Airbnb CEO: AI Agents Need Their Own Operating System

Chesky said he has not changed his view that a chatbot-only interface is not right for e-commerce or travel. Chatbots are poor for browsing or shopping because they give users a few options at a time, and users have to take multiple turns before reaching a result, he said. But he also said Airbnb’s new search is not the endgame. He described the eventual interface as something between a chatbot and what the company shipped in the first version. “I think what you’re seeing today is also not the endgame for e-commerce or for travel or shopping,” he said.

For travel, Chesky said, speed and ease matter for some tasks, such as booking a flight. But Airbnb users often enjoy looking at the product, and travel studies show that people derive more pleasure from planning and anticipating a trip than from traveling itself, he said. That means the company does not want to take away inspiration, browsing and the experience of dreaming.

Chesky identified a second problem with chatbots: they are built for one person and are not a collaborative interface. That is a poor fit for travel, which people often use with family and friends. Over the next three to six months, Airbnb’s task is to explore interfaces that enable “multiplayer” AI, meaning AI that several people, such as a group planning a trip together, can use at once.

Airbnb is not the only company rethinking what an AI interface should look like. Startups such as Monogram and, more recently, Wabi are experimenting with “generative” interfaces, which are screens an AI builds on the fly rather than ones a designer fixed in advance. Chesky said he thinks people will see a mix of deterministic, pre-designed interfaces and generative ones.

Asked how Airbnb is preparing for consumer AI agents, tools like Instinct and Muse that carry out tasks on a person’s behalf, Chesky said agents will need robust software infrastructure on Airbnb’s end. The company wants to make itself more agent-friendly, and he said agents could be good lead generators. “I don’t necessarily know if I believe in a world that there’s apps in the future, but I don’t believe in a world where the apps become data layers of a universal simple UI,” he said. He added that when users open an iPhone, there is a reason Airbnb’s app is different in design from DoorDash, Shopify and Amazon. Interfaces should adapt to the utility and function of each service, he said. Airbnb wants to make its app more agentic and build more user interface controls.

In Chesky’s description, a user might have an agent while also wanting strong browse functionality, the ability to message a host, compare options, verify identity, use a map feature and add other items. If the user wants to stay on the platform, either the agent has to hand off or the agent application would need to build a more robust software developer kit to bring the app to the user, he said. Chesky said concerns had existed that AI tools could cut companies like Airbnb out of the customer relationship, but chatbots have proved good at lead generation.

He is also planning for agent-to-agent interactions through the platform. Some interactions might take place within Airbnb, and some might involve agents from other apps, he said. His thesis is that apps have not always been interoperable unless companies struck a business deal. Agents, by contrast, are like “apps with faces,” and when they talk to each other, all apps can become interoperable. Chesky said he thinks of the Airbnb app as one day becoming an agent, with an agent in the explore tab, an agent in customer service and another agent under development in a different service area. Eventually, there will probably be a macro Airbnb agent that is interoperable with MCP, a standard that lets AI agents connect with other software, and with other agents.

Chesky argued that the bigger problem is that the industry lacks a true operating system built for AI. Right now, people are building AI apps for iOS, macOS and Windows, he said.