Transcriptly

How Gong Built a $7B+ Company in the Age of Vibe Coding | Amit Bendov

EO19:15Added Aug 31, 2026

Amit Bendov, co-founder and CEO of Gong, argues that code is only one piece of the software moat. The real advantage comes from customer context, proprietary...

Watch on YouTube →
Contributed by 刘嘉琪

Transcript

Transcript format

  1. Intro

  2. 00:00A lot of companies stall not because of
  3. 00:03external conditions. It's because maybe
  4. 00:05the market like isn't large enough. If
  5. 00:07your market is too small, very little
  6. 00:09else matters. There is like in math this
  7. 00:12thing of local optimum, right? You're
  8. 00:14climbing a hill and you reach the top,
  9. 00:16but you're not getting to the mountain.
  10. 00:18You can get to like 50 million, 100
  11. 00:20million, right? You're not going to
  12. 00:22reach like a billion. Which market do
  13. 00:24you play in? That's like more than
  14. 00:25anything [music] else. We've built the
  15. 00:27system over 10 years. There's a lot of
  16. 00:28complexity. You know, you can't vibe
  17. 00:30code it even like with a 100 engineers.
  18. 00:32Otherwise, why are we not seeing one
  19. 00:34right now? Why are we not seeing like a
  20. 00:36competitor? If you're in an early
  21. 00:37market, which we are right now in AI,
  22. 00:40there is so many possibilities. Why
  23. 00:42would you like try to like one up a
  24. 00:45competitor, right? There more. Just just
  25. 00:47go and keep looking. Why don't go where
  26. 00:49all the other kids are going?
  27. 00:50Competition should be more like like an
  28. 00:52Easter egg hunt than like a football
  29. 00:54match that you're you got to match like
  30. 00:56every play that the other opponent is
  31. 00:57doing. It's not a zero- sum game like in
  32. 00:59in the early stages. Hi, I'm Amit
  33. 01:02Bendov, co-founder and CEO of Gong. Gong
  34. 01:05is the leader in revenue AI which helped
  35. 01:08company maximize their revenue
  36. 01:09potential. We saw an opportunity in
  37. 01:12[music] automating revenue with AI. That
  38. 01:14was like in 2016 before AI was like uh
  39. 01:17all the rage and uh today we're one of
  40. 01:19the fastest growing company. Our growth
  41. 01:21rate is over 50%. Every quarter we're
  42. 01:23actually growing faster than the
  43. 01:25previous quarter for almost like 12
  44. 01:27consecutive quarters now. So pretty
  45. 01:29incredible times.

  46. Design Product Like a Music Producer

  47. 01:39I grew like in a very humble background.
  48. 01:42I mean great family, loving family, but
  49. 01:44most of my family like did not even like
  50. 01:46go to college or university. So from my
  51. 01:49family's perspective, if I was like
  52. 01:50seller at a retail store, it'll be just
  53. 01:52fine. They love me still the same
  54. 01:54because it's a job and that's what you
  55. 01:56do. So my original dream as a teenager,
  56. 01:59I wanted to be a guitar player at a rock
  57. 02:01band. I was pretty good to be honest,
  58. 02:04but not like not top 1%. And I was like
  59. 02:07in a small country, Israel, it's not you
  60. 02:09can't even if you're big, like it's not
  61. 02:11all that big. That's a miss. But I
  62. 02:13started my career actually selling
  63. 02:15professional audio equipment at a retail
  64. 02:17store like like a guitar center. I was
  65. 02:19hoping to uh get close to the uh the
  66. 02:22show. But I realized okay maybe like you
  67. 02:24know I don't want to play in weddings
  68. 02:26when I'm 40 [laughter] to pay the bills
  69. 02:29and I was really intrigued by [music]
  70. 02:31computer science. So I signed up. There
  71. 02:33was a fork in the road for me. My first

