Why AI Automation Starts With Tasks — Transcript

Full transcript

1. 0:00AI probably occupies half of my time
2. 0:02right now
3. 0:03in terms of like what I'm thinking about
4. 0:05within the business.
5. 0:11And so
6. 0:18Okay.
7. 0:20Okay, so from a beliefs perspective,
8. 0:22right? Like one of the big shifts that
9. 0:23everybody here has to make is shifting
10. 0:25away from thinking about roles based
11. 0:27growth and expansion
12. 0:29towards workflow based automation. So
13. 0:33Imagine that you're you're like I want
14. 0:35to hire
15. 0:37I want to hire an editor. All right, I'm
16. 0:38just making this up. I want to hire an
17. 0:40editor.
18. 0:41This is how we used to think about
19. 0:43things. I need to hire an editor so they
20. 0:44can do XYZ.
21. 0:46The way that we all have to think about
22. 0:47our businesses now is that this editor
23. 0:50really was going to do
24. 0:53these four or five things, right? And
25. 0:55each of these things should exist
26. 0:59in a workflow. Like the reason that like
27. 1:01and this is the editor is this one and
28. 1:03they're doing this one
29. 1:05in this workflow and maybe they're doing
30. 1:06these three in this workflow. And so
31. 1:09basically you have to go more granular
32. 1:12to the level of like because every
33. 1:13single business, if you really think
34. 1:15about it, can be connected at the most
35. 1:16intricate level of like this action to
36. 1:18this action to this action to this
37. 1:19action all the way through. Now might be
38. 1:21120 actions that go from
39. 1:23inbound lead to dirt, you know, turf,
40. 1:26whatever, right? And it's getting great
41. 1:29and this is the thinking that most
42. 1:30people won't do, which is why they will
43. 1:31lose and why I think a lot of people are
44. 1:32going to get [ __ ] But if you have a
45. 1:33little bit more horsepower and have a
46. 1:34little bit more time to think about
47. 1:36things, that's where instead of saying I
48. 1:38need to hire an editor, you say I need
49. 1:40an agent who just does this and I need
50. 1:42an agent that just does this and an
51. 1:44agent that just does this. And so it's
52. 1:47switching from
53. 1:49again roles to workflows.
54. 1:52And everything just has to be organized
55. 1:53in workflows. And so it's like a very
56. 1:55fun project is thinking like can you
57. 1:57actually draw your entire business in
58. 1:59one linear workflow? Cuz it does happen
59. 2:01and if you can't draw it,
60. 2:03how the [ __ ] do you have a business,
61. 2:04right? But it really starts with like
62. 2:09Like the business for us, for example,
63. 2:10starts with an idea. It sounds silly,
64. 2:12but like it starts with an idea. So, how
65. 2:14does an idea come to happen?
66. 2:16Well, the what is the workflow around an
67. 2:18idea? For us, it would be like, okay, um
68. 2:21I check our school community and I look
69. 2:24at what's happening. Okay, that is a
70. 2:26workflow, which is
71. 2:28uh search, you know, search this
72. 2:31database, which is school, but it could
73. 2:33be, you know, trends.com or you know,
74. 2:35one of 10.com, whatever.
75. 2:37Um where you look at what things are
76. 2:39outperforming in this particular market.
77. 2:41And then it says, okay, then it gives
78. 2:43you this list
79. 2:44of the outperformers. And then there's
80. 2:46going to be a match between this list
81. 2:48and our competency. So, it's like I have
82. 2:51a list of things that I like talking
83. 2:52about. These are list of things that
84. 2:54people want to hear about. This is the
85. 2:55intersection between those things.
86. 2:57Great.
87. 2:59What does this look like in terms of uh
88. 3:01uh scripting, right? So, it's like,
89. 3:03okay, we're going to have some sort of
90. 3:04scripting around this.
91. 3:05So, I want uh templated ideas. AI does
92. 3:08this. So, it gives us
93. 3:09um okay, this is a seven-step, whatever.
94. 3:12And it has each of the steps underneath
95. 3:14of it and we like, great. Now that the
96. 3:15scripting's done,
97. 3:17we're going to go into
98. 3:18packaging.
