Airbnb's new AI search, rolled out this week, marks a deliberate step for a company that has moved cautiously on AI. In an interview, CEO Brian Chesky reiterated his long-held view that chatbots are the wrong interface for e-commerce: they force users through multiple turns and hide the browsing and inspiration that travelers enjoy. He described the current release as an early form, not the endgame, and predicted a hybrid interface somewhere between a chatbot and a fixed screen.
Chesky also highlighted a second limitation: chatbots are single-user, while travel planning is often collaborative. Over the next three to six months, Airbnb plans to explore 'multiplayer' AI interfaces that groups can use together. He acknowledged that startups like Monogram and Wabi are experimenting with generative interfaces, but he expects a mix of deterministic and AI-built screens rather than a wholesale replacement.
On agents, Chesky sees them as potential lead generators rather than threats, provided Airbnb builds the right software infrastructure. He wants the app to become more 'agentic' and to support handoffs between agents and rich user controls. He also envisions agent-to-agent interactions, with Airbnb's own agents interoperating via standards like MCP. The bigger problem, he argues, is that the industry lacks an AI-native operating system—one where AI works at the kernel level. Without that, he says, attempts like ChatGPT's app store will fail, because they lack the developer kit and OS foundation that made the iPhone ecosystem work. His point: the race to be the 'quarterback' agent is premature until the underlying platform exists.