Airbnb rolled out new AI-powered search this week as part of its fall update, a significant step for a company that has been cautious and slow when it comes to AI. Co-founder and CEO Brian Chesky has said in the past that a chatbot-only interface is not suitable for travel discovery. As the Airbnb app continues to incorporate more AI, I caught up with Chesky on Zoom to talk about how the company is approaching AI, where the interface is heading, why consumer AI isn’t working yet, and what role voice will play.
Chesky hasn’t changed his mind about chatbots. He said these are not suitable for browsing or shopping as they display several choices at once and require multiple turns to reach a result. But he doesn’t think the current version of Airbnb’s search is its final form. Its interface will evolve over time.
“I think I’ve believed for a long time that chatbots are not the right interface for e-commerce. I don’t think what you’re seeing today is the end-game for e-commerce or travel or shopping. I don’t want to say a middle ground, but I think there will be some kind of interface between chatbots and what we saw in the first version that we shipped,” he said.
Mr. Chesky continues. “As you know, a lot of people want something…quick and easy, like ‘Book a flight. I don’t want to see the plane.’ On Airbnb, we enjoy looking at products. Travel research shows that people get more pleasure from planning and anticipating a trip than actually taking a trip. So we don’t want to take away the inspiration, the browsing, the dreamy experience you get at a library.”
Chesky believes there is a second problem with chatbots. That is, chatbots are built for one person. He feels that these are not collaborative interfaces, but this is important for products that are frequently used with family and friends. Over the next three to six months, the company’s challenge will be to explore interfaces that enable “multiplayer” AI, or AI that can be used by multiple people at the same time, such as a group planning a trip together.
Airbnb is not alone in rethinking what an AI interface should look like. As AI tools evolve, new interfaces are emerging in apps. Startups like Monogram and, more recently, Wabi are experimenting with “generative” interfaces, screens that AI builds on the fly rather than screens that are fixed in advance by a designer. Chesky thinks people will see a mix of deterministic (pre-designed and predictable) and generative interfaces.
I asked Chesky how Airbnb prepares consumer AI agents (tools like Instinct and Muse that perform tasks on behalf of humans). He said agents need a robust software infrastructure on Airbnb’s side, and the company wants to make itself more agent-friendly. In his view, they can be great lead generators.
“I don’t necessarily believe in a world where apps exist in the future, but I don’t necessarily believe in a world where apps are a data layer with a universal, simple UI. There’s a reason why when you open your iPhone, Airbnb’s app has a different design than DoorDash, Shopify, or Amazon. Our interface has to adapt to the utility and the functionality that we’re doing. We want our apps to be more agentic, and we want them to be more agentic and more user-friendly. “We want to build an interface. We want to control it,” he said.
“If you have an agent, but you also want really good browsing capabilities, if you want to be able to message hosts like on Airbnb, if you want to be able to compare, if you want to be able to verify your identity, if you want to be able to use the map feature, if you want to be able to add other items, if you want to stay on the platform, the agent will do the handoff, or the agent I think the application needs to build a little more robust software, so to speak, a developer kit for delivering apps.
Chesky said he believes there have been concerns about AI tools separating companies like Airbnb from customer relationships, but chatbots have proven to be better at generating leads.
The Airbnb CEO also plans for interactions between agents through the platform. He explained that there may be some interaction within the platform or with agents from other apps. His point is that apps won’t always be interoperable unless companies have a business agreement. Agents, on the other hand, are like “apps with a face”; when agents talk to each other, all apps become interoperable.
“I think the Airbnb app will one day be an agent. Basically we have an agent in the Explore tab, we have an agent in our customer service. We’re developing other agents in different service areas. We’ll probably end up with a macro Airbnb agent. And it’ll probably be interoperable with MCP (a standard that allows AI agents to connect to other software) and other agents,” Chesky said.
The bigger problem, Chesky argues, is that the industry lacks a true operating system built for AI. Today, people are building AI apps for iOS, macOS, and Windows, which are not “AI operating systems.” He believes that in an ideal world, AI would be able to operate at the kernel level (at the core of the operating system) and all components, such as agents, would be able to interact with each other. He said the race now is to become the “quarterback,” the main AI agent that users go through to do everything.
“The problem with that model[of trying to be a primary distributor]is that no one has developed a complete operating system software developer kit. Let me give you an example,” he continued. “ChatGPT tried to do a small app store early last year, and I said to Sam[Altman]: If you want to be an app store like the iPhone, you need a software developer kit and an operating system like the App Store. Otherwise, none of our features exist. And of course it didn’t work out very well,” Chesky said.
“I’ve used Airbnb with Muse and Instinct,” he continued. “And it doesn’t work very well. I mean, even hotels don’t work that well, but Airbnb works even less well. So I don’t think we’ve been able to crack consumer AI. And to crack consumer AI, you have to crack it at a lower base, and you have to (turn) agents into apps. All apps have to be agents. All apps have to be interoperable, and all have to have richer user interfaces.”
Chesky believes there are two misconceptions in Silicon Valley. One is that interfaces will go away and people will just talk or type. Second, the software will be completely generative. His argument is that designers can prompt better software than consumers.
He also believes that as long as the iPhone remains dominant and Apple and Google continue to dominate the major platforms, apps are likely to remain central.
“It’s up to Apple or Google or whoever to build a new platform for us to make a true transition from apps to agents. Otherwise, we’re going to end up with agents built into apps instead of whole apps being robust agents that are interoperable with other agents,” he said.
Finally, I asked Chesky about audio. With apps like Wispr Flow, Willow, and Superwisper, a new wave of note-taking apps, and dictation built into AI tools, voice is becoming more prevalent as people increasingly speak into their devices. He thinks people will eventually talk into their computers rather than type. He said voice plays an important role for Airbnb in search and customer service, and the company plans to roll out voice agents to assist users this fall.
The company also benefits greatly by using AI to deliver features faster, he volunteered. On a personal level, Chesky said AI allows him to get information faster and saves him from many meetings.
“I think AI is reducing the communication layer. One of my jobs as a CEO is to get information, and the way I used to get information was in meetings. Now I can just ask the AI a question and I don’t need to get as much information,” he said.
If you buy through links in our articles, we may earn a small commission. This does not affect editorial independence.
