YouTube2Text

Как создать ИИ маркетолога: парсит, разбирает конкурентов и создаёт мощный контент | ИИ Сотрудник — Transcript

by Игорь Зуевич · 5,076 words · 764 segments · language en · Watch on YouTube

Full transcript

  1. 0:00In my opinion, a marketer should do
  2. 0:01three things. First, they should
  3. 0:03collect data. Second, they should
  4. 0:05analyze this data. And third, they
  5. 0:07should create content based on this
  6. 0:08analysis, based on the data generated.
  7. 0:11In this video, I'll show you how I
  8. 0:13assembled my own marketer who collects
  9. 0:15data. For example, he analyzed or
  10. 0:17conducted marketing research in the
  11. 0:20electrical services niche in St.
  12. 0:22Petersburg and gave me this score,
  13. 0:24which completely describes what's
  14. 0:26happening in the electrical services
  15. 0:28market in St. Petersburg, meaning who
  16. 0:31provides what services, who the leaders
  17. 0:33are, and which companies. He analyzed
  18. 0:36not just a website, but also Avito,
  19. 0:38which seems impossible to analyze. He
  20. 0:41analyzed photos of Proruns and analyzed
  21. 0:44Yandex services. He analyzed over 4,000
  22. 0:47different ads and, accordingly, came to
  23. 0:51me with this report. The second thing
  24. 0:54he should do is, he's collected the
  25. 0:56data, he's completed the first part of
  26. 0:58the work, he can perform this part of
  27. 1:00the work based on a request we send him
  28. 1:03in a Telegram chat. All my agents work
  29. 1:05through Telegram. It's easier for me to
  30. 1:07interact with them this way. It's code,
  31. 1:10it's antigravity, it's code, it doesn't
  32. 1:12matter. So, all my agents work through
  33. 1:14Telegram, not through a terminal. It's
  34. 1:16inconvenient for me to interact with
  35. 1:18them through a terminal, for example,
  36. 1:21from a phone. Okay? So, he's done the
  37. 1:23research for me. So, he found the data
  38. 1:25I needed. The second thing he should do
  39. 1:28is conduct a competitor analysis. What
  40. 1:31did he do here? He came to me with this
  41. 1:34report as a competitor analysis. We
  42. 1:37took several YouTube videos and looked
  43. 1:40at what was in those YouTube videos.
  44. 1:42Why were these YouTube videos, or SHO
  45. 1:45videos, TikTok videos, Instagram ReS
  46. 1:47videos, why did they get so many views?
  47. 1:49So, he analyzed the hook, he looked at
  48. 1:52the structure, he looked at what drives
  49. 1:55audience retention, what strong wording
  50. 1:58was used in this particular video. We
  51. 2:01analyzed literally several videos. The
  52. 2:04more videos, the more, well, data for
  53. 2:06analysis you have, the better,
  54. 2:09accordingly, he will do this work.
  55. 2:11That's it. So, we now know what works
  56. 2:14in our niche, in this case, the
  57. 2:16electrical niche. And thirdly, we must
  58. 2:19now create content based on what works,
  59. 2:22based on the data we have. And the
  60. 2:25third task, the third task was
  61. 2:26precisely content creation. We asked
  62. 2:29our marketer to create content based on
  63. 2:33his analysis and, accordingly, on the
  64. 2:37data we have. In the end, he came to us
  65. 2:40with these three posts, because he was
  66. 2:43asked to create three posts; They might
  67. 2:45ask you to create 10 posts that you can
  68. 2:48simply copy and paste and publish
  69. 2:50somewhere on Instagram, Telegram, maybe
  70. 2:53VK, or any other convenient platform
  71. 2:56you like, even Max. We did this for
  72. 2:59text content. The same can be done
  73. 3:01using this as a basis: making videos
  74. 3:04out of this—not videos, but carousels
  75. 3:07, generating images. We can also create
  76. 3:11a script for short videos from this. So
  77. 3:14, you'll have a constant stream of
  78. 3:16content generation. And here I'll show
  79. 3:19you how to do this using simple queries
  80. 3:21. All of this, in theory, should work
  81. 3:24within the system. And I'll show you
  82. 3:26how to do it within the system. For
  83. 3:27example, you could have your agent do a
  84. 3:30market analysis on Mondays, do an
  85. 3:32analysis of your competitors 'market on
  86. 3:35Tuesdays, or maybe even on Wednesdays,
  87. 3:38and then bring you content for the
  88. 3:40following week on Fridays. You set all
  89. 3:43of this up in a scheduled posting, and
  90. 3:46it's all created in the scheduled
  91. 3:48posting in just a few minutes. So, the
  92. 3:51main idea of this video is to show you
  93. 3:53an AI marketer, to build an AI marketer
  94. 3:56together with you, who will collect
  95. 3:58data, analyze this data, and create
  96. 4:00content for you based on this data.
