Airbnb now spends more than 17 percent less on support for each booking than it did a year ago, and CEO Brian Chesky gave most of the credit to the company’s AI assistant. CX Dive reported the figure from Chesky’s remarks on the Q3 2026 earnings call. It is a hard cost number in a category that usually trades on demos.
The assistant resolved about 40 percent of third-quarter support issues without a human agent, according to the company. Airbnb launched it a year and a half ago, so the savings so far come from a system that closes a minority of cases end to end.
Phone support is next. In the U.S., guests who call Airbnb for help will now be answered by the assistant, which until now handled text. Chesky added that customer service NPS is as high as it has ever been. CX Dive’s report does not include the score, and it does not say how Airbnb defines a resolved issue.
The bigger announcement is a product, not a metric. Chesky described Airbnb as an “AI-native” company and said it will launch an AI travel agent in the first half of 2027. The agent will cover every stage of a trip and will not be confined to Airbnb’s own platform.
Chesky made the case against general-purpose assistants directly. Meta’s Muse and Instinct, from Spear Street Technology, are good at routine chores such as booking a restaurant table or scheduling an appointment, he said. Planning a trip is a different job. “Travel is aspirational,” Chesky said, and choosing where to go and stay is part of the enjoyment, so he does not expect one agent to handle everything.
He listed three advantages for Airbnb. The company has 20 years of experience with its marketplace. It already sits inside the reservation, payment, and support flow, a position a third-party agent would struggle to copy. And it is building an interface made for travel rather than a standard chat window, one that assembles maps, photos, calendars, and itineraries on the fly from a library of hundreds of travel-specific components.
The support results bear on that pitch more than the quarterly cost line suggests. Of the three advantages, the second is the one Airbnb has now demonstrated: its assistant works inside booking and customer-service workflows at real volume. The first and third, deep marketplace knowledge and a purpose-built interface, remain claims about a product that has not shipped. Support is a problem with a clear right answer. Trip discovery is not, and nothing in Chesky’s remarks offers a way to measure whether the agent does it well.
Chesky’s own framing sets the test. If discovery is the experience, an agent that answers quickly but recommends blandly would fail the job he says only a travel-specific design can do. Airbnb has also not said what the agent will cost to run or how its savings on support will translate to the booking funnel.
Travel operators benchmarking their own support automation now have two Airbnb numbers to beat: 17 percent lower cost per booking and 40 percent of issues closed without a person. Competitors have until the first half of 2027 to answer on discovery, when Airbnb’s agent is due.
Reported by CX Dive (Customer Experience Dive) on November 10, 2026.
