Transcriptly

How We 5X Our Revenue While Others Chase AI Updates | Lindy, Flo Crivello

EO12:29Added Aug 31, 2026

Flo Crivello built Lindy, the world's easiest no-code AI agent platform, by breaking every conventional startup rule - here's his unconventional playbook. Most AI founders chase the latest models and worry about competition, but Flo took the opposite approach. After his previous startup Team Flow fai

Watch on YouTube →
Contributed by 刘嘉琪

Transcript

Transcript format

  1. Intro

  2. 00:00If you're not embarrassed by your first
  3. 00:01version, you shipped to late in AI. I
  4. 00:04would even go as far as to say like if
  5. 00:05your first version works, you're not
  6. 00:07thinking big enough. You should build
  7. 00:10for the future generations of the model
  8. 00:12so much and the models are improving so
  9. 00:13rapidly that the current generation of
  10. 00:16models should barely support your
  11. 00:17vision. Just barely. If you spend any
  12. 00:19time on Twitter, you feel like there is
  13. 00:21so much competition in AI agents. You
  14. 00:23feel like there is a new a new company
  15. 00:24coming out every week. And yet when
  16. 00:26you're out in the market like us, you
  17. 00:27don't run into competitors that often.
  18. 00:30So I sometimes compare it to like
  19. 00:31there's like trillions of trillions of
  20. 00:33stars in the universe and yet the
  21. 00:35universe is 99.9% empty space. It's the
  22. 00:37same, right? There's a lot of us
  23. 00:39nominally, but somehow it feels empty.
  24. 00:40And I think because the market is just
  25. 00:42so enormously huge. These people have
  26. 00:44gone about their entire life hearing
  27. 00:47everyone tell them the reason why they
  28. 00:48couldn't do things. But actually, if you
  29. 00:50do it, you're going to be very, very
  30. 00:52successful. You're going to make a lot
  31. 00:53of money. you're going to build
  32. 00:54something very impactful with the twof
  33. 00:55world. It becomes addictive. It's like a
  34. 00:57ha like tell me you can't I can't do it.
  35. 00:58I'll show you I can.
  36. 01:04I'm Flo Killo. I'm the CEO of Lindy
  37. 01:07which is the world's easiest way to
  38. 01:09build AI agents with no code. Lindy is a
  39. 01:12setup of building blocks that you can
  40. 01:14put together to build your own AI agents
  41. 01:17and to get them to do anything. One big
  42. 01:18use case that we have is a customer
  43. 01:20support. Step one, customer support
  44. 01:22ticket arrives by email. Step two, you
  45. 01:24can have a condition building block that
  46. 01:26basically branches up the agent. So you
  47. 01:28can be like, hey, is this a customer
  48. 01:30asking a question or is this a customer
  49. 01:31asking for a refund? Step three, for
  50. 01:33example, if they're asking a question,
  51. 01:35you can use our knowledge base building
  52. 01:37block, your FAQ, your website, a PDF,
  53. 01:39your notion, whatever you want. Then you
  54. 01:41can ask maybe finally step four to
  55. 01:43answer the support ticket based on the
  56. 01:46knowledge that it extracted in the
  57. 01:47knowledge base. We have quintupled uh in
  58. 01:49revenue over the last 6 months and
  59. 01:51roughly doubled the size of the team
  60. 01:52since then.

  61. Focus on What Never Changes

  62. 01:58I think it's Jeff Bezos who said
  63. 02:00something like people always comment
  64. 02:02about how fast everything is always
  65. 02:03changing like he likes to focus on what
  66. 02:05doesn't change right in a world thing is
  67. 02:06always changing it finds these pillars
  68. 02:08of stability to hang on to and I think
  69. 02:10what doesn't change is the user and a
  70. 02:12lot of preferences of the user will just
  71. 02:14immutable that really surprised us. We
  72. 02:17released the platform and we started to
  73. 02:19have quite a few users ask us for HIPPA
  74. 02:21compliance which is this regulatory
  75. 02:23standard in the US that allows you to
  76. 02:25work with sensitive health information.
  77. 02:27So people started to be like can you be
  78. 02:29HIPPA compliant? Can you be HIPPA
  79. 02:30compliant? I want to use legal for
  80. 02:31healthcare use cases. And we were like
  81. 02:33why like what do what do you want us to
  82. 02:34be HIPPA compliant for? And that's when
  83. 02:36we we understood the use case which is
  84. 02:38turns out it's huge and it's a big
  85. 02:39industry today. There's a lot of really
  86. 02:41big companies dedicated solely to this
  87. 02:42use case where doctors spend a lot of
  88. 02:45time taking notes during patient visits.
  89. 02:47They call it charting because there's
  90. 02:48like a very specific standard of notes
  91. 02:50that they call soap notes. Some of them
  92. 02:52started building Lindy's to help them
  93. 02:54build get their soap notes. So they use
  94. 02:56Lindy on a mobile phone and they
  95. 02:58recorded the conversation using the
  96. 02:59record feature that we have and then
  97. 03:01they would prompt Lindy to give them the
  98. 03:02format of notes. In the case of AI
  99. 03:04agents, the user wants something that's
  100. 03:05easy to use. We want something that's
  101. 03:07effective, something that works,
  102. 03:08something that's simple to use,
  103. 03:09something that's affordable, something
  104. 03:10that's fast. These things are always
  105. 03:12going to be the case. And the problem
  106. 03:13that the user is trying to solve is is
  107. 03:15fairly constant as well.

