Airbnb CEO on Why AI Will Create a New Era of Consumer Products

0:00Brian Chesky:Brian, how are you doing? Good. How are you guys doing? We're doing fantastic. Welcome back to the show. It's been too long. Yeah. Why don't we kick it off with just an update on how 2026 is going, how the business is going, what's new in your world? Yeah, I mean, things are really good. We've accelerated growth for the first time since the pandemic. We grew 10 % last year in revenue. And this quarter, we announced that revenue was 18%. So from 10 % to 18%, which is a pretty big acceleration. Yeah. Thank you. And marketplaces are really, really hard, right? It's kind of like gravity. Once a marketplace at our size doing around$100 billion in gross bookings a year started to come down, it's really, really hard to tip that curve.

0:40So this has been a pretty big feat for the team to be able to do it.

0:43Brian Chesky:So how'd you do it? You know, it's kind of funny. A number of years ago, I noticed that as we got bigger, we started losing a little bit of that startup intensity, that startup energy. And I asked myself, how can we get that energy back? And we basically took a very small team. We named the team Project Hawaii. The name doesn't really matter. But basically, I took a very small team of people, and we said, we're going to focus on a very, very small service area. And we decided to focus on conversion rate, the guest journey. And I basically tried to work with a team as if we were at Rouse Street.

1:17Rouse Street was the apartment that Jona and I started. And I said, we're going to act like a startup. And the team just grinded really, really hard. We weren't working like a big company. We were a very small team, grinding really hard, focusing on obsessing over the customer experience, really looking at the data, and we really got a lot of points in the board. And then we really started taking these pods and we really started working with the teams, trying to coach them how we worked in the early days. And I just think the pace increased, the intensity increased, we really tried to bring in world-class people onto the team, they got very, very focused, we tried to get all the management and bureaucracy out of their way.

1:51And it's been a couple years. In fact, we were doing this last year, but one of the things is, as you know, financial results are lagging indicators, especially when you're a big company. It takes sometimes a while to get the financial numbers to reflect what's happening inside the company. But I feel like we're a startup again, more than ever. We feel much younger and smaller than earlier. I think with AI, that would be the other thing, is 60 % of our code is now written by AI, which is twice a benchmark of our competitors and peers. It's really, really helping us. AI, I think is a huge boon to us.

2:25I don't know if it's helped the OTAs, but it's helped us with customer service. The cost per customer service tickets down 10%, 40 % people who contacted me, the AI solves the problem for them. And we brought it through the entire journey. So everything is really accelerating.

2:40Brian Chesky:On the special team that went into optimized conversion rate, I imagine you have folks internally whose job was basically the customer journey already. then you bring in your special team. And is there some sort of culture clash there? Like, how do you set people up for success to not get too political in that environment, actually see it as an opportunity for a win, some fresh eyes, some fresh ideas? Like, what is required to actually have success? Because I think a lot of big companies that bring in McKinsey and they put together a big deck and everyone freaks out and thinks they're getting fired and maybe some of the good ideas are surfaced, but it never really goes through.

3:20Brian Chesky:So what do you have to communicate to the team that is receiving information from this new QI team? Yeah, so it's a great thing. I mean, really what I'm talking about has played out over like really five years. And the term founder mode, what it really meant that Paul Graham wrote was about me like skipping layers of management and going into the details with teams. And instead of trying, I mean, here's my advice. If you're a CEO or a leader of a company that's big, don't try to change the whole company. try to change a corner of the company. It's kind of like don't renovate the whole house, pick one room and make it perfect and then go room to room.

3:55So I really told teams, hey, I actually didn't replace the team. I took the team that was already working. I handpicked some people on the team, but I really just taught them the pace and I would review work very regularly. So I'd sometimes review the work weekly or even daily to just teach them a level of intensity, a level of perfection. and it was unfamiliar and uncomfortable. And I will tell you that not everyone liked it. And some people didn't stick around the company because they didn't like that way of working. But those who stuck around and the vast majority did, they realized, wow, actually when the CEO is involved, it's actually easier.

4:32There's less bureaucracy. I try to make the work better and I try to clear all obstacles. And so I basically did that group by group. And the great thing is like, I was in massive number of details reviewing everything for years, but eventually it's like muscle memory. It's like an instructor or a coach. You have to teach them, but once they learn and they have the muscle memory, like a golf swing, you can step back and now it's muscle memory. So now I don't have to be in all these details, all these teams. But I would just say a couple things, people listening, like great leadership is presence, not absence.