  97. 4:02Let's go. Okay. How do we build such a
  98. 4:03marketer? As always, we have two
  99. 4:05options. The first option is that you
  100. 4:08build this marketer together with me
  101. 4:10now, meaning I'll show you the tools
  102. 4:12we'll be using. The second option is
  103. 4:15that you download these seven files in
  104. 4:17the first pinned message—how many?
  105. 4:20That's seven files, which you download
  106. 4:23directly from this Telegram channel.
  107. 4:25And you either manually send these
  108. 4:28files to your agent working on CLDcode,
  109. 4:32MS, Codex, Antigravity, or, if you use
  110. 4:35an app like Codex, an app like clkD,
  111. 4:38well, the official app from Claude,
  112. 4:41then you simply send this entire folder
  113. 4:44with all this documentation here in the
  114. 4:47chat and ask him to build a similar
  115. 4:50system, a similar structure. It works
  116. 4:53in both places. It doesn't matter where
  117. 4:56you use your agents. Even if you use
  118. 4:58them through the terminal, you simply
  119. 5:00upload all these materials to the
  120. 5:02appropriate documentation or folder. If
  121. 5:05it's complicated, or if you don't
  122. 5:07understand how to do it or implement it
  123. 5:10somewhere, you immediately ask the
  124. 5:13agent what instructions you need to
  125. 5:15upload. It will tell you, you upload
  126. 5:17them, and that's it. After that, it
  127. 5:19implements everything the same way I do
  128. 5:21. That is, there are several files that
  129. 5:25you send it, and it uses these files,
  130. 5:28these instructions, they are verified,
  131. 5:31meaning they work in my agents, and we
  132. 5:34installed them on Codex, ID Code, and
  133. 5:36Antigravity. And they understand what
  134. 5:39these files are, how to use them, and
  135. 5:42how to implement them for everything to
  136. 5:45work. So, you're downloading from the
  137. 5:47first pinned message now. I'm now
  138. 5:48showing you how this structure works.
  139. 5:51Here's a step-by-step breakdown in this
  140. 5:53video. Okay, for our marketer to work
  141. 5:55properly, they need to perform three
  142. 5:57steps. The first step is data
  143. 5:58collection. How will we collect the
  144. 6:00data? Of course, we can train or
  145. 6:03instruct our agents to simply collect
  146. 6:06data, meaning they go online and search
  147. 6:09for specific data. I have a similar
  148. 6:12structure in place, but to be honest,
  149. 6:14it's not entirely ideal. It's not
  150. 6:17parsing, because most content creators,
  151. 6:20or most services, or most databases,
  152. 6:24especially Avito, Viberis, large
  153. 6:27resources where it's desirable to pull
  154. 6:31in specific data, impose restrictions.
  155. 6:35The restriction is that you have to
  156. 6:38register with Avito, or even Viber. And
  157. 6:42they impose endless restrictions
  158. 6:44everywhere, which most agents can't get
  159. 6:47around. If you need to find some
  160. 6:49superficial information, like news or
  161. 6:52news digests, then your marketers, oh,
  162. 6:55your agents, will handle it without a
  163. 6:58problem. But if you need to do any
  164. 7:01serious work, serious social media
  165. 7:03parsing, parsing, parsing, analyzing
  166. 7:06platforms like Avito, or more in-depth
  167. 7:09parsing, uh, based on some data, then
  168. 7:12it's best to do all of this through the
  169. 7:15Firecoll service. I have a separate
  170. 7:17video about how to connect Firecoll to
  171. 7:20your agent. I won't go into that now.
  172. 7:23The advantage of Firecoll is that it
  173. 7:27gives you about a thousand or fifteen
  174. 7:30hundred free tokens daily per month,
  175. 7:34which you can use to train your Agent
  176. 7:38to parse any platform. The advantage,
  177. 7:41perhaps, of Firecoll is that it's one
  178. 7:43of the services that bypasses all these
  179. 7:45platform restrictions. So, we'll ask it
  180. 7:49now. I already have Firecoll connected
  181. 7:51to my agent. This was done using one of
  182. 7:54these instructions that are written
  183. 7:57here. These are the parsing
  184. 7:59instructions, the market parsing
  185. 8:01instructions, the full instructions. I
  186. 8:03sent these instructions to my agent,
  187. 8:05and he installed everything himself, he
  188. 8:08connected himself, he set up the
  189. 8:11connection himself, yes, the free plan,
  190. 8:13about 1,000 requests a month, and he
  191. 8:16hasn't had any problems now. Let's, you
  192. 8:19know what we'll do? Let's do some
  193. 8:21parsing. That is, we'll ask him to
  194. 8:23parse an ad on Avito. And I most likely
  195. 8:27won't be able to access Avito myself
  196. 8:28because I have a VPN enabled right now
  197. 8:30and it will block me. That is, avita.ru
  198. 8:33will block me, but we'll ask...Yes, you
  199. 8:37have a VPN enabled. So, you see, Avito
  200. 8:40is imposing some restrictions. We'll
  201. 8:43ask him to parse. Oh, I don't know,
  202. 8:45let's ask some electricians in St.