  72. Gong’s Origin: CRM Had Nothing in It

  73. 02:36job was an engineer. I wrote code. Uh
  74. 02:39many people don't know that. Then I I
  75. 02:40led an R&D team. Then I moved from Tel
  76. 02:43Aviv to the US to start uh my first
  77. 02:45company, Click Software. It's not part
  78. 02:47of Salesforce. And I became a CEO of a
  79. 02:50company called Cyence in Business
  80. 02:52Intelligence. We're growing pretty fast.
  81. 02:54And then all of a sudden, we had a
  82. 02:55quarter that wasn't great at all. I have
  83. 02:58no idea what was going on. We're
  84. 02:59analyzing a lot of data from the CRM
  85. 03:01that everybody thought this is the way
  86. 03:03to manage customer. It's a single source
  87. 03:05of truth. There was nothing there. it's
  88. 03:08actually relies on people telling it
  89. 03:09[music] what happened and that didn't
  90. 03:11make sense. So only 1% of the
  91. 03:14information ever [music] actually makes
  92. 03:15its way to the CRM. It's all in people's
  93. 03:18heads. Uh now what happens the companies
  94. 03:20only half the sales people are making
  95. 03:22their numbers. Nine out of 10
  96. 03:24opportunities are not one without the
  97. 03:27company ever knowing why. And I thought
  98. 03:30wow this just totally doesn't make
  99. 03:32sense. like uh there was the year that
  100. 03:34AI was already beating the world
  101. 03:36champion in Go. You remember like Alpha
  102. 03:39Go and Deep Mind [music] and the thought
  103. 03:41was can we create like a system that
  104. 03:43will relieve people from the needs to
  105. 03:46update CRM. That was the idea that when
  106. 03:49I left Cyence I thought hey maybe we
  107. 03:51could build a company around that. The
  108. 03:53idea was like an autonomous system uh
  109. 03:55that automatically captures context from
  110. 03:58conversations and uses AI to turn it
  111. 04:01both to insight and and action. Are you
  112. 04:03using uh like a meeting assistant today?

  113. How Gong Invented the Meeting Assistant

  114. 04:06We invented this. This was from my
  115. 04:08background in music industry. First, if
  116. 04:10we capture a meeting that we have
  117. 04:13usually would have like multiple people,
  118. 04:15but we didn't want people to listen to
  119. 04:18like 45 minute meetings, right? We want
  120. 04:21to know like here are like the
  121. 04:22interesting points out there. So things
  122. 04:25like markers,
  123. 04:26topics. Today it's pretty standard but
  124. 04:30that did not exist. So we have uh first
  125. 04:33drew and I plagorized a little bit from
  126. 04:36uh Soundcloud. That was the play buttons
  127. 04:39and uh where we put in and you have
  128. 04:41those like tracks. If you was like
  129. 04:42Garage Band or something, you'll see
  130. 04:44like here's a drum, guitar, bass, right?
  131. 04:47So if you look at gong, you'll see those
  132. 04:49tracks. all the concept like listen to
  133. 04:51talk ratio. This is the first
  134. 04:53application. It was like measuring how
  135. 04:54much you're listening versus talking.
  136. 04:56How it tracks competitors. We invented
  137. 04:59this thing that was uh pretty pretty
  138. 05:01innovative and we went to raise [music]
  139. 05:04money. I said, "Let's start a company."
  140. 05:06I was pretty sure it's going to be a
  141. 05:07walk in the park, but it [music] was a
  142. 05:09far cry from that. It was like very
  143. 05:11frustrating and almost nobody wanted to
  144. 05:14invest in us. They obviously took a
  145. 05:16meeting and like another meeting and
  146. 05:18what about this and what about that and
  147. 05:20people thought that sellers are going to
  148. 05:22hate Gong because they're going to see
  149. 05:24it as a big brother. Google and Amazon
  150. 05:26are going to own this space that the
  151. 05:29technology is not there. When we started
  152. 05:31like AI was nowhere near uh where it is
  153. 05:34right now. So it was it was a hard job.
  154. 05:36So, we struggled to raise money, but
  155. 05:38finally we found like a couple investors
  156. 05:40that gave us the money and that allowed
  157. 05:42us to uh start the company and it was
  158. 05:44pretty smooth from that point on.

  159. EO Partner Highlight: Meet Attio

  160. 05:50What if your CRM kept up with the
  161. 05:52conversations that matter for you? Meet
  162. 05:55Atoio, the AI CRM for modern businesses.
  163. 05:58[music] Connect your email, calls, and
  164. 06:00product data, and AIO instantly builds
  165. 06:02your CRM with complete context. [music]
  166. 06:05Then ask anything, spot a deal at risk,
  167. 06:08draft outreach, uncover new prospects,
  168. 06:11spend less time in your CRM and more
  169. 06:14time building relationships. Whether
  170. 06:16you're running productled growth or
  171. 06:17enterprise sales, AIO adapts [music] to
  172. 06:20how your team sells and grows. And if
  173. 06:22you're building with AI agents, plug any
  174. 06:24MCP client straight [music] into your
  175. 06:26CRM, so your agents always have the
  176. 06:28context they need to act. 9,000
  177. 06:30companies build with AIO, including
  178. 06:32Granola, Modal, and Whisper [music]
  179. 06:34Flow. Try Atoio free at aio.com/eo.

