Airbnb CEO on Why AI Will Create a New Era of Consumer Products
TL;DR
Airbnb says it reignited growth by acting like a startup again — Brian Chesky said revenue growth accelerated from 10% last year to 18% this quarter after small, high-intensity teams like “Project Hawaii” obsessed over conversion and the guest journey.
Chesky’s management philosophy is blunt: leadership is presence, not absence — he argued big-company CEOs shouldn’t “renovate the whole house,” but fix one room at a time by getting deeply involved in details, skipping layers, and teaching teams startup-level pace.
AI is already materially changing Airbnb’s operations — Chesky said 60% of Airbnb’s code is now written by AI, customer service cost per ticket is down 10%, and 40% of people contacting support have their issue solved by AI.
He thinks chatbots are the wrong interface for travel and commerce — Chesky’s case is that text-first UX breaks on photos, filters, comparison shopping, and collaboration, and that the future is agentic but far more visual, immersive, and interactive than today’s chat apps.
Consumer AI is still wide open because too many startups are chasing enterprise — citing a recent YC batch where only 16 of 175 companies were consumer-focused, Chesky argued most labs are staffed with AI researchers, not designers or product people, so the real consumer breakthrough hasn’t happened yet.
His advice to founders and designers is basically the same: stop following the crowd — whether it’s the 10th agent startup or creatives sitting out AI, Chesky’s point was that the winners will be people who claim an uncrowded space and build something weird enough that consumers can’t ignore it.
Summary
Airbnb’s growth snapped back — and Chesky says that’s hard to fake
Chesky opened with a strong business update: Airbnb grew revenue 10% last year, then 18% this quarter, which he framed as a real acceleration for a marketplace doing roughly $100 billion in gross bookings annually. His point was that once a marketplace of that size starts slowing, “gravity” makes it brutally hard to bend the curve back up.
The “Project Hawaii” playbook: fix one room, not the whole house
The turnaround started with a small internal team focused narrowly on conversion and the guest journey, working with the intensity of Airbnb’s early “Rouse Street” days. Chesky said the lesson for leaders is not to transform the whole company at once, but to make one part excellent, then move room by room until the muscle memory spreads.
“Leadership is presence, not absence”
Asked how that didn’t turn political, Chesky said he didn’t replace teams — he got deeply involved with them, reviewing work weekly or even daily, removing bureaucracy, and modeling a higher bar. Some people hated it and left, but his core argument was memorable: a CEO should be “on the battlefield,” not somewhere remote “writing blueprints,” and AI will only make hands-on leadership more necessary.
Chesky’s AI numbers are already pretty concrete
He said 60% of Airbnb’s code is now written by AI, which he claimed is about double peers and competitors, and that AI is also helping customer service in measurable ways. The cost per support ticket is down 10%, and 40% of inbound support cases now get resolved by AI before needing a human.
Why he thinks chatbots won’t own travel
Chesky pushed back on the idea that travel just becomes a ChatGPT wrapper, saying text-first interfaces are bad for photos, direct manipulation, comparison shopping, and collaboration. His analogy was sharp: asking a chatbot to handle all commerce is like trying to use iMessage for everything when every real app has its own richer interface.
The app is only 20% of Airbnb
He also made the strategic defense of Airbnb against AI disruption: the visible guest app is only about 20% of the company. The harder part is everything behind it — payments, support, adjudication, host tooling, guarantees up to $3 million for theft or property damage, and operating a network with 5.5 million hosts and 4–5 million guests staying nightly in 100-plus countries.
Consumer AI hasn’t had its breakout because everyone is chasing enterprise
Chesky said nearly every serious AI company is still enterprise-first, pointing to a YC batch with 175 companies and only 16 consumer startups. His explanation: too many new labs are full of pure AI researchers without product or design instincts, so they’re all attacking coding, science, and similar infrastructure problems while the real consumer interface revolution remains unbuilt.
Designers should run toward AI, not away from it
One of the strongest sections was his warning that designers and creatives may repeat the early-web mistake of treating a new medium as low-status. If they sit this out, he said, engineers and PMs will design the future for them; if they lean in, photo/video generation and vibe coding could make this the biggest opportunity for visual and interaction design in his lifetime.
Marketing still works — but only if it feels a little unhinged
On growth, Chesky said old channels don’t disappear, they just stop being magical once everyone copies them. His best example was Airbnb turning a Malibu house into the Barbie Dreamhouse: the internet talked about it, and the ROI beat any traditional ad campaign, reinforcing his broader belief that the best marketing is being unmistakably different.
Was This Useful?
Share
Keep Reading
Make Alcreon Yours
Tune your feedFive quick questions, and the feed ranks what matters to you first.Or just get notified
The weekly Echo. Signal worth keeping in your inbox.
Every new piece, announced on X.
Read Next
See all
Playbook
The Retirement Email Isn't a Warning
Model retirements now arrive every few weeks; the config-eval-rehearsal loop turns each deprecation email from a fire drill into an afternoon swap.

Playbook
The Cheapest Model That Passes
OpenRouter lists 400 models behind one API. The fix for choosing isn't a better leaderboard, it's a four-step protocol that ends in a real eval.

Playbook
Cheap Models, Hard Tasks
Most agent workflows route every step to the frontier model by default. The bill scales with how chatty the agent gets, even when most steps don't need that brain.