  203. 8:47Petersburg, so he'll find an ad for
  204. 8:49those same electricians. Why do we need
  205. 8:51to parse electricians? We can choose
  206. 8:53any niche. And my main goal is to show
  207. 8:57you that for a tool, for a marketer who
  208. 9:00will be doing work every week, every
  209. 9:03day, or at our request, it can handle
  210. 9:06any parsing without a problem,
  211. 9:08including Avito parsing. You can give
  212. 9:11it complex tasks, or you can simply ask
  213. 9:13it to parse the market, because it will
  214. 9:16already have the information, skill,
  215. 9:18and experience on how to parse the
  216. 9:20market. That is, it has complete
  217. 9:22instructions on how it should parse the
  218. 9:25market for your specific situation. You
  219. 9:28can, of course, modify this prompt; you
  220. 9:31can ask it to modify it specifically
  221. 9:33for you, for your niche. I'll use its
  222. 9:35classic prompt. I log in to my agent,
  223. 9:38who already has access to Firecoll. The
  224. 9:41only thing I've changed here is the
  225. 9:42folder where it should save the report.
  226. 9:44So, let's send this request, and in the
  227. 9:46meantime, let's familiarize ourselves
  228. 9:48with the prompt I sent it. Conduct
  229. 9:50market research in the electrician
  230. 9:52services niche in St. Petersburg. Use
  231. 9:54Fcrawl to analyze websites, catalogs,
  232. 9:58and ads. Generate a detailed Markda
  233. 10:00report and save it to the Dump folder
  234. 10:02in Obsidian. Report structure: Key
  235. 10:04segments include minor repairs,
  236. 10:05electrical installation, turnkey
  237. 10:07electrical services, electrical panels,
  238. 10:09smart home, and commercial properties.
  239. 10:12Targets for each segment include
  240. 10:13minimum price and average price. This
  241. 10:15is, roughly speaking, a general report,
  242. 10:17a general analysis, not necessarily for
  243. 10:20Avito. So, if we asked it to do the
  244. 10:23same here, but for Avito, and after
  245. 10:26this analysis, we'd ask it to do this,
  246. 10:30we could also look at specific
  247. 10:32platforms to see what it can do. Right
  248. 10:36now, we're starting by asking it to
  249. 10:38create a full report, a global report,
  250. 10:40roughly speaking, covering the entire
  251. 10:43internet, but with regional
  252. 10:44implications. That is, electrician
  253. 10:46services in St. Petersburg. Now, in 5-7
  254. 10:49minutes, I'm pausing this video, in 5-7
  255. 10:53minutes he'll do this work, come back
  256. 10:56to us with a report, and we'll review
  257. 10:59it in detail. The agent hasn't
  258. 11:01responded to me yet; he's probably
  259. 11:02getting ready to write me something on
  260. 11:04Telegram, but I saw this file in the "
  261. 11:06dump" folder. Look, here it is. Dump,
  262. 11:08Electric, SPB Market Search. Let's take
  263. 11:12a look. So, you see, he immediately
  264. 11:14linked this note to the other notes. He
  265. 11:17wrote that he does this using FCW
  266. 11:20plusdel Avitonaliz. Oh, so he took the
  267. 11:23data from the Avito analysis from the
  268. 11:26previous note, that is, when I was
  269. 11:28analyzing Avito, uh, on June 4th, it
  270. 11:31turns out, on the same topic, he pulled
  271. 11:33these notes from there too. I honestly
  272. 11:36don't remember that I analyzed Avito
  273. 11:38anymore. So, he's saying there are
  274. 11:404,000 ads in St. Petersburg on Avito
  275. 11:42alone. And the structure of the players
  276. 11:45, that is, the multiservice Landremont,
  277. 11:48that is, he talks about who is in this
  278. 11:51niche, aggregators and an army of
  279. 11:53private sellers on Avito. There. That
  280. 11:56is, he has completely written it down,
  281. 11:58that is, he most likely also parsed
  282. 11:59Avito. He should indicate a list of
  283. 12:01sources for the main segments at the
  284. 12:03bottom, but if he hasn't, you can ask
  285. 12:06him to do so. So, look at what he did.