99. 3:20All right, what are going to be the
100. 3:20headlines, what are going to be the
101. 3:21thumbnails?
102. 3:24This is an AI workflow.
103. 3:25And then we're going to get into
104. 3:26production, right? And so all of this is
105. 3:29broken down just so that we can get one
106. 3:30thing to happen, which is a click.
107. 3:32Right, but every single element of the
108. 3:34business can be broken down into the
109. 3:36into its constituent workflows. And so,
110. 3:38I think the mistake that a lot of people
111. 3:39are making right now with AI is being
112. 3:40like, well, AI can't do what a human can
113. 3:42do, which is obviously [ __ ] but
114. 3:44like
115. 3:46you have to train it the way you train a
116. 3:47human.
117. 3:48So, saying like, well, I'm I'm to bring
118. 3:50this this human in and tell them nothing
119. 3:52and expect them to do a good job is
120. 3:53ridiculous, but we're somewhere else we
121. 3:55like believe that a machine should be
122. 3:56able to do that. But instead it's
123. 3:57thinking, okay, this editor or
124. 4:00videographer or whatever it is,
125. 4:02um actually does 16 actual activities
126. 4:07and each of those activities we have to
127. 4:09take one at a time and we can have an
128. 4:10agent that's trained on that thing.
129. 4:13And so like right now like Nomadic is
130. 4:14working super hard with Julian on this,
131. 4:16but like
132. 4:17we have so I do like my Hormozi
133. 4:18hotlines. And so I do a hotline and then
134. 4:21that then goes into
135. 4:23um
136. 4:24basically an AI thing that does uh that
137. 4:26breaks down the the the transcripts.
138. 4:28That's like step one. From the
139. 4:29transcript it then says, okay, notice
140. 4:31when there's different speakers. All
141. 4:33right? So it has to do that two times.
142. 4:35One is notice when
143. 4:36it's Alex versus the other person. So
144. 4:38it's ABABAB, imagine I'm A, but then it
145. 4:41eventually switches to ACACAC and then
146. 4:43it goes to ADADAD because there's new
147. 4:47person that I'm on the phone with. So
148. 4:48it's going to say segment first between
149. 4:51when it goes from AB to AC. So great,
150. 4:53now we have 10 underneath clips. From
151. 4:55there uh it'll run a separate script
152. 4:57that's like look cuz it's like what
153. 4:59would an editor do? The first thing
154. 5:00you're going to clip out the individual
155. 5:01portions. Okay, there's 10 different
156. 5:03conversations Alex had. Okay. Now that
157. 5:04we have these 10 conversations, look
158. 5:06through this and try to find the highest
159. 5:07tension moment. And so it'll read the
160. 5:09words and say this is the highest
161. 5:11tension moment. Be great, start the clip
162. 5:13with the highest tension moment and then
163. 5:15get into the rest of the clip. What else
164. 5:17do we want? Remove all ums and ahs.
165. 5:19Great. Now from there, what are the um
166. 5:22what data is required in order to have
167. 5:24the insight that Alex delivers? Only
168. 5:26these four points. Great, collapse the
169. 5:28rest. Now export final file of that.
170. 5:32That's the workflow. And so it's really
171. 5:34just building this on a consistent basis
172. 5:36and continuing to update that prompt,
173. 5:37which is literally just like updating
174. 5:38the training for a human being of like
175. 5:40actually do this too, just like you
176. 5:41would with an editor, but doing that
177. 5:43across every function of the business.
178. 5:46So that is like that is the future. And
179. 5:48the nice thing is that with
180. 5:50there's a lot of businesses that
181. 5:52like agencies, for example, have been a
182. 5:54notoriously shitty business
183. 5:56to run.
184. 5:57Um
185. 5:59but in the future, I think they'll be
186. 6:01incredibly lucrative businesses for the
187. 6:03people who are really good with AI
188. 6:04because
189. 6:06the demand for more customers is
190. 6:07infinite.