5:04I think a generation of management consulting or management school, management teaching taught us that CEOs should trust their people and get out of the way. And I don't think trust and get out of the way are the same thing. Again, I think leadership is presence. And you should actually be partnering with your people. You should be on the field with them. If you're a cavalry general, you should be on a horse. You should be on the battlefield. You're not overseas somewhere else just writing out blueprints. You've got to be on the field with them. And you've got to be leading from example. I think leadership is from the front, not from the back.

5:36And so these are just some of the things we do. And I think AI is going to create the equivalent of an AI founder mode, which is now you've got to be even more hands-on. And I don't think in the era of AI there should be any pure people managers because you're so close to the details, to the data, that everyone has the opportunity to be hands-on. And it's hard to imagine only managing people and not agents. So I think this hands-on approach is for everyone inside of a company.

6:03Brian Chesky:Jordy, please. Do you think a lot about what you would do if you were you from 10 or 15 years ago and you were trying to disrupt Airbnb? Like, is that a helpful exercise? Yeah, and it's slightly scary because I think that I've told, I had a meeting with our team recently. I said that 26-year-old me and Joe and Nate could F us up if we wanted to. And so I told them this is what I would do. But that's because you understand this market better than anyone else. You know the key drivers. Like, I don't think any 26-year-old off the street. Hey, I'm paranoid. I think that if everybody sits still, I think a different group of 20 or 30 year olds could also disrupt us.

6:48And I think that's true of every one of us. And maybe, by the way, maybe that's not true. Maybe that's not reassuring because we do have a brand that's a noun and a verb. We've got a global network effect. I think it's actually hard to build a network. That may not be possible. But the software and the app, I told our team, like, we can't sit still. Like, our app is beautiful. It's really nice. But we got to be in a world of AI native. And here's a key point I would make. I do not think a chatbot is the right interface for travel or e-commerce. That might be a radical statement. ChatGPT launched third-party apps last year.

7:20In March, they shut them down. I don't think a chatbot is the right interface. It's got four or five problems. The first problem is it's text-based. Photos are an afterthought. The second is there's no direct manipulation. You have to type every single prompt, which is fine for a conversation, but you can't add filters. You can't cook around. The third problem is it's hard to compare. A lot of e-commerce and travel is comparison shopping. If you have thousands of options, the AI has to know exactly what you want to be able to show you one or two things. But you usually want to see more choices and you get lost.

7:53And most AI is single player. It's not collaborative, let alone the fact that Airbnb requires people to have an account. 85 % of people send a message. So what I'm trying to do is imagine. Sorry to push back there, but my wife's planning a vacation, and there's like three-ish hotels that we're looking at, and we're looking for specific dates. And she, theoretically, I'm not saying the products are there yet, but theoretically, you could ask a chatbot, I'm interested in staying in this location. I'm interested in these hotels. I have X, Y, Z number of people and it could go and the agent could go and pull together like relevant sort of like listings or room types, et cetera, pricing, show me pictures, and then actually do an analysis of the trade-offs based on all the information available as well as information in other parts of the internet and pull it together.

8:52And then she could share that chat with me and we could both review it together. Yes, I agree. That's the future and that's not a chatbot we described. That's not a chatbot. It's going to be a completely different interface. It's going to be, well, I guess you'll have to wait and see, but I think the future are not apps. The future are agents, but I don't think they're going to be text forward. I think they're going to be really rich user interface. And so I think the current chatbot paradigm, what you're describing, it can do it. I just don't think it's the best way to do what you just described.

9:23I think there's a much more immersive way to do that. Yeah.

9:27Brian Chesky:As you think back, I mean, it feels like at various points, clearly there's been like, oh, is all of this going to move to chatbots? Or is all of, is, are there going to be a million competitors that are all vibe coding exactly what you have? And so is software remote. The stock has not been beaten down during the SaaSpocalypse, but have you had to process those with investors? Have you had to walk people through Airbnb strengths again? I'm sure you did after the, like, the 2028 intelligence crisis. Thankfully, they picked DoorDash instead of you guys. But thank God, I really appreciate them picking on someone else for once.

10:05But, yeah, no, absolutely. I think there were entire, like, actually when ChatGPT launched a third-party app, their stock probably went down, like, 7%. And, by the way, I thought it was a really good idea for them to do third-party apps. I think it could have been successful. But they would have needed a richer SDK for it to work. And it would have, like the App Store, Apple's App Store was good for Airbnb, and it was good for every company, because they had a really rich user interface. Maybe this is the point I'm making. That imagine using iMessage on your iPhone to do everything, when in fact, like, every other app has a unique interface.