  286. 12:08Here's a report. The main segments:
  287. 12:10minor repairs, electrical installation,
  288. 12:12turnkey electrical work, control panels
  289. 12:14, competition, where is the lowest,
  290. 12:16average, and highest and, accordingly,
  291. 12:18the average bill. And who is working
  292. 12:20and, accordingly, where they look for
  293. 12:23clients for companies with permits.
  294. 12:25Minor repairs, the minimum price is
  295. 12:27200-350 rubles per operation. Socket
  296. 12:30and switch in a finished place.
  297. 12:32Callouts are free for most. Rough
  298. 12:34electrical installation. And, turnkey
  299. 12:37electrical work, apartment. Control
  300. 12:39panels, the minimum price for a simple
  301. 12:41control panel. Here is 1,300. Smart
  302. 12:43home, minimum price per element—that
  303. 12:46is, I've already analyzed the Proiru
  304. 12:48specialists and scenarios above. It's a
  305. 12:50good thing, by the way, that we didn't
  306. 12:52ask him to analyze Avito exclusively.
  307. 12:54He immediately took a large sample,
  308. 12:56including private contractors from
  309. 12:58Proiru, and I don't know what kind of
  310. 13:01keyword this is. Oh, what kind of
  311. 13:02website is this? Ah, some kind of
  312. 13:04specialized, maximum-price website,
  313. 13:06accordingly, here. He took the entire
  314. 13:09internet in St. Petersburg, and
  315. 13:12completely, ah, ads on sites like
  316. 13:15Proiru, Avito, and probably some other
  317. 13:18sites. Analysis of competitors'
  318. 13:20packaging. I found Lenremon
  319. 13:22Multiservice—by the way, it's also a
  320. 13:25service that allows you to find private
  321. 13:28specialists. The only thing is, I think
  322. 13:31they operate centrally. A specialized
  323. 13:33company, two electricians, that is,
  324. 13:35proi.ru, Yandex services, aggregators,
  325. 13:37a website, perfect packaging, high
  326. 13:39trust in the platform, price,
  327. 13:41accordingly, so much for the services
  328. 13:43participants. So, well, it's a pretty
  329. 13:45good report, yes. And everything we
  330. 13:48asked him to do, by the way, he wrote
  331. 13:52to me on Telegram. Ah, with 1988 lines.
  332. 13:57All the sections according to your
  333. 13:58structure, respectively, market
  334. 13:59such-and-such. So here he is, ah, a
  335. 14:02brief summary so we don't have to dive
  336. 14:06too deep, so we can see what he
  337. 14:09actually did. Top 10 ideas for future
  338. 14:11content based on the identified pain
  339. 14:13points. So, he immediately pulled out
  340. 14:15the pain points and wrote down ideas
  341. 14:18for our future content. If we're an
  342. 14:20electrician, what do we need, and how
  343. 14:23do we attract an audience, respectively
  344. 14:25. Four FireCroll queries, and
  345. 14:28competitor Avito scripts parsed from
  346. 14:32existing data. Prices are current as of
  347. 14:35July 2026. But listen, yes, our
  348. 14:37e-marketer's eyes, they didn't let him
  349. 14:39down, so he did an excellent job with
  350. 14:40the parsing. So now we have data on
  351. 14:43competitors. We can either enter this
  352. 14:46data manually or in the handouts found
  353. 14:50in the first pinned message. Here's a
  354. 14:53guide on how to train an agent to turn
  355. 14:56this into a rhythm. A good agent is one
  356. 14:59you assign a task to, and they complete
  357. 15:01it. An excellent agent is one you
  358. 15:03assign a task to once, and after that,
  359. 15:05they complete it every so often. Here's
  360. 15:08a guide on how you can turn a good
  361. 15:11agent into an excellent agent, so they
  362. 15:14can send you these kinds of reports
  363. 15:16every Monday, or reports on the areas
  364. 15:19you need, or use Chrome—maybe once a
  365. 15:22week, once a month, or something else,
  366. 15:25or at some other time. In other words,
  367. 15:28they can collect data for you without
  368. 15:30your request. They can simply ask you
  369. 15:32for it, or they can come to you with
  370. 15:35ready-made data and say, "Hey, I've
  371. 15:37done the work. Go ahead and check it
  372. 15:39and see what we can do with this data."
  373. 15:42And then we have to work with this data
  374. 15:44. It would be foolish to start turning
  375. 15:46this data into content now. We must
  376. 15:48first look at what type of content is
  377. 15:50popular, that is, what's popular in the
  378. 15:52market, what's already out there. That
  379. 15:54is, we shouldn't just continue to pull
  380. 15:57data, but rather look at what's
  381. 16:00happening in our market specifically.