  180. How to Build a Cult and Pull Ahead

  181. 06:47At Gong, we have our number one
  182. 06:49operating principles like raving fans.
  183. 06:50We're not happy with just happy

  184. Make Raving Fans, Not Just Happy Customers

  185. 06:52customers. Happy is like, yeah, I could
  186. 06:54be happy with a with a chair, right?
  187. 06:56That's not We want something people talk
  188. 06:57about that is a level above, right? What
  189. 07:00if I sit on a chair like here in the
  190. 07:02kitchen and all of a sudden it start to
  191. 07:04give me like a back rub and it adjust
  192. 07:06the temperature. Oh, what just happened
  193. 07:08here? Right? I mean, that's you get
  194. 07:10people to talk about it and it surprises
  195. 07:12you for the better. It does something
  196. 07:14that you don't uh don't expect. You got
  197. 07:16to understand your mission is not
  198. 07:18feature or AI or all of that. That's
  199. 07:20like what people buy. That's what they
  200. 07:21ultimately want to accomplish. First we
  201. 07:24focus on value not just features. So we
  202. 07:26said what does it do? So we have like
  203. 07:28two things that the gong will try to do.
  204. 07:30First like 75% of a seller's time is
  205. 07:33non-productive. All right it's like
  206. 07:34updating CRM it's like doing forecast
  207. 07:37doing training doing like sifting
  208. 07:39through a list. So what can we remove
  209. 07:40out of that second [music]
  210. 07:42when they're with talking with customers
  211. 07:44how do we help them be better on their
  212. 07:46job? So these are the two northstar
  213. 07:50metric people move from like somewhat
  214. 07:52reluctant to use it like ah is it like
  215. 07:55invasive to go oh wow like you just did
  216. 07:58that right so you get that that wow all
  217. 08:01the time and the product surprises you
  218. 08:02and what it does and it needs to work
  219. 08:04well and it needs to be above what you
  220. 08:06expect. We do have like a rough patch in
  221. 08:092023 because that was the postcoid

  222. How Gong Pushed Through a Crisis

  223. 08:13growth right the hangover a lot of our
  224. 08:15customers have overhired people and got
  225. 08:18into uh financial stress so we saw like
  226. 08:22uh higher churn we also had competition
  227. 08:24that went in and and bought like a bunch
  228. 08:27of competing products and they're going
  229. 08:29to our customers and consolidating. They
  230. 08:32say oh we're going to do like all three
  231. 08:34or four products in one. the products
  232. 08:36were not very good but at that time
  233. 08:38customers were just looking at a
  234. 08:39spreadsheet and say okay like yeah 1 2 3
  235. 08:41and so our growth slowed down
  236. 08:44substantially 23 I wouldn't say
  237. 08:46existential but it was definitely like
  238. 08:49uh not not fun days so we raised a lot
  239. 08:51of money but we barely touched it we
  240. 08:54just saved it for if and when we're
  241. 08:56going to need that so that gave us a lot
  242. 08:59of confidence and and the ability to
  243. 09:00execute and we knew that if we struggle
  244. 09:03right now all of our competition is
  245. 09:04struggling If you're in an early market,
  246. 09:07which we are right now in AI, there is
  247. 09:09so many possibilities. Why would you
  248. 09:12like try to like one up a [music]
  249. 09:14competitor, right? Just try go like in a
  250. 09:17different direction and there's so many
  251. 09:19option that could be like way better.
  252. 09:21I'm stealing from Jeff Bezos, right? Who

