Chesky: AI agents need an operating system of their own

Airbnb rolled out AI-powered search this week as part of its fall update — a significant step for a company that has been slow and deliberate about AI. In an interview, CEO Brian Chesky said he hasn’t changed his view that a chatbot is the wrong interface for travel discovery: it gives users only a few options at a time and forces multiple turns before reaching a result. He doesn’t see the new search as Airbnb’s final interface either, saying some middle ground between a chatbot and what the company just shipped will likely emerge. People enjoy browsing and planning trips, he said — travel studies show people get more pleasure anticipating a trip than taking it — so Airbnb doesn’t want to design that dreaming experience away.

Chesky also sees chatbots as poorly suited to group decisions, since they’re built for one person at a time. Over the next three to six months, he said, Airbnb’s task is to explore “multiplayer” AI interfaces that let several people — a family planning a trip together, for instance — use the same AI at once. He expects a mix of pre-designed, deterministic interfaces and “generative” ones an AI builds on the fly, the kind startups like Monogram and Wabi are experimenting with.

On AI agents that act on a person’s behalf, such as Instinct and Muse, Chesky said Airbnb wants to make its own app more agent-friendly and build richer interface controls, since he doesn’t believe apps will simply become data layers for a single universal interface. He’s also planning for agent-to-agent interaction, envisioning Airbnb’s explore tab, customer service, and other functions eventually becoming their own agents under one “macro” Airbnb agent, interoperable with other agents through the MCP standard.

Chesky’s core argument is that the industry lacks a true operating system built for AI — one where agents interact at the software’s core, the way apps interact through an OS today. He pointed to ChatGPT’s short-lived app store from the beginning of last year as a cautionary example, recalling that he told OpenAI’s Sam Altman the project needed its own developer kit and operating system to succeed. “The problem with that model [of trying to become primary agent] is no one is developing a full operating system software developer kit,” he said. “And of course, it didn’t work very well.” Voice will also play a bigger role for Airbnb, with voice agents for search and customer service planned for this fall.

Why it matters

Chesky’s framing suggests the next competitive battle in consumer AI won’t be won by whichever chatbot app is most popular, but by whoever builds the operating-system layer that lets agents from many companies work together.