191. 6:09And so there's like no one's like, "Hey,
192. 6:10what's product market fit for an
193. 6:11agency?" You don't have to ask. People
194. 6:13want customers. It's not hard to like
195. 6:14you don't need to ask for product market
196. 6:15fit. The issue is that
197. 6:18the ops around being good at getting
198. 6:20customers is significantly harder. But
199. 6:22if you can operationalize each of the
200. 6:23flows, it becomes much better. And this
201. 6:25is why like
202. 6:27like I'm built for this movement because
203. 6:29it's all about being black and white
204. 6:31with observable behavior. And that's
205. 6:32like what I geek out on. And so this is
206. 6:35where vague language like machines do
207. 6:37not understand like vague language.
208. 6:38Also, neither do you [ __ ] humans.
209. 6:40Um we just nod our heads because we've
210. 6:42been reinforced for doing so. But this
211. 6:43is where being very specific about like,
212. 6:45"Hey, I need you to be more
213. 6:45charismatic." means nothing. We have to
214. 6:47say, "I need you to raise your voice in
215. 6:49this way. I need you to talk faster. I
216. 6:50need you to raise your head when you're
217. 6:51speaking. I need you to nod when the
218. 6:52other person's talking." We have to
219. 6:53break it down into what someone would
220. 6:55do. And so um as though you were
221. 6:57teaching a toddler, you have to work the
222. 6:59same way with the machine, but the
223. 7:00machine never forgets.
224. 7:02Hopefully that answered the question.
225. 7:04>> It does.
226. 7:04>> Okay.
227. 7:06But yeah, this is what I'm thinking a
228. 7:07lot about.
229. 7:07>> Yeah.
230. 7:10So, it starts with a clear process map
231. 7:12or workflow map and then every action in
232. 7:14the business.
233. 7:14>> Exactly. And I would I would start like
234. 7:16start start with the big ideas first.
235. 7:19So, it's like, "Okay, I do I I I have an
236. 7:21idea."
237. 7:22And then I make a script and then I uh I
238. 7:25shoot it and then I you know, then we we
239. 7:27export the clip and then we edit it and
240. 7:29then we post it. Okay, that's the that's
241. 7:31the big picture. But
242. 7:33now underneath of each of those, there's
243. 7:34probably six or seven things that
244. 7:36actually happen in order to post it.
245. 7:39It's like, "So, I actually go to this
246. 7:40page. I look at the schedule of what
247. 7:42was, you know, last posted, I add 3
248. 7:44hours to it. I like you just have to
249. 7:46like
250. 7:47one word turns into six six actions. And
251. 7:50it's just continuing down until there's
252. 7:52you can't reduce it anymore, and then
253. 7:53that's what you automate.
254. 7:56>> Do you have a special AI implementation
255. 7:57team to build this?
256. 7:58>> Uh yeah, we have a full dev team. Yeah.
257. 8:00>> Yeah, but they're pretty good at all
258. 8:02this stuff.
259. 8:03>> Yeah. So, I think I think many of you,
260. 8:05especially those of you who are online,
261. 8:08uh an amazing high-leverage hire is just
262. 8:10an AI like an AI automation person.
263. 8:13Um they will not lack for work to do.
264. 8:15Like they can just sit next to teams and
265. 8:17just watch what they're doing and then
266. 8:18just say like I can automate that. I can
267. 8:20automate that. I can automate that.
268. 8:23>> All right.
269. 8:24Is that something that we should do
270. 8:25right now?
271. 8:27>> Yeah.
272. 8:27>> That hire?
273. 8:28>> Yeah.
274. 8:29If you have the cash flow for it. Mhm.
275. 8:31>> What does a dev hire run?
276. 8:33>> Mhm?
277. 8:33>> What does a
278. 8:34somebody like that run?
279. 8:36>> Depends on how good they are. I think
280. 8:38though as Lindsay Lohan said it best,
281. 8:40limits do not exist. So,
282. 8:42like there there's a there's a
283. 8:44there's a price tag for every level of
284. 8:45skill you want.
285. 8:47But even just somebody who would I would
286. 8:48just consider like a hacker, and I say
287. 8:50hacker like loosely, like somebody who
288. 8:51just duct tapes tools together,
289. 8:53probably like 150 grand a year.
290. 8:58>> And then in terms of prioritizing for
291. 8:59us, would you focus on the delivery side
292. 9:01first?
293. 9:02>> I think that you getting um
294. 9:06the elements that we walked through
295. 9:07specifically around the expectation
296. 9:09setting number one, um
297. 9:12I think CAC expectations number two,
298. 9:14and then the controllables around brand.
