Airbnb CEO Brian Chesky Sounds Alarm: AI Agents Demand Their Own Operating System

Airbnb recently launched its new AI-powered search as part of its fall update, marking a significant step for a company that has adopted artificial intelligence deliberately. Co-founder and CEO Brian Chesky has previously expressed skepticism about a chatbot-only interface for travel discovery, but with more AI features slated for the Airbnb app, he shared his insights on the company's AI strategy, the evolution of interfaces, challenges in consumer AI, and the future role of voice technology.

Chesky maintains his stance that chatbots are not ideal for browsing or shopping in e-commerce, particularly for travel. He believes they offer limited options at a time, requiring multiple interactions before yielding a satisfactory result. Furthermore, he points out that current chatbots are inherently designed for individual users, lacking the collaborative functionality essential for group travel planning, which often involves friends and family. However, Chesky views the current iteration of Airbnb's AI search as an intermediary step, not the ultimate interface. He anticipates an evolution towards a hybrid interface that bridges the gap between traditional chatbots and existing search UIs.

The CEO emphasizes the importance of preserving the inspirational and browsing aspects of travel planning, noting that studies suggest people often derive more pleasure from anticipating and planning a trip than from the trip itself. Consequently, any AI interface must not diminish this 'library experience of dreaming.' Over the coming three to six months, Airbnb plans to explore 'multiplayer' AI interfaces, designed to facilitate collaborative use by multiple individuals planning a trip together. This aligns with broader industry trends, where startups like Monogram and Wabi are experimenting with 'generative interfaces,' which are screens dynamically built by AI, moving beyond pre-designed layouts. Chesky predicts a future with a blend of both deterministic (pre-designed) and generative interfaces.

Addressing the emergence of consumer AI agents, such as Instinct and Muse, which perform tasks on a user's behalf, Chesky stated that Airbnb is preparing by building robust software infrastructure to make its platform 'agent-friendly.' He sees these agents as potentially excellent lead generators. While questioning the long-term existence of traditional apps, Chesky argues against a future where apps merely become data layers for a universal, simple UI. He believes that interfaces should adapt to the specific utility and function they serve, hence the distinct designs of apps like Airbnb, DoorDash, and Amazon. Airbnb aims to make its app more 'agentic' and incorporate richer user interface controls, positing that for a truly robust experience, an agent application would need to develop a comprehensive software developer kit (SDK) to integrate the full app functionality.

Chesky acknowledges concerns that AI tools could disintermediate companies like Airbnb from their customer relationships but notes that chatbots have proven effective for lead generation. He also envisions agent-to-agent interactions on the platform, both within Airbnb's ecosystem and with agents from other applications. His thesis is that while traditional apps often require business deals for interoperability, agents, acting as 'apps with faces,' can inherently achieve interoperability when they communicate with each other. Chesky foresees the Airbnb app itself transforming into an agent, with dedicated agents for exploration, customer service, and eventually a 'macro Airbnb agent' that is interoperable with other agents via standards like MCP.

A significant challenge, according to Chesky, is the industry's lack of a true operating system built for AI. He argues that current AI applications are developed on conventional operating systems like iOS, macOS, and Windows, which are not 'AI operating systems.' In an ideal scenario, AI would operate at the kernel level, allowing all components, including agents, to seamlessly interact. The current race to become the 'quarterback,' or primary AI agent, is flawed, as no one is developing a full OS SDK. He cites ChatGPT's failed app store attempt as an example, attributing its struggle to the absence of a foundational operating system and SDK comparable to the iPhone's App Store.

Chesky's personal experience with Airbnb on third-party agents like Muse and Instinct has been suboptimal, leading him to conclude that consumer AI has not yet been 'cracked.' For consumer AI to truly succeed, he advocates for development at a lower substrate, where all apps become interoperable agents with rich user interfaces. He identifies two common misconceptions in Silicon Valley: that interfaces will disappear in favor of pure speech/typing, and that software will be entirely generative. Chesky contends that human designers can prompt better software than consumers, and that as long as Apple and Google control major platforms, apps will remain central. A fundamental shift from apps to agents, he believes, will necessitate a new platform built by entities like Apple or Google.

Looking ahead, Chesky foresees a growing role for voice technology. He believes people will eventually speak to computers more than they type, and voice will be crucial for Airbnb's search and customer service functions. The company plans to roll out voice agents this fall to assist users. Internally, Airbnb has already realized substantial benefits from AI, accelerating feature development. On a personal level, AI has streamlined Chesky's access to information, reducing the need for numerous meetings by allowing him to directly query his AI assistant.