  382. 16:03Basically, we need to help our agent
  383. 16:08with oversight, so they understand, not
  384. 16:14what data, but how to create content
  385. 16:19that goes viral. What do we do for this
  386. 16:22? The easiest way is to turn on TikTok,
  387. 16:26YouTube, Shorts, whatever—that is,
  388. 16:29any social media. And we look at what's
  389. 16:32already out there. What hooks are
  390. 16:35popular videos using, what locations
  391. 16:38are popular videos using, what type of
  392. 16:41content—reels, long videos—what the
  393. 16:46author says, why these videos are
  394. 16:48getting views. So, we need to analyze
  395. 16:51what's already out there. So, the
  396. 16:54second step is analysis. It's not an
  397. 16:56analysis of the data we've collected;
  398. 16:59he's already done the analysis—
  399. 17:02analysis to convey this data in a
  400. 17:05format that the audience, roughly
  401. 17:07speaking, will accept. So, now we need
  402. 17:11to find some viral videos on our topic.
  403. 17:14In this case, we've chosen electricians
  404. 17:16. So, we need to find viral videos on
  405. 17:18the topic of electricians to help the
  406. 17:21agent create content that's already
  407. 17:23resonated. See? I'll open my TikTok now
  408. 17:27, type in some keyword like "
  409. 17:29electricians," and I'll find a few
  410. 17:32videos and show you how I'll send these
  411. 17:35videos to my agents for them to analyze
  412. 17:38. Although, I think we could probably
  413. 17:40find something on YouTube, too. Let's
  414. 17:42go to YouTube and type in the word "
  415. 17:44electrician." Yes, here we're no longer
  416. 17:48limited to just electrician. So,
  417. 17:51electrician, or electrician. Here we're
  418. 17:54no longer limited to location. Here
  419. 17:57we're looking for everything that
  420. 17:59catches our eye and will adapt it to
  421. 18:01St. Petersburg. Let's see. 2.5 million
  422. 18:04views. Well, this is just about
  423. 18:06assembling a junction box. Okay, an
  424. 18:08electrician should know that. Uh, so we
  425. 18:11can look at, yeah, that format. This
  426. 18:13guy, by the way, I've met him a few
  427. 18:16times on YouTube, he comments on
  428. 18:18certain videos, but that's probably not
  429. 18:20of interest to us yet. Let's do this.
  430. 18:23It's him again. So, you see, this guy
  431. 18:25has flooded the entire internet. We
  432. 18:27need some kind of content, where...Well
  433. 18:31, we can take this guy. Let's take a
  434. 18:35look. Yeah, I'll take this video. It
  435. 18:37looks like someone is explaining
  436. 18:39something and I'm sending it to my
  437. 18:41agent. I can send it to this agent, but
  438. 18:44in my marketing department, for example
  439. 18:47, it's distributed among several agents
  440. 18:50. I have Antigravity. It's better at
  441. 18:53extracting the main meanings from
  442. 18:55videos. I'm sending him a link to this
  443. 18:58video and, accordingly, I'll wait for
  444. 19:01him to turn on, activate, and give me
  445. 19:03an in-depth analysis of this video. Yes
  446. 19:05, I'm still looking for a different
  447. 19:08type of content. That is, we can
  448. 19:10probably give you long-form content,
  449. 19:13because there are a lot of people on
  450. 19:16YouTube who make great content. For
  451. 19:19example, five years ago, ah, all the
  452. 19:21stages of electrical installation in an
  453. 19:23apartment, well, this is a very long
  454. 19:25video at 2.12 hours, I think. Of course
  455. 19:27, I should have prepared in advance.
  456. 19:29Well, let's also watch a twenty-minute
  457. 19:31video, we'll use that. Uh, how is an
  458. 19:35electrician's apartment set up? Well, I
  459. 19:37don't think this is an electrician.
  460. 19:38Okay, the agent has already sent me a
  461. 19:40transcript of one of the videos. Okay,
  462. 19:43let's move on. Electrical installation
  463. 19:46in an apartment can only be done after
  464. 19:47alignment. He also immediately added
  465. 19:50this video to my YouTube. Okay. And
  466. 19:53let's, well, I think two videos are
  467. 19:55quite enough. We go to the folder in
  468. 19:57Obsidian where all these materials are
  469. 20:00saved, that is, all these eyes. Here,
  470. 20:02ah, electrical, installation,
  471. 20:06everything's okay. The essence, the
  472. 20:09context, and, accordingly, the words "
  473. 20:10video transcription" are here. All
  474. 20:12these notes are saved exactly as I need
  475. 20:14them. Since electrical is not my niche.
