Pod Insights Y Combinator Podcast How To Use AI In Your Startup About this episode AI continues to improve at an exponential pace. So what can you do as a founder to take advantage of it?
In this episode of Office Hours, YC Partners discuss what you should consider if you’re thinking about pivoting to or incorporating AI as part of your startup.
Transcript 0:00 Speaker 00 We're talking about all this AI stuff and it's all awesome, but the fundamental things that the founders still need to do to like enable the technology to produce value for customers are the same. And if you're not doing that, just switching your idea over to something that makes calls to OpenAI is not going to change your fate as a startup.
0:17 Speaker 01 If you are building a cloud company, obviously you're going to use the cloud. Same thing with AI. It doesn't make sense for anyone creating a new company today to not leverage AI within their company.
0:28 Speaker 02 Go to your friend who works in this and go and sit next to them and ask them how they do their job and watch their screens. And you will get ideas right away. I'm very convinced that like you don't have to do, spend more than like a few hours in front of their screens and see what of their repetitive tasks can be done better.
0:51 Speaker 00 Welcome to another episode of Office Hours. I'm Brad and this is Pete and we're here with Gustav and Nikola. We've sat across the table from thousands of founders over the years, digging into just about every question you can possibly imagine. And we want to talk with you today about some of those things that have come up. We are in the age of AI. That can mean a lot of things though, and there's a lot of different directions you can go with that. So what are some of the ways that startups are taking advantage of the fact that there's so many new opportunities coming on board from LLMs?
1:17 Speaker 03 If I were a founder today, what I would be asking myself is what's going to be possible when all of these models become twice as good as they are today?
1:24 Speaker 01 So if you are a startup today, in the old world of how you build the startup, should you pivot to AI? How you see that going? Should you use AI?
1:34 Speaker 00 I have two answers to that. The straw man answer to that is that no, you should not pivot to AI. You shouldn't do that just because it seems like a good idea. But yes, you should almost absolutely be working on something that's using LLMs at the heart of it today.
1:47 Speaker 01 What's an AI company anyway?
1:48 Speaker 00 Exactly. What does it even mean? I think anyone that's using LLMs either internally for to make things more efficient. I've got a company that I worked with a few batches ago that's literally building an HOA management company. Just like straight up, we are an HOA management company. We sell to condo boards and then we do all the stuff for them. But they're using LLMs to make everything more efficient internally.
2:12 Speaker 01 You don't ask people if you are building a cloud company, obviously you're going to use the cloud. Same thing with AI. It doesn't make sense for anyone creating a new company today to not leverage AI within their company, right?
2:24 Speaker 03 That's exactly right. And it's in the early 2010s when a lot of software businesses started moving onto the cloud. There was this moment in time as a founder where there were just a lot of good ideas lying around because you had the opportunity to replace some legacy piece of software that was on prem. And I think we're in a similar moment right now where there's just a lot of good ideas, where you can take some existing piece of software and build it from scratch with AI, a native part of that, and it's going to be much better.
2:55 Speaker 00 I've talked to the CEO of Workday about this, and he was previously at PeopleSoft, and just saw this 2000 or whenever, like, okay, PeopleSoft is huge, but now the cloud is here. Someone's going to build a whole new version of all the same stuff. Just have it be based in the cloud and it'll be better. And similar moment now for a lot of companies.
3:13 Speaker 02 I think if I look back on my time, I arrived here in 2007, roughly the same time for you. When you had these startups, the mobile wave had not happened yet. Like, my first week of YC was Steve Jobs presenting the iPhone. And if you haven't lived through these cycles, it was hard for me to imagine what would happen from that week on. But our company was too early. Like, the apps didn't come until summer after, and the permissions that we needed didn't come until the summer after that, 2009. But the following five years were just like that period where all the interesting mobile companies were generally started. Like, a lot of the stuff that we now use, and I remember in my match was Joe, and Joe sat next to me at a hackathon and was like, "What are you working on?" I was like, "Building Facebook's first mobile app. You can now build apps for the iPhone."
3:58 Speaker 02 And it was like, "Oh, cool." That was my reaction. It wasn't like anything more than that. But if you look back, I realize that we were right in the middle of the beginning of that cycle. And I think because I saw that, or we all saw that, we now see this. And a founder who didn't see any of the previous cycles that actually was real, that had a really big impact, they can't see the, they might not be able to see this cycle in the same way. Big companies generally are preceded by big technology shifts. And this one happened two years ago, and we're increasingly getting more intense throughout every month. And whatever we saw in the beginning of last year, it's like we've moved much further already, and there's a lot more things you can build now that you couldn't build a year and a half ago.