  108. Build AI That Barely Works

  109. 03:21LinkedIn's founder Hoffman, who said
  110. 03:23this famous thing, if you're not
  111. 03:24embarrassed by your first version, you
  112. 03:26shipped too late in AI. I would even go
  113. 03:28as far as to say like if your first
  114. 03:30version works, you're not thinking big
  115. 03:32enough. You should build for the future
  116. 03:34generations of the model. so much and
  117. 03:36the models are improving so rapidly that
  118. 03:38the current generation of models should
  119. 03:40barely support your vision just barely.
  120. 03:42That was very much our case at Lindy
  121. 03:44like for a long time the product simply
  122. 03:46did not work because the models were not
  123. 03:47good enough. The very first version of
  124. 03:50the product was just a prompt. So
  125. 03:53instead of building this AI agent using
  126. 03:55these building blocks that I mentioned,
  127. 03:56it was literally just writing text.
  128. 03:58Because it's just a prompt, you're just
  129. 03:59trusting this AI model to basically
  130. 04:02follow your instructions. It may not
  131. 04:03follow the instructions. You're you're
  132. 04:05like praying the LM gods. Even to this
  133. 04:07day, even to the frontier models, like
  134. 04:09fast forward two or three years, to this
  135. 04:11day, Claude Opus 4 or OpenAI, I use the
  136. 04:15very most frontier models right now,
  137. 04:17even they very often struggle. We are
  138. 04:19building for the next generation of
  139. 04:20models. The current product doesn't work
  140. 04:22and yet we keep building because we
  141. 04:23expect that the next generation of
  142. 04:25models will make it work. And the way I
  143. 04:27characterize it today is it's it's sort
  144. 04:28of AI in turn. It can perform simple
  145. 04:31enough tasks. Anything that you can
  146. 04:32describe in a step-by-step uh fashion is
  147. 04:35something that that Lindy can do. Over
  148. 04:36the long term, the vision is to build a
  149. 04:38full-blown AI employees, an AI agent
  150. 04:40that can do anything that the human
  151. 04:42employee can do on a computer. So,
  152. 04:44that's my piece of advice is like skate
  153. 04:46for the puck is going to be in in Steve
  154. 04:47Jobs world and skate for where the the
  155. 04:49next generation of models is going to be
  156. 04:51and build with that in mind.