10:36So what I think is e-commerce, you want a very rich user interface. It would be agentic, you could be able to have a conversation with it, you could talk to it, I could talk back to it, but I think the point is it has to be more visual. I think a text-based interface is for some solutions, and a chatbot that's visual, yeah, that would work. I'm just saying today's chatbots aren't the right solution. But the answer to your question, I had a couple new employees. One of our new team members is a guy named Ahmed, who is the CTO of Airbnb. Now he led the LLAMA models at Meta. And one of the comments he made to me was he said, wow, yeah, Airbnb is so much more than the app.

11:13And in fact, the app that you see is like 20 % of Airbnb. We have typically four or five million people staying at Airbnb every night in more than a hundred countries around the world. So there's a lot more to do around payments, around customer service, adjudicating everything. We have a$3 million guarantee against theft or property damage for a million homes a night. That's$3 trillion. There's just so many types of things around managing five and a half million hosts. We have a host app. So there's a lot of things that are beyond the guest app. I actually think the guest app would be pretty easy to copy.

11:45And I think in an age of AI, you can make a better app than ours. And we want to make that app before anyone else does. And we want to be agentic. But I think the key is most of Airbnb is not the app that you use. It's mostly the offline experience. It's the operation.

11:57Brian Chesky:Yeah, that makes sense. Yeah, the iMessage example is good. It's maybe even going deeper into old SMS. Because iMessage has hydrated so many things with the reactions. And you paste a link, and it hydrates it. And it's becoming more of a visual tool. But it is a long road. Question for you. things. Yeah. Well, before you say it, one of the things is the last time I was on, I did make a comment and I'll make it again. Almost every AI company is an enterprise company. And the last stat I think I said on TBPN was like three months ago and stats haven't changed. I think it was 175 companies in a YC batch and I think 16 were consumer.

12:33That trend hasn't changed. Almost everyone is going into coding. Almost everyone's going to enterprise. I think that's great. My whole point though is that the consumer experience hasn't fundamentally changed that much being a chatbot. And I think the consumer is a massive opportunity for AI. And I think it's going to need to be a richer user interface. I think people that want to do more things, they want it to be. I mean, the modern AI today is text, photos, and buttons. Yep. Mostly, in some videos. And I think there's a more breakthrough visual paradigm that can be much more immersive. And with the new image and video generation models, you could do something so much more immersive.

13:10So the exact example you gave of trying to book a hotel, yeah, you can do it on a chatbot, but there's probably some more breakthrough way to imagine that, to visualize it, to see the neighborhood, to see the map, to be able to talk to it, to understand where it is, to be able to compare photos to different hotels. There's something, I think, richer on the horizon. Yeah, I was surprised at the push into browsers last year from some different AI companies. specifically because I already felt like an LLM, like the chatbots were helping me browse around the internet. It was like just double down on that and helping me find information all over the internet.

13:48I like that it's brought together in a standardized way, right? If like the example, you know, of comparing like different hotels, right? It's nice to have it just like formatted the same way. So I'm looking at it, I'm not being like, I want to be influenced by the images and like the property, but not necessarily the design of the website, right? Because that they can be disconnected. I wanted to ask if you think LLMs will impact travel in the world in the way that social media did and travel trends specifically because Instagram nowadays, like a place like Marfa and like Texas, right? It's like, it's random town or, you know, artistic type town in Texas.

14:29and then it just becomes a cool place because of Instagram, and then the entire kind of town evolves because of that. I can imagine people researching places to go with LLMs could ultimately drive some of that and then ultimately reinforce each other. But what do you think? Yeah, maybe the simple framework is think about a travel journey. Step one is destination discovery. Where should I travel? Step two is flights. How do I get there? Step three is where do I stay? Airbnb or hotel. Step four is like, what do I do when I get there? Restaurants, activities. And step five is typically like logistics, car rental, services.

15:09And then step six is you're in the city. And then you might want to do things that time spontaneously. I think LM, or let's call it the LM technology, applied to, I'm arguing, a slightly richer user interface than typical chatbot, will be revolutionary for step one, step two, for destination discovery and flights. also because flights are not very hard to build there's just three global distribution systems anyone can pipe in an API and have a flight booking app pretty pretty quickly and so I think the LMs are really really good at destination discovery if you want to say like hey I want to go somewhere that's like Paris but it's a little more affordable it's good to go in August I like Opera like it's going to give you like a very rich like suggestion of places to go now I think the current chatbot doesn't have really rich maps.

15:55I think it could do a lot more visually. I think eventually chatbots or this new interface could be very video-based. So imagine a chatbot that was actually video-based or very, very photo-based rather than a little less text-based. I think that would probably be what you'd want for travel. I think what I'm describing would disrupt travel more than the current chatbots, which are more acting like Google sending referral traffic. And they actually, the referral traffic from these chatbots are converting higher than Google. And so actually the chatbots are actually additive to travel companies. To disremediate travel company, you would really want to disremediate the travel journey.

