them. Just say
364. 10:49that you'll help them fix their
365. 10:50marriage.
366. 10:50>> So, what would he do if super
367. 10:51intelligence really was only a few years
368. 10:53away, which it really might be?
369. 10:55Steven asked him exactly that.
370. 10:56>> From an entrepreneurship perspective, I
371. 10:58would continue to build trust and
372. 10:59distribution.
373. 11:01From a personal perspective, I would
374. 11:03think really hard about what it means to
375. 11:05be human
376. 11:06and why I'm here to be in love.
377. 11:08>> So, if you boil down everything that he
378. 11:10said about AI, use it everywhere in your
379. 11:12business, aim it at your one real
380. 11:14constraint, automate your own job before
381. 11:17someone else does,
382. 11:18>> [music]
383. 11:18>> keep the big decisions for yourself, and
384. 11:21spend the time it saves building proof
385. 11:23that AI can't fake. And if you want to
386. 11:25see what that looks like when you're
387. 11:26starting from zero, you'll really like
388. 11:28this video where Anthropic CEO breaks
389. 11:30down how to build a one-person business
390. 11:32with Claude. I'll tag that right up here
391. 11:34and it picks up right where this one
392. 11:35leaves off. But anyways, that is going
393. 11:38to do it for this one. So, if you guys
394. 11:39enjoyed and learned something new,
395. 11:40please give it a like, it definitely
396. 11:41helps me out a ton. And as always, I
397. 11:42appreciate you guys making it to the end
398. 11:43of the video and I'll see you on the
399. 11:44next one.
400. 11:45Thanks, everyone.