4:40 Speaker 01 And you shouldn't fear existing big companies. They are moving so slow. Like, look, ChatGPT is like two years old, like a little more now. And we still are like such a dumb Alexa. Oh, come. I can tell you it's been obvious for two years that that's the way things change. And still, we have the whole Alexa. So of course, it's going to eventually get there. But yeah, I would not fear big companies' speed of execution. And I think that opens up so many opportunities for startups.
5:09 Speaker 02 And I have a story that I find quite inspiring, actually. It's about a company that applied to Y C. I looked up the application yesterday. They applied in the fall of 2020 as Invest Better. And they were building a financial investment platform. They came into Y C and then within a month or so, they had pivoted and they were building a productivity tool to help you with Zoom calls. This is right in the beginning of COVID, like in January of 21. We got used to being on Zooms 20 times a day. And it turned out that every time you had to open Zoom, you had to go to your calendar, click and click and click and then open. And they're like, no, we just built something to just like drop down and click one time and then everything is perfect. They built that, it's called Superpowered. It became, I thought it was one of the most beautiful consumer products or prosumer products I've seen in Y C. They grow to thousands of DAUs and they grow to like a decent sized business. They can probably survive themselves.
5:54 Speaker 02 They were like at least default live. But what happened is like a year went by and another year went by and they moved back to Toronto. They were here despite COVID. And in Toronto, they were like, L L M's are happening, something is happening. We need to do something else. And they took two very specific decisions in the beginning of last year, which one to move to San Francisco from Toronto. And the other one, which is to just try to stop working. Well, they tried to sell it, but they stopped working on the previous product and then they wanted to build something new. And that company is now called Vapi. And the last 15 months, it's not even 18 months. They've grown from nothing to powering a big portion of the voice AI companies in Y C and outside of Y C. There are other companies like Plan and Retail as well.
6:37 Speaker 02 But I've worked closely with Vapi, so that's a story that I know. It's doing incredibly well and they are a very important part of sort of like bringing this forward. And I think those founders just kind of saw the writing on the wall and they then they did get the insight. Okay, if we move to S F and we. Kind of embed ourselves in this community that was happening right now. Good things will happen, and that was absolutely true. And this company is on fire right now.
6:59 Speaker 00 I think you raise a really interesting point. I have some companies that I've worked with that were not working on AI ideas, and things were not going that well. And they said, OK, we're going to pivot to an AI idea. It also didn't work well. And in those cases, the things that I think they did wrong were, they had kind of a very obvious approach. You know, we're going to be customer support agent company number 50, and didn't really have any new insights on what to do differently there. They didn't change up any of like the environment that they were working in. They didn't embed themselves in any communities. They didn't get deeper into any customer insights. And so we're talking about all this AI stuff, and it's all awesome. But the fundamental things that the founders still need to do
7:41 Speaker 00 to like enable the technology to produce value for customers are the same. And if you're not doing that, just switching your idea over to something that makes calls to OpenAI is not going to change your fate as a startup.
7:54 Speaker 01 You got a good point here. That's what you discovered in Europe here. Like, everything in AI is happening in SF right now. Moving here is actually a great way to open yourself up to new ideas, to new insights, and to not be left behind. And realize when you're later than voice AI is the thing.
8:10 Speaker 00 And to know what the state of the art is. And to know like, OK, I can do this. And to know that maybe that's not impressive enough yet. That's not useful enough yet. That's not going to get a deal done yet.
8:19 Speaker 02 It's very hard to know what you're missing without actually going to experience it. So if you're not ready to move here, well, at least come here for three or four weeks. And like camp out here, go to the hackathons that you want to go to. Try to meet with other companies.
8:30 Speaker 00 Stay at Gustav's house.
8:31 Speaker 02 Yeah. And just try to embed yourself in this community that is now very apparent. And I remember I used to get asked this question in the past. It's like, why is it so good to be in the Bay Area? And I gave this example, and I worked at Airbnb. And I was like, well, we wanted to learn how to be good at SEO. And we're like, who is really world-class in SEO in the world? Pinterest. Where are they? Oh, they're downstairs. So we walked downstairs to the Pinterest, and we talked about the MOSS. We learned from the Apple bass team. And I think that is the state of LMS here right now. Like, you go next door. Or three doors down and then you have the company that's probably the best at the thing that you want to learn about.