  476. 20:17Let's try it. So, let's return to our
  477. 20:19agent. So, what do I do? Let's send the
  478. 20:22agent a task right away. So, what's
  479. 20:24going on here? I'm not asking it to
  480. 20:26analyze the video now, because
  481. 20:27antigravity is responsible for analysis
  482. 20:29, and it's precisely configured for
  483. 20:31that. I'm asking our marketer, based on
  484. 20:33the video analysis done by "Antigravity
  485. 20:36"(meaning he went out, pulled out all
  486. 20:38the transcripts, all the data, and
  487. 20:40created a brief structure), to analyze
  488. 20:43these videos as the chief marketer and,
  489. 20:45accordingly, extract the entire
  490. 20:47structure, all the ideas, all the hooks
  491. 20:50, everything that works, because these
  492. 20:52videos have garnered a large number of
  493. 20:55views. I sent him two videos. Two
  494. 20:57videos isn't much. After he does all
  495. 21:00this, I'll ask him to, with the data we
  496. 21:03received and the analysis that works,
  497. 21:06that is, we'll, roughly speaking, begin
  498. 21:10to combine all this for our future
  499. 21:13content. Let's wait until he does this
  500. 21:15analysis of the two videos. Of course,
  501. 21:17it's better not to rely on two videos
  502. 21:19alone. It's better to take 10, 15, 20,
  503. 21:2130 videos, for example, and I regularly
  504. 21:24upload all my content for the last two
  505. 21:26years to my marketers who create
  506. 21:28content for my Telegram channel, along
  507. 21:31with all my text content for the last
  508. 21:33year and a half to two years, so they
  509. 21:36understand how I write. For my agent,
  510. 21:39who creates the scripts for my videos,
  511. 21:42I upload all my statistics and data
  512. 21:45from YouTube, plus I transcribe them
  513. 21:48for the last two years so they can
  514. 21:50write the structure of my future videos
  515. 21:54for me. I can just show you; if you're
  516. 21:56interested, message me. And we can
  517. 21:58implement something similar there for
  518. 22:00anyone. So, I'll just show you the
  519. 22:03structure of my videos while he does
  520. 22:05the analysis. Here are the ideas, oh,
  521. 22:08ube videos from my channel. And here,
  522. 22:10accordingly, are some fresh ideas, no,
  523. 22:13development, here's the channel
  524. 22:15development strategy. And, accordingly,
  525. 22:18here's this thing where the analysis of
  526. 22:21all my videos didn't give you. And when
  527. 22:24it's not brainstorming ideas for my new
  528. 22:27videos, it focuses on these metrics. It
  529. 22:30calculates how many views each video
  530. 22:33gets, how long each video is, which
  531. 22:36networks, and why they perform better.
  532. 22:40And based on subsequent videos—it's
  533. 22:42constantly learning, and based on
  534. 22:44subsequent videos, it gives me better,
  535. 22:46better, best, roughly speaking, content
  536. 22:48. I had to switch agents because this
  537. 22:50one ran out of content. I'd asked them
  538. 22:52too many questions before filming the
  539. 22:55video. So, all my agents are
  540. 22:57interchangeable, which is convenient.
  541. 22:59So, there's an agent based on CLC code,
  542. 23:01an agent based on antigravity code, and
  543. 23:03the advantage of these agents is that
  544. 23:05they share a single database. So, they
  545. 23:08all have the same data. So, they can
  546. 23:11access obsidians without any problems.
  547. 23:15And I'm a VZA, meaning I replace it. So
  548. 23:17if something stops somewhere, my work
  549. 23:20will never stop because I have other
  550. 23:22agents that can pick it up. And all
  551. 23:25these agents have their own systems.