  253. Are You Truly Customer-Obsessed?

  254. 09:23said that like most people say they're
  255. 09:25obsessed with customer. They're not.
  256. 09:26They're really competitor driven. They
  257. 09:28look, oh, they did this, I'm going to do
  258. 09:29that, too, and I'm going to do better,
  259. 09:31right? That's the wrong thing. So, what
  260. 09:33do you do? We step on a gas pedal,
  261. 09:36right? So, we know that everybody's
  262. 09:38going to struggle. We have the cash and
  263. 09:39we're going to invest in engineering.
  264. 09:41I'd much rather invest in people and
  265. 09:43engineering than than in office space.
  266. 09:46Cuz like I remember like in 99 I was
  267. 09:50stuck with like an office that was like
  268. 09:51all like steel and plastic bags, all the
  269. 09:54area and chairs and like you know
  270. 09:55millions of dollars. It was like giving
  271. 09:57goosebumps even to watch it and say like
  272. 09:59why do we do that? I mean we we we spent
  273. 10:01money on real estate and not on people,
  274. 10:03right? Is there anything that I could
  275. 10:05have done right that would have like
  276. 10:06changed the outcome? So we only do like
  277. 10:09two or three year leases right I think
  278. 10:11for now for the first time we did like a
  279. 10:12four-year lease in an office but
  280. 10:14generally we've been in we work at
  281. 10:16weiworks like since day one and we still
  282. 10:18are offices like either they're too
  283. 10:21small you grow so fast that you need to
  284. 10:23move out or if you don't grow then
  285. 10:25you're stuck now with a lot of space. So
  286. 10:27we built like two more products. Now we
  287. 10:30added our fourth. Then we started like
  288. 10:32uh consulting and now they're they're
  289. 10:34behind.
  290. 10:36I'm an amateur cyclist. When you're
  291. 10:38climbing, right, you know, everybody's
  292. 10:40exhausted, then you attack, right? You
  293. 10:42just give it like all you and then just
  294. 10:44psychologically it breaks your if you're
  295. 10:47good, right? If you're able to
  296. 10:48accelerate that, everybody just gives up
  297. 10:50because they know that they're barely
  298. 10:52breathing and now you're charging ahead.
  299. 10:55And that's exactly what we did.
  300. 10:56Competition should be more like like an

  301. Competition Should Be an Easter Egg Hunt

  302. 10:58Easter egg hunt than like a football
  303. 11:01match that you're you got to match like
  304. 11:03every play that the other opponent is
  305. 11:04doing. It's not a zero sum game like in
  306. 11:06in the early stages. Just there they're
  307. 11:08more just just go and keep looking. Why
  308. 11:10don't go where all the other kids are
  309. 11:11going?

  310. Local Optimum: The Trap Founders Miss

  311. 11:16A lot of companies stall not because of
  312. 11:20external conditions. It's because maybe
  313. 11:22the market like isn't large enough,
  314. 11:24right? So there's like uh in math there
  315. 11:26is, you know, everybody suddenly knows
  316. 11:28there's this thing of local optimum,
  317. 11:30right? You're climbing a hill and you're
  318. 11:32you're climbing and first it looks like
  319. 11:34very very exciting and you reach the top
  320. 11:36but now you're you're not getting to the
  321. 11:38mountain. The most important thing is

