Airbnb launched its new AI-powered search this week as part of a fall update, marking a notable step for a company that has taken a cautious, deliberate approach to AI. Brian Chesky, Airbnb’s co-founder and CEO, spoke over Zoom about how the company is approaching AI, where interfaces are headed, why consumer AI is not yet solved, and the role voice will play.
Why a chatbot-only interface falls short
Chesky reiterated that chatbot-only interfaces are not well suited for travel discovery or browsing: they typically present a few options at a time and require multiple turns to reach a result, which undermines the visual and inspirational experience users enjoy. He also said the version of Airbnb’s search released this week is not the final form and that the interface will evolve.
He argued many people derive pleasure from planning and anticipating travel, so Airbnb should preserve browsing, inspiration, and the library-like discovery experience. He also noted a second shortcoming of current chatbots: they are built for single users, whereas travel planning is often collaborative and used by families or groups.
"Multiplayer" AI and the next three to six months
Over the next three to six months, Chesky said Airbnb’s focus is to explore interfaces that enable “multiplayer” AI — AI that multiple people (for example, a group planning a trip) can use simultaneously. He mentioned startups such as Monogram and Wabi experimenting with generative interfaces, where an AI builds screens on the fly rather than a designer fixing them in advance. Chesky expects a mix of deterministic (pre-designed) and generative interfaces going forward.
Agents, interoperability, and the future of apps
Chesky said Airbnb is preparing for consumer AI agents like Instinct and Muse, which act on users’ behalf. These agents will require robust software infrastructure on Airbnb’s side, and the company wants to become more agent-friendly. He suggested agents could serve as effective lead generators.
He also described plans for agent-to-agent interactions on the platform: eventually Airbnb’s Explore tab, customer service, and other services could each be agents, and a macro Airbnb agent might interoperate with other agents through standards such as MCP. He noted that historically apps only became interoperable through business deals, but agents—“apps with faces,” in his words—could enable broader interoperability when they communicate with one another.
The missing AI operating system
Chesky argued a major industry gap is the absence of a true operating system designed for AI. Today, builders create AI apps for iOS, macOS, and Windows, but not an "AI operating system" that manages agents and components at a lower substrate. Ideally, he said, AI would be operable at the kernel level so all components can interact; otherwise the market becomes a race to become the single “quarterback” agent that users go through for everything.
He referenced ChatGPT’s early attempt at an app-store-like model at the beginning of last year and told Sam Altman that to be like the iPhone App Store, a platform needs a software developer kit and an operating system. Without such tooling, Chesky said, functionality will be limited. He shared that his experiences using Airbnb on Muse and Instinct were not smooth and concluded: “I don’t think we’ve cracked consumer AI.” To do so, he believes agents must become apps, all apps must become agents, everything must be interoperable, and user interfaces must become richer.
The role of design and platforms
Chesky identified two misconceptions in Silicon Valley: first, that the interface will disappear and people will only talk or type; second, that software will become entirely generative. He argued a skilled designer can prompt better software than a typical consumer can.
He also noted that as long as the iPhone remains dominant and Apple and Google control major platforms, apps are likely to remain central. A true shift from apps to agents, he said, would require Apple, Google, or another platform provider to build a new foundation; otherwise apps will contain agents rather than becoming fully agentic and interoperable themselves.
Voice and internal benefits of AI
Chesky said he expects voice interaction to grow and eventually people will speak to computers more than they type. Voice will be important for Airbnb’s search and customer service; the company plans to roll out voice agents this fall to assist users.
He also volunteered that AI has helped Airbnb ship features faster and has improved his personal productivity by reducing the number of meetings he needs. “AI is reducing communication layers,” he said, describing how he can now ask his AI for information instead of scheduling as many meetings.
Conclusion
Airbnb’s fall AI update signals a continued emphasis on user experience and agent integration. Chesky said the coming months will focus on experimenting with multi-user, agent-friendly interfaces and advocating for a lower-level, interoperable AI infrastructure to enable consumer agents to work reliably.