299. 9:17Like the the brand controllables, I
300. 9:18think that should be largely automated.
301. 9:21Um because it is automatable. It's
302. 9:23literally pattern detection, which is
303. 9:24like machines do it better than humans
304. 9:26do anyways.
305. 9:27And so, what what humans do shitty at is
306. 9:29giving the pattern to recognize. And so,
307. 9:32that's like in terms of like what skill
308. 9:33will you need to be get good at is when
309. 9:36someone says like this is off brand, be
310. 9:37like what the [ __ ] does that mean? Like
311. 9:39tell me in describables so that
312. 9:41otherwise how am I if you can't tell me,
313. 9:43how am I going to know?
314. 9:45So it has to be in black and white
315. 9:46language.
316. 9:50Yeah.
317. 9:51I think for like home services and brick
318. 9:53and mortar, a lot of um basically all
319. 9:56administrative work should just not
320. 9:58really be a thing.
321. 9:59Like, you know, invoicing, matching
322. 10:02billables, like back office for sure is
323. 10:05or like that's already there. Like
324. 10:06that's like now.
325. 10:08Um
326. 10:09receivables, that's like now. Like you
327. 10:12don't need to There's nothing there.
328. 10:13Lead nurture,
329. 10:15that's now. Like you don't
330. 10:17Like a lot of that stuff's already
331. 10:18already kind of done. The advantage that
332. 10:21um y'all have for in person is that like
333. 10:23I don't know how horses will respond to
334. 10:25robots. So you probably have a little
335. 10:27bit of You probably have a little bit of
336. 10:28security, you know, like
337. 10:30otherwise you're getting you know mule
338. 10:31kicks into the into the side of the
339. 10:33stall.
340. 10:34>> [laughter]
341. 10:34>> Um
342. 10:35uh
343. 10:37I think that from a landscaping
344. 10:38perspective,
345. 10:40there will be kind of like once Optimus
346. 10:42comes out and you can teach it how to
347. 10:43like mow lawns and [ __ ] like that'll be
348. 10:44a thing for sure.
349. 10:46Um
350. 10:47Hm?
351. 10:49Yeah.
352. 10:50Yeah, yeah, I mean it can do it.
353. 10:52Um there'll also be I think like job
354. 10:54specific ones, but most likely I think
355. 10:56the biggest proliferation's going to be
356. 10:57around generalized humanoid robots
357. 10:58because the world is already built for
358. 10:59us. And so like as much as it's like,
359. 11:02you know, I'm sure you've seen like the
360. 11:03like the Roomba of lawn care, I'm sure
361. 11:06you've seen these things where it's like
362. 11:07you drop it around, but like I haven't
363. 11:09seen anyone really take off with that.
364. 11:11At least I haven't seen it. And I think
365. 11:13it's just because like there's too many
366. 11:14unique scenarios, it costs too much like
367. 11:16the cost doesn't make sense yet. Maybe
368. 11:18in three or four generations it'll be
369. 11:20there, but in the meantime the skeleton
370. 11:23key for the world is just the way the
371. 11:24human body works. Door handles,
372. 11:26everything's already built for that. So
373. 11:28it's like as silly as it sounds, it's
374. 11:29like the The already We already built
375. 11:30the world for ourselves.
376. 11:32If there's a robot that can fit
377. 11:33underneath the door, we could make a
378. 11:35special purpose robot that does that,
379. 11:36but it's like you'd have to do that in
380. 11:37every door in every building. Or you
381. 11:40just build a [ __ ] human robot, which
382. 11:41is what everyone's doing.
383. 11:44Yeah, I think I think you're fine for
384. 11:45the short term. The The stuff that you
385. 11:46should be focusing on from an automation
386. 11:48perspective is going to be back office.
387. 11:50Yeah. And there's And the thing is it's
388. 11:52great at that. It's [ __ ] awesome at
389. 11:54it. And you should totally use it for
390. 11:55that.
391. 12:00The UAJ.
392. 12:01>> Yeah, like would that be something like
393. 12:02using AR for like the nurture like that
394. 12:05stuff and like
395. 12:05>> 100%.