  552. 23:28For example, they have some things of
  553. 23:30their own, of course, but most of their
  554. 23:32solutions overlap. So, I have an agent
  555. 23:34architecture, and each one looks at the
  556. 23:37architecture of the other and picks up
  557. 23:39the skills it needs from there. It
  558. 23:42works really well. And thanks to this,
  559. 23:45I never have any downtime. So, if my
  560. 23:47agent stops, I can easily turn on
  561. 23:49another agent, and it continues the
  562. 23:51first one's work because the entire
  563. 23:53system is here. I have separate
  564. 23:55instructions for assembling such agents
  565. 23:58. And, accordingly, all the files for
  566. 24:00assembling and training these agents
  567. 24:02are in the first pinned message. I
  568. 24:04already showed you this. Good. So, this
  569. 24:06one picked up the work first and,
  570. 24:08accordingly, completed the task. So,
  571. 24:11look what it did. Ah, I analyzed both
  572. 24:13videos like a marketer. So, he analyzed
  573. 24:16both of these videos. Right now, for
  574. 24:18the sake of demonstration, I'm throwing
  575. 24:20everything into the "dump" folder
  576. 24:21because I'm not an electrician, I don't
  577. 24:23need this data. If I wanted to combine
  578. 24:25them and do something with them, I
  579. 24:27would, of course, create a separate
  580. 24:29folder. And I would save all these
  581. 24:30analyses there, as I showed you with
  582. 24:32the YouTube example, so that I could
  583. 24:34compile the entire structure in the
  584. 24:36future. So, good. So, he, ah, fully
  585. 24:39described why each video worked. He
  586. 24:43explained the idea of the video, he
  587. 24:45chose the hooks, he chose the structure
  588. 24:47, he described, ah, what causes,
  589. 24:49accordingly, the retention of this
  590. 24:51video, the strong wording that was used
  591. 24:53, the emotions that were used in this
  592. 24:55video. The same thing applies to the
  593. 24:57second video. So, he was able to
  594. 24:59analyze these videos without any
  595. 25:01problem. And now we have data that
  596. 25:03allows us to—we have the data that we
  597. 25:07collected. We have an analysis of what
  598. 25:10works. And now we can easily create
  599. 25:13content or do whatever we want with
  600. 25:15this data and this analysis. Since
  601. 25:17we're hiring a marketer, and this will
  602. 25:20be an e-marketer, the e-marketer must,
  603. 25:22accordingly, create content. Therefore,
  604. 25:25the third task is content creation. All
  605. 25:28these agents, they work precisely
  606. 25:30according to these files, where all the
  607. 25:32instructions are fully written out on
  608. 25:34how each of these agents should perform
  609. 25:36this or that task. Same thing. We don't
  610. 25:39have to manually send them links to the
  611. 25:41videos we like, although I still do
  612. 25:43that manually, because the easiest way
  613. 25:44for me to simply forward the videos
  614. 25:46that resonate with me to my antigravity
  615. 25:48agent is to simply forward them to my
  616. 25:49antigravity agent. Antigravity will
  617. 25:51then save them, and other agents will
  618. 25:53pick them up and be able to work with
  619. 25:55these materials. Because, I repeat,
  620. 25:58their entire database is in one place.
  621. 26:01Very convenient. I use the dump purely
  622. 26:04to demonstrate various topics. Good. We
  623. 26:07have the data, we have, uh, everything
  624. 26:09that resonates. Let's create some kind
  625. 26:11of content. I don't know, let him write
  626. 26:14me a post, or three versions of a post
  627. 26:17based on all this data and all this
  628. 26:20content analyzed. So, I'm sending him
  629. 26:23the following prompt. Let's do it this
  630. 26:25way. And I need to tell him to put all
  631. 26:27this in the dump folder, otherwise
  632. 26:29he'll throw it in with my Telegram
  633. 26:31posts, throw all this information, and
  634. 26:33then they'll get mixed up. I won't
  635. 26:35delete it, as always. So, the dump
  636. 26:37folder asked him to do this work,
  637. 26:39because, I repeat, he'll throw all this
  638. 26:41at me. So, use the material that's
  639. 26:43already in memory. Sources, the latest
  640. 26:46market research report. And the latest
  641. 26:48analysis of a video competitor. Choose
  642. 26:50the most promising topic. Your task is
  643. 26:52to prepare a post for Telegram for
  644. 26:54publication. You can ask him to do more
  645. 26:56than just a post for Telegram; you can
  646. 26:59ask him to do some kind of carousel for
  647. 27:01Instagram, that is, do anything. In
  648. 27:03this case, he asked for a post for
  649. 27:05Telegram. So, explain for each topic.
  650. 27:07So, it's completely described here what
  651. 27:09he needs to do. The same will be
  652. 27:11described in your instructions, which
  653. 27:14you'll download. And the same here in
  654. 27:17the prompts. He'll know exactly what he
  655. 27:19needs to do. He just won't need to
  656. 27:22write such a long prompt, such a big
  657. 27:24request. But since my topic isn't
  658. 27:27electrical, I'd rather let him do it
  659. 27:30correctly in the video. And when I
  660. 27:33assign tasks, I mean, for content
  661. 27:35creation, for example, I don't assign
  662. 27:37tasks at all for Telegram, because they
  663. 27:40create the content for me themselves.
  664. 27:43So, they write posts for me. Here's my
  665. 27:46Telegram in my style, here's Telegram,
  666. 27:48and here are ideas for text posts. Each
  667. 27:51of their agents, respectively, throws
  668. 27:54in ideas here, uh, every few days.
  669. 27:57Claude wrote the code based on, uh,
  670. 27:59what worked for me. So, you see, he's
  671. 28:02already analyzed my content, analyzed
  672. 28:05the spare videos that I have, and,
  673. 28:07accordingly, creates ready-made text
  674. 28:10content. Either for Instagram or
  675. 28:12Threads. They do the same for Instagram
  676. 28:15. I get ready-made ideas for carousels
  677. 28:18every day, and I tweak them. Maybe I
  678. 28:22choose different colors, maybe I choose
  679. 28:24different headlines, but they already
  680. 28:27know what I need. And, accordingly, my
  681. 28:29e-marketers create content for me.