  322. Why Your Market Matters Most

  323. 11:39the market like which market you're in
  324. 11:41like which market do you play in that's
  325. 11:43like more than anything else if you get
  326. 11:45it right then you could be okay in the
  327. 11:47others. There are two choices. Either
  328. 11:49you join an existing category. Let's say
  329. 11:51the advantage it's already established.
  330. 11:53People are buying and you just need to
  331. 11:55explain why you're better. There's
  332. 11:56already budget for this. Starting
  333. 11:58category is like uh is very compelling
  334. 12:00people. Oh, I'm going to start like a
  335. 12:01whole new category of of music or
  336. 12:04whatever, right? It's also hard because
  337. 12:06people don't have budget. They don't
  338. 12:07know how to think about it. So, there's
  339. 12:09a lot of education. And if you want to
  340. 12:12be precise, you can start a category. A
  341. 12:14category is by definition like a number
  342. 12:17of companies doing the same thing. If
  343. 12:19it's just you, you're not a category.
  344. 12:21You're a unique product. So you start by
  345. 12:23creating a unique product. If you're if
  346. 12:24you're not successful, then nothing
  347. 12:27happens. If you're successful, it's
  348. 12:28guaranteed that other will try to like
  349. 12:31either copy or do something better than
  350. 12:33you or like something different. That's
  351. 12:35when you start to have a category. It
  352. 12:37does take time. Today, Gong is
  353. 12:39considered or something like Gong
  354. 12:41revenue AI. Most companies say we need
  355. 12:45that but it wasn't like a must-have when
  356. 12:47we started like people had to find
  357. 12:50budget for it. It didn't exist right it
  358. 12:52wasn't like a CRM or a contact database.
  359. 12:54[music] So creating a category sounds
  360. 12:56very exciting and I'm I have no regrets
  361. 12:59but if you're an entrepreneur don't
  362. 13:01think that oh I need to start like a new
  363. 13:02category. So first understand if you're
  364. 13:04stuck understand if this is like
  365. 13:06intrinsic to your business or something
  366. 13:09that's happening right now. There are a
  367. 13:10lot of companies that are for example,
  368. 13:12oh we develop solutions for SDRs, right?
  369. 13:15Don't do that. If you live in the Bay
  370. 13:17Area, you think like everybody has SDR.
  371. 13:19Companies in Wisconsin don't have. So
  372. 13:20this is example of a small market. I
  373. 13:22mean you can get to like 50 million 100
  374. 13:24million, right? You're not going to
  375. 13:26reach like a billion. So understand if
  376. 13:28you're stuck uh because you've reached a
  377. 13:30local optimum of your market. Uh also if
  378. 13:32if it's a small market and it's very
  379. 13:34competitive, that that's that's the
  380. 13:36worst thing. It's like five hungry dogs
  381. 13:39like fighting for a very small piece of
  382. 13:41[music] meat, right? For us, we
  383. 13:42understood, listen, we have like a
  384. 13:44trillion dollar opportunity. Like, if we
  385. 13:46don't succeed, like it's not the market,
  386. 13:48it's us. The noise that's happening

  387. Advice for AI Software Builders

  388. 13:50right now with AI, I know it's very
  389. 13:52confusing. Just just make a bet and just
  390. 13:55keep going. Ignore the noise. Right now,
  391. 13:57people are very emotional. It's almost
  392. 13:58like when, you know, people fall in love
  393. 14:00with someone that, you know, it's more
  394. 14:02like the hormone thinking versus like
  395. 14:04the brain. The dust will settle. At some
  396. 14:06point it'll be clearer. So if you have a
  397. 14:08direction, hopefully it's good. Just
  398. 14:10keep going, keep plowing, do your own
  399. 14:12thing. Look a little bit like to what's
  400. 14:14going on, but don't be enamored with
  401. 14:17like things around you. If I have to
  402. 14:19distill it to one, this is it. Like who
  403. 14:21do you cater to and what do you offer?

  404. Why the “SaaSpocalypse” Is Overblown

  405. 14:31Apocalypse. There's some truth in that
  406. 14:33but it's like a way like overreaction of
  407. 14:35the market right now. It's like uh uh
  408. 14:37investors are worried like if AI is has
  409. 14:40this like power of god anybody can do
  410. 14:43software. So what what is the mode? It's
  411. 14:45a natural question but it's way way
  412. 14:48exaggerated cuz most companies will
  413. 14:51probably like um open source their code
  414. 14:54and it will feel the difficult for
  415. 14:56someone to compete with them. The code
  416. 14:57is just a part. What about like running
  417. 14:59it? Like there's like a security
  418. 15:01vulnerability that you need to patch a
  419. 15:03package, a library, like what do you
  420. 15:04call it like 3:00 a.m. Just writing the
  421. 15:07code is one part of a software company,
  422. 15:10but actually like running it right and
  423. 15:13supporting production driving adoption
  424. 15:16is extremely hard. So even if like work
  425. 15:19there or Salesforce's like code was out
  426. 15:21there on the internet, it would still be
  427. 15:23extremely difficult for anybody to
  428. 15:26compete with them. CRM and HR are
  429. 15:28incredibly complex, right? People just
  430. 15:30underestimate it. They think, you know,
  431. 15:32you can't vibe code that like uh even
  432. 15:34like with 100 engineers or otherwise,
  433. 15:37why are we not seeing one right now? Why
  434. 15:39are we not seeing like a competitor? We
  435. 15:40have to understand that sales is
  436. 15:42asymmetrical engagement. So buyers don't
  437. 15:46necessarily tell you everything. They
  438. 15:48might even lie to you. I don't have a
  439. 15:49budget, right? Or I am the decision
  440. 15:52maker even though you're not. We're
  441. 15:53looking at a few of their competitors.
  442. 15:55they might not tell you what they are
  443. 15:56and they might not be looking at
  444. 15:57anything. So this is like already like
  445. 15:59AI needs to have like excellent context
  446. 16:02or data right when you negotiate like
  447. 16:05you want to lower the price you want to
  448. 16:07increase the price. So that's like uh
  449. 16:09one thing. Second at least for us like
  450. 16:12we help mostly with B2B sales which is
  451. 16:15complex. It's like multiple people
  452. 16:17selling to multiple people.
  453. 16:18Understanding the power dynamics, the
  454. 16:20political power structure, all of that
  455. 16:23something that AI will struggle.
  456. 16:24Ambiguity is not where it's good.
  457. 16:26Reading the room and judgment,
  458. 16:28creativity, right? AI is very
  459. 16:30straightforward. It's trained on u being
  460. 16:33plausible. Sometimes the creative moves
  461. 16:36aren't the ones that are plausible. It's
  462. 16:38just doing something like that seems
  463. 16:39unreasonable, but still like a winning
  464. 16:42place.
  465. 16:44when we're alone, like nobody cares.
  466. 16:46Once you're you're head there's a
  467. 16:48crosshair on your back, so everybody
  468. 16:49wants to kill you, right? And that's
  469. 16:51like we take it as a compliment. We've
  470. 16:53built the system over 10 years and and
  471. 16:55again, it's there there's a lot of
  472. 16:56complexity. Our vision was we thought of