  157. Don't Hesitate to Pivot Hard

  158. 04:57Don't build a mine until you've you've
  159. 04:59found gold. I really identified these
  160. 05:01two different distinct phases in a
  161. 05:03startup life cycle. One, you prospect.
  162. 05:05You look for the right mining site where
  163. 05:07you can find gold. And once you've
  164. 05:09struck gold, but only then do you build
  165. 05:11a mine. And building a mine is basically
  166. 05:13hiring a team and raising money and all
  167. 05:15of that stuff. It is really hard to move
  168. 05:17a mine. And so if if you build too large
  169. 05:19a mine and you haven't found gold, what
  170. 05:20you need to do is you need to destroy
  171. 05:21your mine. You need to fire your team.
  172. 05:25So I started at Uber as a software
  173. 05:27engineer. I took a six months break
  174. 05:29after Uber. So COVID breakout, everybody
  175. 05:31was stuck at home. So I was just on my
  176. 05:33computer at home and I I read a blog
  177. 05:35post by John Palmer called the spatial
  178. 05:37interfaces sort of extolling the virtues
  179. 05:39of of spatial design. And there's almost
  180. 05:41a call to action in that blog post
  181. 05:42that's like I wonder what a special zoom
  182. 05:44would look like. I was like oh it's a
  183. 05:45good question like I started like
  184. 05:46hacking it together and in the span of a
  185. 05:48couple of hours they just had like a
  186. 05:50prototype of like a special zoom. It's
  187. 05:51like you see your video in a bubble. You
  188. 05:53can move the bubble around on like a two
  189. 05:54dimensional canvas and you can only hear
  190. 05:56and be held by people around you. And
  191. 05:58again, this coincided perfectly with
  192. 05:59COVID. And so I started hosting virtual
  193. 06:01happy hours in there with my friends
  194. 06:02like, "Hey, join me in that space. I I
  195. 06:03built and I sort of got sucked in little
  196. 06:05by little. I just kept adding features
  197. 06:07and I realized it could be like a
  198. 06:08collaboration space. I added a
  199. 06:10whiteboard and so forth. And then I
  200. 06:11realized like oh my god, like
  201. 06:12everybody's working remotely right now.
  202. 06:14Like you know this could be a virtual
  203. 06:15office. Like there is a market here." I
  204. 06:17released it. It got traction. like bunch
  205. 06:18of users started sagging up, raised some
  206. 06:20money and then like the company was just
  207. 06:22after the races. When COVID subsided and
  208. 06:26when people returned to the office, Team
  209. 06:28Flow started not going so well and the
  210. 06:30growth stalled completely. What finally
  211. 06:33convinced me that there was simply no
  212. 06:34market anymore was like looking around
  213. 06:36and seeing all the competitors of which
  214. 06:38we had many like good teams like really
  215. 06:41good people and they had raised a lot of
  216. 06:43money and none of them were taking off.
  217. 06:44So that's when I just realized like
  218. 06:46there is no there is no market. That's
  219. 06:47what I did. It was extremely painful,
  220. 06:48but I fired twothirds of the team and it
  221. 06:50coincided with the release of the GPT3
  222. 06:53API. So, it was before the GPT. It was
  223. 06:56like mid 2022 and we built a meeting
  224. 06:59recorder in team flow and we were very
  225. 07:01interested in the GP3 API and so we were
  226. 07:04like what can we do with it? So, we we
  227. 07:06did what now a lot of applications do
  228. 07:07which is a meeting summarization. And
  229. 07:10then sales team came to me and they were
  230. 07:11like flow can this thing also update our
  231. 07:13sales force after sales calls and we
  232. 07:15were like sure I don't see why not. And
  233. 07:16so we basically kept building and kept
  234. 07:19fleshing out that like Salesforce
  235. 07:20integration and then HubSpot integration
  236. 07:22and and that's when it clicked for us
  237. 07:24like we realized like wait a minute we
  238. 07:26can get the LLM to not just generate
  239. 07:28text for us but generate API calls like
  240. 07:30generate actual actions for us and so
  241. 07:32that's when we were like wait a minute
  242. 07:34like this can be like an an entire AI
  243. 07:36employee basically we got sucked in to
  244. 07:38that and again it coincided with first
  245. 07:40team flow stalling out we got excited
  246. 07:43about this AI thing jp came about in the
  247. 07:45middle exploded and like oh my god like
  248. 07:47AI is real it's happening and so we just
  249. 07:49decided in 2023 to completely fool the
  250. 07:52team flow and just focus on the AI
  251. 07:54opportunity and and I I did that very
  252. 07:56deliberately and it's painful as a
  253. 07:57founder you need to deliberately
  254. 07:59mismanage your team you need to let them
  255. 08:01alone and they're going to hate you like
  256. 08:03where is our CEO what's he doing all day
  257. 08:05I'm looking for gold like the mind
  258. 08:06doesn't matter the mind is so very
  259. 08:08mismanaged right now and you guys are
  260. 08:10lost at sea and I decide not to care
  261. 08:12because I'm going to spend my time
  262. 08:13digging for gold um that's that's what I
  263. 08:16Don't don't don't build a mind before
  264. 08:17you you found the gold.