  682. 28:32Every day they bring me a carousel so I
  683. 28:34can publish these carousels on social
  684. 28:37media. There are prompts here that you
  685. 28:39can copy, go somewhere, I don’t know,
  686. 28:41to the GPT chat. Let me show you. Here,
  687. 28:44I think, yesterday I manually sent them
  688. 28:46. Prompts to the GPT chat. And you can
  689. 28:48do the same thing using the a key. So.
  690. 28:52Ideas for something else. Here’s the
  691. 28:55carousel, see? Here’s yesterday’s
  692. 28:56carousel that I published. I uploaded
  693. 28:59the prompts here and, accordingly, got
  694. 29:02a ready-made carousel. The prompts were
  695. 29:05pre-made in my Obsidian. I just had to
  696. 29:09copy and paste them, or message the
  697. 29:11agent. I'd run out of budget on Kiap.
  698. 29:15It's the service we use to generate
  699. 29:17carousels. And I was too lazy to
  700. 29:20replenish it. I'd simply log into the
  701. 29:23GPT chat on my phone in my free time
  702. 29:26and send these requests, receive these
  703. 29:28images, save them, and publish them on
  704. 29:31Instagram. Okay, our agent has already
  705. 29:34completed the job. Let's see what they
  706. 29:39did there. So, we asked them to save
  707. 29:41all this information to the dump folder
  708. 29:44. Dump. Here's the post. Ah, Telegram
  709. 29:47posts for the audience. A socket from
  710. 29:50300 rubles turns into 15,000. How do
  711. 29:52you calculate an estimate? And how do
  712. 29:54you read an electrician's estimate? Why
  713. 29:56doesn't a normal electrician bother you
  714. 29:59with plastering? He's based on content.
  715. 30:02Aluminum fretwork panel, which is
  716. 30:04quietly waiting for a fire. Well,
  717. 30:05basically, not bad. And now the text
  718. 30:07posts themselves. Why this post in
  719. 30:09particular? So, he's selling us an idea
  720. 30:11. Post number two. And why this
  721. 30:13particular post? Post number three. And
  722. 30:15, accordingly, why this particular post
  723. 30:17? Everything we asked him to do was
  724. 30:20based on the materials, the sources we
  725. 30:24sent him. So, look, yes, our marketer
  726. 30:26found the data, analyzed the
  727. 30:28competitors, created the best content,
  728. 30:31and created the content. All of this
  729. 30:34setup, all of it, took us literally
  730. 30:36half an hour because we have ready-made
  731. 30:39materials, ready-made instructions,
  732. 30:41which you can download as the first
  733. 30:44pinned post. And then it's just routine
  734. 30:46. So, we tell him to do the same thing
  735. 30:49once a week. Once a week, he searches
  736. 30:51for data, and during the week, we send
  737. 30:54him some links for him to analyze and
  738. 30:56video content. And, ah, once a week, he
  739. 30:58creates not three, for example, but 10
  740. 31:01text posts for us, or seven text posts,
  741. 31:03and we simply copy, paste, copy, paste,
  742. 31:05and publish these text posts on our
  743. 31:07Instagram. And the marketer works. If
  744. 31:10we also connect some N8N system, we can
  745. 31:14automate content publishing and connect
  746. 31:18some chatpoints and attract clients
  747. 31:22through these automations. In other
  748. 31:25words, the AA marketer works without us
  749. 31:28. Our main task is simply setting them
  750. 31:31up initially. Guys, I hope this video
  751. 31:33was useful to you. You will definitely
  752. 31:35build yourself such a marketer for your
  753. 31:37business. I'm not saying you need to
  754. 31:40fire your existing marketers, but this
  755. 31:42one will work for you, it will work
  756. 31:44from standard subscriptions, from chp,
  757. 31:46antigravity, claude, the herm code
  758. 31:48works on all systems. The main thing is
  759. 31:51to just start building it, and then you
  760. 31:54can gradually delegate such tasks and
  761. 31:56increase the number of employees in
  762. 31:58your project, in your business. This
  763. 32:00was Grizuevich. Be sure to like and
  764. 32:02subscribe to the channel. Bye.

About this transcript

This page contains the full transcript of Как создать ИИ маркетолога: парсит, разбирает конкурентов и создаёт мощный контент | ИИ Сотрудник by Игорь Зуевич, generated from the public captions YouTube serves with the video. The transcript has 5,076 words across 764 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.