  473. Gong’s Vision: A Self-Driving Revenue System

  474. 16:58it as a self-driving car, right? So when
  475. 17:00we started Gong, we wanted to build an
  476. 17:02autonomous system. Our inspiration was a
  477. 17:04was a Tesla self-driving car like five
  478. 17:06levels. Five is like full self-driving
  479. 17:08and one is getting like lanekeeping
  480. 17:10alerts. And he thought if you want to
  481. 17:12build a self-driving car and this is a
  482. 17:14self-managing customer management
  483. 17:16system, you need three things. First,
  484. 17:18you need the sensors to capture all the
  485. 17:20data. In a car, it'll be the radars, the
  486. 17:23lighter, cameras. So, this is our
  487. 17:25revenue graph. Second, you need to make
  488. 17:27sense of all of it, right? So, you see
  489. 17:29like a red light blinking. Is it a
  490. 17:31traffic light or is it a car or a
  491. 17:33motorcycle? Is it accelerating or
  492. 17:35decelerating? Right? So, this is like
  493. 17:37the AI level. And then there is the
  494. 17:38application layer. So we always thought
  495. 17:41about like the five levels and we said
  496. 17:44okay today we're like uh 2017 we're like
  497. 17:47a level one right and when can we get to
  498. 17:49level five so we built uh even though
  499. 17:52something like a lighter didn't exist or
  500. 17:54the AI wasn't strong enough to be full
  501. 17:57self driving so we built the framework
  502. 17:59so we can plug in like the technologies
  503. 18:01as they mature if you could vibe code it
  504. 18:03like overnight then uh then we have a
  505. 18:05problem but it's it's not it's not easy
  506. 18:07there's a lot of uh nuance in details
  507. 18:10that goes into especially deploying at a
  508. 18:12large scale. So I think it's like a
  509. 18:15underestimation of what it takes to
  510. 18:16really run a company right me not an
  511. 18:19investment advice I think it's a good
  512. 18:20good buying opportunity for the
  513. 18:22companies that are real [music] from the
  514. 18:23outside when you see a company growing
  515. 18:25right it looks like this but from the
  516. 18:27inside there are multiple revenue
  517. 18:29sources right and you need to layer them
  518. 18:31in a way that it looks smooth from the
  519. 18:33outside but each one is kind of like S
  520. 18:36shape right at some point it'll hit a
  521. 18:38glass ceiling you know you ask me about
  522. 18:40the people getting stuck because you're
  523. 18:42you're doing one thing and and that's
  524. 18:44it. Like it's not it's not growing as
  525. 18:46fast anymore. So, you always need to
  526. 18:47layer those things. You need to think
  527. 18:48about like two or three years in advance
  528. 18:50cuz uh when you realize that you need
  529. 18:52it, like it's already too late.