  265. Stop Avoiding Micromanagement

  266. 08:24My leadership philosophy is the job of a
  267. 08:26CEO is to win. And so that that's my
  268. 08:28guiding light. Epistemologically that is
  269. 08:30how I will define what is right or
  270. 08:32wrong. Perhaps I would compare it to
  271. 08:33gardening. You know it starts with like
  272. 08:35selecting the seeds. So I really do
  273. 08:36believe like half the job is just
  274. 08:38recruiting like getting the right people
  275. 08:39into the door. Like again I think it's
  276. 08:41Kra Boy who said the people you hire is
  277. 08:43the company you build. and I'm reminded
  278. 08:44every day of how true that is. It's half
  279. 08:46the job. So just getting the right
  280. 08:48seeds. If you don't have that, the plant
  281. 08:49won't be good. And then giving a
  282. 08:51direction to the team and to the seed
  283. 08:52like that's the sign. You grow in this
  284. 08:54direction, you know. So that's that's
  285. 08:56the second big part of the job is just
  286. 08:57setting the vision and guiding the plant
  287. 08:59as it grows, but you can't grow for the
  288. 09:02plant. Plant won't grow the way you want
  289. 09:04it to. So you've got to fire people.
  290. 09:05There's this amazing book by Bill Walsh
  291. 09:07called The School Takes of Itself,
  292. 09:08right? And so the philosophy of Bill
  293. 09:10Walsh is like success consists of just
  294. 09:12two things. one, identifying the right
  295. 09:14inputs that lead to success. And two,
  296. 09:16nailing each and every one of these
  297. 09:17inputs perfectly, right? If you do these
  298. 09:19two things, you will win. If you don't
  299. 09:20win, it means you either don't
  300. 09:21understand what it takes to win or
  301. 09:22you're not doing what it takes to win.
  302. 09:24That's just one of these two things. And
  303. 09:25so, it's actually funny. We on boarded a
  304. 09:27new employee this morning. He's the
  305. 09:29first one of his function. And I spent
  306. 09:30the day yesterday, Sunday, like
  307. 09:32literally seven plus hours just writing
  308. 09:34an 18 pages long document about this is
  309. 09:37precisely exactly we we call it a
  310. 09:39standard of excellence. Step by step by
  311. 09:40step. This is the job. Again, that's the
  312. 09:43plan. It's like if you if you see them
  313. 09:44doing something wrong, it's like no.
  314. 09:45It's like your guard is like, "No, this
  315. 09:46is not what you this is not what this is
  316. 09:48what you do." And you just keep putting
  317. 09:49these guardrails and you refer back to
  318. 09:50the dock that you wrote. Absolutely, you
  319. 09:52should micromanage. The reason why you
  320. 09:53are in a position of of of authority is
  321. 09:55because someone with capital most likely
  322. 09:58trusted you and your taste. So you you
  323. 10:00are entitled, you are in fact obligated
  324. 10:02to micromanage. That that's why you're
  325. 10:03in your seat. You exercise your taste.
  326. 10:04You exercise your judgment. You exercise
  327. 10:06your your knowledge about your business,
  328. 10:07your industry, your user, your product,
  329. 10:09and you guide the plant. That's my
  330. 10:10philosophy.

  331. Be a Definite Optimist

  332. 10:17I view building a company as a way to
  333. 10:20bring an ambitious bold vision to
  334. 10:23reality. I don't think the Lin startup
  335. 10:26model leads you there. I'm a huge fan of
  336. 10:29Peter Thiel's thinking on this. He calls
  337. 10:31it definite optimism. It's like a 2x2.
  338. 10:33You've got a definite and indefinite and
  339. 10:34optimism and pessimism. And so definite
  340. 10:36or indefinite is like do you know or do
  341. 10:38you not know? There's a lot of
  342. 10:39indefinite optimists in tech. I view
  343. 10:42crypto as the the eternal indefinite
  344. 10:44optimist industry. It's like we don't
  345. 10:46know what's going to happen. We don't
  346. 10:47know why things are going to go well,
  347. 10:48but wag me, we're all going to make it.
  348. 10:50You know, everything's going to go so
  349. 10:51well, but we we can't exactly tell you
  350. 10:53half. Instead, Peter Tiel pushes for
  351. 10:55this vision of a definite optimism.
  352. 10:57Like, you must have a crisp opinion, a
  353. 11:00well-defined opinion of what the world
  354. 11:02will look like in the future, what it
  355. 11:03should look like, what it must look
  356. 11:05like, and bring that vision to life.
  357. 11:07What percent of your time did you spend
  358. 11:08in 2022 speaking to an AI? It was 0%.
  359. 11:11Today in 2025, it's it's a lot more than
  360. 11:13that. And I think the trend is going to
  361. 11:14continue and continue and continue. AI
  362. 11:16agents are basically in the short term
  363. 11:18going to turn us all into managers.
  364. 11:20Instead of doing individual contributor
  365. 11:22work, you're going to manage a swoop of
  366. 11:24AI agents doing your work. I sometimes
  367. 11:27describe it as like imagine if any
  368. 11:30random 25year-old could have the same
  369. 11:32impact in the world as Google. And so I
  370. 11:35I think that AI agents are going to do
  371. 11:37to the world of business at large what
  372. 11:39the internet has done to media. It's
  373. 11:41going to be a very powerful leveling of
  374. 11:44the playing field. It's going to unlock
  375. 11:46opportunity to for everyone regardless
  376. 11:48of their network or their capital or
  377. 11:50their access or their geography. It's
  378. 11:53just really going to come down to their
  379. 11:55vision and the ideas they have and their
  380. 11:57ability to leverage AI agents.