16:31It would have to be much richer.

16:33Brian Chesky:Yeah, I feel like, do you remember that company Hipmunk? I think it was founded by some YC founders, right? Yeah, it was really cool. Yeah, actually the CEO of Reddit. That's right. Steve Huffman, right? Wait, are you talking about Hipcamp? No. Hipmunk. Hipmunk was a really cool flight booking. dashboard and it would rank the flights by pain, like a pain index. So it would say, well, you're not going to face financial pain because it's a really cheap flight, but you do have a stopover or you do have to get to the airport super early. And so it would blend all these things together. And it's something that it feels like could be vibe coded over a weekend now.

17:13Brian Chesky:And yet it is weird that when I open up the app store charts, I see ChatGPT, Gemini, Claude, and then it's just AI chat apps. And we haven't had that breakout moment. Demis was on stage at YC actually talking about the fact that it feels weird that we have this super intelligence in the enterprise and vibe coding and 60 % of Airbnb's code is written by AI now, and yet we don't have a new AAA game or a new video game that everyone's playing. And I'm wondering, is that a lack of creativity? Is it a lack of risk-taking that there's just too much money? It's too obvious to be able, if you're good in AI, just go into enterprise because you'll just raise a bunch of money, get a bunch of customers.

17:54Brian Chesky:It's really easy to make money. It's more risk-on in the consumer. What do you think needs to happen? Or is it just the intelligence isn't quite there and maybe the next model is the one that unlocks it? This is such a good question. I think there's like three or four factors going on. The first factor is I saw a tweet recently that I think there's like 60 new Neo Labs being formed. Yeah. And I've met a number of these researchers, and a lot of them are interested in doing the same things. And I think what I'm noticing is most of these teams are purely AI people. They don't have product people or designers on their team.

18:27They're picking things like science. They're picking things like coding. They're picking things like we want to create a different type of model. This is great, but I don't know if all 60 companies should be doing the same problem. And so what you're seeing is you don't see a big focus on people wanting to do consumer. I think some people think that the AGI can just figure out consumer. I think that's maybe overly simplistic thinking. I do think you have to have a point of view about consumer. So I think that's the first point. I think the second point is Silicon Valley has become more vibe and trend-based than when I came to Silicon Valley in 2007, although it was back then.

19:02And one of the things I think when people see enterprise companies doing well, they go into it. And I think there's this natural flywheel. The other thing is I know in Y Combinator, for example, we teach the Y Combinator companies to use the other companies as their customers. Now, back in the day, we would get them to use our product as consumers. But I think everyone's figured out actually it's way more efficient to get them to become customers as a business. Like five code gen startups all being like, you code with me. Yes, exactly. And so I think enterprise makes a lot of sense until everyone does enterprise.

19:36And then suddenly, everyone's saying consumer's hard, but you know what's also hard? Competing with 10 other companies in enterprise doing your idea, too. That's also hard. I think additionally, the image in video generation models are having a... I think image two is almost like a cloud code breakthrough. It is so good now in the C-Dance video models that I think we're about to enter this new era where suddenly going beyond text is possible. I think you're going to even be able to see real-time AI-generated interface in the next year or two. So my prediction is this has been the era of enterprise.

20:12Even ChatGPT's breakout success, I would predict that most of the revenue is going to be going to Codex. I might be wrong, but just given how much money is going to Claude. And I think in the next two years, you're going to see a massive revolution in consumer. And I think you're going to need a couple of companies to lead the way. I think we need reference points. I mean, even me saying the chatbots, not the interface, people are probably asking, well, what is? And someone needs to be the one to do that. I mean, we're going to try to do our part. I'm not sure Airbnb will be the company to blaze the trail, but I think someone has to.

20:44I think in another era, it would have been Apple, like the Steve Jobs era. Apple would have done that. Maybe Apple will do that, but they would have been typically the company that would have done that.

20:54Brian Chesky:Yeah. Are you worried that everyone is working on the exact same thing? And when I say exact same thing, I mean agents that can do things on your computer and or, you know, look at the iPhone app store chart or look at look at the early stage startup market. Right. A lot of companies, you know, the full spectrum of companies from hyperscalers to labs to neolabs to. I'm not I'm not too worried. I might be a little worried for the entrepreneurs if they're like the 10th company doing something. I mean, like Peter Thiel was one of our earliest investors and he used to hear this book Zero to One that a lot of people listening probably read.