396. 12:09Right now, voice is almost there. This
397. 12:11is supposed to be the year of voice.
398. 12:13Um it's close. It's not.
399. 12:16There's a lot of tools out there. I
400. 12:17won't make any recommendations cuz like
401. 12:19they change all the time.
402. 12:21Um I would say voice is very close.
403. 12:23Text is there.
404. 12:25>> Yeah, I [snorts] was just thinking about
405. 12:26like
406. 12:27off hours having
407. 12:29like a voice like auto
408. 12:32auto attendant.
409. 12:34>> I went to a gym
410. 12:35that was local here. It's a It's like a
411. 12:37very exclusive like whatever. I think
412. 12:39it's like 10 grand a year or something
413. 12:40like that. It has like
414. 12:42400 members and that's the deal.
415. 12:44Um pretty interesting model. But I went
416. 12:47to the site and it was just a chat
417. 12:48interface. It just said like what do you
418. 12:50want to know?
419. 12:51And so I was like hours and it was like
420. 12:52here's the hours. I was like what else
421. 12:53do you want to know? And I was like
422. 12:54membership prices and it was like here's
423. 12:56the membership prices. What else do you
424. 12:57want to know? I was like uh
425. 13:00I mean it should have asked me for my
426. 13:01information, but
427. 13:03It's not per It was not well trained
428. 13:04enough.
429. 13:04>> [laughter]
430. 13:05>> Um but like I was like this is pretty
431. 13:07slick.
432. 13:11Like why navigate when it just already
433. 13:12has the answers?
434. 13:16Not to get very odd for you guys, but in
435. 13:17the next I would say 12 to 24 months a
436. 13:19lot of like the UX is going to
437. 13:21disappear.
438. 13:22Like we're not going to be clicking
439. 13:23things.
440. 13:24Like the mouse Like what brought
441. 13:27computers into mainstream was a mouse.
442. 13:30Like that was the big innovation was
443. 13:32having a mouse you could point and click
444. 13:33and then all of a sudden people are
445. 13:33like, "Oh, I just want that thing." It's
446. 13:34like click and it worked. And so,
447. 13:37nothing's really kind of happened in the
448. 13:39computational world besides speed and
449. 13:40power and all that stuff, but it's all
450. 13:41been
451. 13:42um more of the same. The next phase
452. 13:45shift is that UX disappears, which is
453. 13:47that you just have an agent that you
454. 13:49just talk to.
455. 13:50Right? And you say, "Hey, you know, what
456. 13:51did we do in cash flow last month?" And
457. 13:53you just text it or you say it and it
458. 13:55just answers.
459. 13:56And like the world of CRMs is going to
460. 13:58Like this is This is the next 12 to 24.
461. 14:00Yeah.
462. 14:01Yeah.
463. 14:03>> Don't you think that we are very into
464. 14:04it, but 99% of people are still
465. 14:06>> Yeah. Oh, yeah. I mean, there are legal
466. 14:08firms that use fax machines now. And so,
467. 14:11there's this idea that like, "Oh my god,
468. 14:12the whole world." It's like, "No, just
469. 14:13like always, humans plus technology will
470. 14:16beat humans with inferior technology."
471. 14:18And so, if you learn how to use the
472. 14:19tech, you'll do better. But there will
473. 14:21be people who still have law firms and
474. 14:23use fax machines that will make money.
475. 14:25It's just there will be people who don't
476. 14:26use fax machines that will make more
477. 14:28money.
478. 14:32So, I'm not like an AI alarmist. I think
479. 14:33it's going to take a while, but
480. 14:35um
481. 14:36I do think there's I won't even get into
482. 14:38that, so I'll just stop, but
483. 14:40we just get too much of a future no one
484. 14:41knows anything about. I can't give you
485. 14:42every single SOP, that would be
486. 14:43ridiculous. It would be like this long,
487. 14:44[music] but these are the top five SOPs
488. 14:46that we have used to scale that Leila
489. 14:48actually put together [music] for us cuz
490. 14:50she's really been the one who operates
491. 14:51the business. And so, yeah, I don't know
492. 14:53what's inside of it, but um it's good.
493. 14:55So, be sure to check it out cuz Leila
494. 14:56made [music] it and she makes good
495. 14:57stuff. It's her.