Airbnb just gave a rare, specific number for something most companies only talk about in vague terms. The time it takes to go from an idea to a live product feature has dropped by 60 percent. The number of features and fixes the company shipped this year is up nearly 80 percent compared to last year. CEO Brian Chesky said the reason is that AI now writes close to 60 percent of the code Airbnb's engineers produce.
That number is worth putting next to what other companies report. Microsoft and Google have said AI writes roughly 30 percent of their new code. Meta's CEO has said AI could handle half of the company's software work within a year. Anthropic, a company that builds AI itself, claims figures well above 60 percent. Airbnb sits closer to the AI-native companies than to its own tech industry peers, which is notable for a company that sells vacation rentals, not software.
Chesky also said something blunter: managers who only manage people, without doing hands-on work themselves, no longer have a place at the company. He described seeing his own design and engineering managers go back to writing code. He did not say whether this means fewer manager jobs going forward, only that it is too early to say. That is a warning worth sitting with even outside tech. If a company can build products faster with fewer layers of coordination, the case for keeping purely administrative management roles gets weaker everywhere, not just in software teams.
On the customer-facing side, Airbnb is moving carefully. The company is testing a toggle that lets users switch between the old search filters and a new AI search that accepts plain language questions and returns visual answers instead of a list. This caution is deliberate. Chesky has said in the past that travel planning does not work well as a simple chatbot conversation, and Airbnb still has not proven that natural language search actually converts to more bookings.
The support numbers are the most concrete win so far. Airbnb's AI assistant now runs in more than 50 languages, and nearly 45 percent of support conversations that start with the AI end without any human involved. Cost per booking for support has fallen 16 percent because of it. Expedia reports a similar pattern, with more than 30 percent of its service interactions now AI-powered and that share still growing. Booking.com is using AI to route customers to the right channel faster. Every major travel platform is converging on the same playbook: let AI handle the routine questions, keep humans for the messy ones, and count the savings.
The lesson for any business watching from outside travel is not about chatbots. It is about the order of operations. Airbnb did not start with a flashy AI feature for customers. It started by rebuilding how its own teams work, then moved to support, and only now is it cautiously testing something customers will notice. Companies that skip straight to a customer-facing AI gimmick without fixing the plumbing first are usually the ones that end up walking it back.