9:05 Speaker 00 And imagine if you're like in Chicago or someplace like that and you have the same insight, oh, I need to talk to someone at Pinterest and you're lobbing emails across and liking someone on LinkedIn and all this stuff and it's just not going to happen for you the same way.
9:17 Speaker 02 It's not going to happen and I think the advice could not be more clear, I think, now compared to any time in the last 10 years, that you should come here, at least temporarily, but like you should consider being here permanently.
9:29 Speaker 00 So it sounds like we've uncovered a one-two punch of yes, you probably should pivot to AI with some caveats and you should come to the Bay Area so that you do it in the best way possible.
9:41 Speaker 02 Should we talk about other applications of AI, of things you've seen, like other parts of the world where AI is having a big impact and anyone has a good example?
9:51 Speaker 03 Yeah, sure. So just looking at the last batch, the first ever fall batch that we just completed, a few of the companies I worked with, so there are two companies that are working on different problems, but they're solving them in a similar way. And so one is a company called Replexica. They're building software that automatically translates your UI from one language into another. So they automated AI localization. And then another is called Gecko Security and they are an AI security engineer. And so both of these companies are taking some previously specialized skill and automating it with AI, putting it behind an intuitive interface and giving software engineers the ability to manage this independently as part of the release cycle. There's another example of
10:35 Speaker 03 a company I worked with in the last batch, who all worked at an insurance technology company before this. And because of that, they had a lot of exposure to the workflow that Medicare Advantage agents go through selling Medicare Advantage insurance plans. And this is a relatively obscure workflow. Most software engineers don't have much experience with that, but because of that experience, they were able to build an AI copilot for Medicare Advantage insurance agents and it's doing very well.
11:06 Speaker 02 That's interesting. Like I have a whole framework in my head for how to go after the healthcare section. Like you said, most people that are building for healthcare have never had a job in healthcare where you actually get, come across these inefficiencies.
11:20 Speaker 03 Yes.
11:20 Speaker 02 But the US healthcare system, I believe it's like something like $4 trillion in spend, of which 1.3 or 1.4 is admin.
11:28 Speaker 02 If you look at the companies that we funded in Y C, it seems like much of this admin is legacy software system plus a human who's moving data from one system into another. And that's what most of this actually is. And that's shocking. And we have more admins per doctor in the U. S. And many, many, many other countries in the world. And I think it has a combination of like, we have lots of softwares and lots of people. And then we have incentive structures that kind of everyone needs to make money in this system. Now, most of the tasks are fully automatable. So there's a company that's doing pre-off, Ti vara, and they are basically taking the information from the doctor plus some other information. And they're summarizing it and they're automatically creating the pre-off request that goes into the payment portal. And you can now do this fully with L L M's.
12:12 Speaker 02 And this is one out of like dozens and dozens and dozens of these manual repetitive tasks that happens as a result of our inefficient healthcare systems. We basically are moving things between one legacy system and another. Like it's both software portals. And this human is translating the information in between. With agents, you can do almost all of these tasks. I'm pretty convinced that like healthcare is not just front desk scheduling. It is like every single back office task. But founders don't know what these tasks are. So the challenge I have to someone who's looking for any healthcare is like, go to your friend who works in this and go and sit next to them and ask them how they do their job and watch their screens. And you will get ideas right away. I'm very convinced that like you don't have to do, spend more than like a few hours in front of their screens and see what of their repetitive tasks can be done better.
12:56 Speaker 01 You can go even further with healthcare. You can actually help patients too. I have a company, Zeez Batch, who's actually using voice A I to call patient in between visits to make sure they are doing well and possibly to schedule a new visit when necessary. It's very simple, but my God, for the practice, first more business, but also it's a way better experience for the patient.
13:14 Speaker 00 Awesome. So those are some great examples that we just heard about of companies that are using L L M's to do awesome stuff and to grow much faster than we've been seeing companies grow for some time here within the Y C portfolio. I hope that you listening and watching this, maybe you've gotten some ideas that can encourage you as you're thinking about whether you should change idea, whether the idea you're thinking of starting is on the right track, whether you should stay where you are, whether you should come here to the Bay Area. We just see a ton of opportunity here and we see a lot of things happening that are tremendously exciting to us. And our jobs has been a really fun, fun time to be group partners here at Y Combinator. And we hope some of this is useful to you. Thanks so much for watching and we'll see you on the next Office Hours.
Previous How To Build The Future: Parker Conrad Next How To Build The AGI Future: Bob McGrew Page updated October 12, 2026 · transcript and summary are machine-generated
