How Brian Chesky Is Redesigning Airbnb for the AI Era

ELI5/TLDR

Brian Chesky says the lesson of his last six years is that founders shouldn’t hand the steering wheel to professional managers — they should keep it, learn the CEO job in the details, and only let go once they know what good looks like. He thinks AI is going to make that even more intense: fewer management layers, no pure people-managers, and every leader expected to actually do the work (read the code, read the case law). For Airbnb specifically, he’s trying to shift the company from being a noun for a house to being a profile for a person — homes plus services plus experiences plus eventually flights, all hung off an authenticated identity. He has not done it yet. The stock has been flat because the second hit hasn’t landed.

The Full Story

The pandemic as a personal reset

Chesky’s origin story for “founder mode” is now familiar but he tells it more honestly here than in the Paul Graham essay. By 2019 Airbnb had ~7,000 people and he had stopped recognising his own company:

I had like 7,000 employees. I didn’t know what anyone was doing. I felt like I was in a car without a steering wheel… I had a dream where I felt like I’d left the company for 10 years and I’d come back and somebody had been running the company for 10 years and they kind of turned it into this giant political bureaucracy and I didn’t even recognize the company. And then I realized, oh my god, it was me the whole time.

The pandemic — losing 80% of revenue in 8 weeks — gave him cover to take everything back. He hired Hiroki Asai from Apple, an “unsung” Steve Jobs creative director, and tried to copy how Jobs ran Apple after 1997. The first attempt at micromanaging produced revolts. He kept doing it anyway.

The frame he settles on: start hands-on, hand off grudgingly. Most founders do the opposite — delegate first, intervene later, by which point the report has built the wrong muscle memory. His golf-instructor analogy: a teacher has to watch your swing thousands of times before they can stop watching.

What he thinks “AI founder mode” is

Chesky is explicit that he doesn’t have it figured out. The principles he’s working from:

Fewer layers. He likes the Catholic-Church-has-four-layers line. Doesn’t think 7,000 people can be managed flat, but is targeting a meaningful flattening.
No pure people managers. Anyone running people also has to do the work. Engineers code. Lawyers read case law. Designers design.

I do not think people who have lots of recurring one-on-ones are going to survive. What they’re doing is — oh, you come to me with whatever your problem is, I’m here to help you, like a mentor or professor. That kind of leadership style is not going to work.

Async over meetings. His old founder mode was meeting-heavy because that was the only way to get information; with AI he expects a shift to async, which suits Airbnb’s remote setup.
The two casualties of the AI transition: pure people managers, and people who don’t want to evolve.

He thinks tools will get easy because there’s economic pressure for them to. Right now, claims Chesky, AI is “almost entirely an enterprise phenomenon” — he cites a YC batch where 159 of 175 companies were enterprise — and his bet is that the next 12-24 months brings a consumer AI renaissance. His diagnosis of why consumer AI hasn’t happened yet: founders feared ChatGPT would kill their startup, no proven business model (subs hit a ceiling because Claude/Gemini are free, ads are blocked by competitors not running ads, e-commerce is hard), and Silicon Valley is more vibe-driven than it admits.

Project Hawaii, and why Airbnb only has one hit

The most concrete operating idea in the conversation. Chesky calls it Project Hawaii: take a giant company and run small elite “Navy SEAL” teams — ~10–12 people, designers/engineers/PMs/data scientists — against one problem at a time. The first one tackled guest-side conversion (the search-to-book funnel). The framework:

Crawl, walk, run, then fly. Crawl: fix the bugs. Walk: develop features, reframe the journey. Run: rethink the entire flow. Fly: completely reinvent.

The numbers he cites: ~$200M in incremental year-one internal revenue, ~$400-500M year two, now running at “over 600 basis points” of $134B in gross sales. Take that with the usual caveats — these are internal attributions on a shared funnel — but the order of magnitude is real, and the operating model has now been copied across pricing, services, experiences.

Underneath Hawaii is a confession. Airbnb has been a one-hit wonder for 18 years. He thinks he finally understands why: Airbnb launched in one city (NYC). Uber in SF. DoorDash in Palo Alto. But when Airbnb launched Services last year, they launched in 100 cities at once — and it didn’t work. The new rule: one to ten to many. Pilot in one market, get it working, expand to ten, then industrialise. He says they now have 10–20 pilots running and want to eventually be running 50–70 verticals.

The first principle, borrowed from Peter Thiel, is to dominate a tiny market rather than nibble a big one — and from Paul Buchheit’s Gmail story: it’s better to have 100 people love you than a million people sort of like you. The metaphor he keeps coming back to: heat a bathtub, not an ocean.

The Apple lessons, after Steve Jobs

Hiroki Asai taught Chesky two principles he says he imported wholesale:

Simplicity is distillation, not deletion. Startups are simple by accident — no money, no people, no choice. Once you raise capital and hire, you fragment. Holding simplicity becomes a discipline.

Simplicity is not removing things. Simplicity is distilling something so fundamentally that you understand its essence.

How you do anything is how you do everything. From Bill Walsh (“the score takes care of itself”) and John Wooden — the famous first-day-of-practice lesson where Wooden spends an hour teaching freshmen how to put their socks on. The point is to stop optimising the scoreboard and obsess over inputs. Chesky says Airbnb formally stopped focusing on growth and focused on getting every input perfect. The implicit claim is that growth followed.

The 11-star exercise gets a clean retelling. Five stars on Airbnb means nothing went wrong (review compression). To find a real six-star experience, you imagine ten or eleven and back into it — the eleven-star check-in is Elon Musk meeting you at the airport and taking you to space; the nine-star is a 1964 Beatles arrival. The exercise is “going beyond the edge of reality and working backwards” — most people can’t even imagine a six-star until they’ve imagined the absurd ones.

What Chesky is actually trying to build now

Three concrete projects at the bedrock level:

Atomic unit shift: home → person. Airbnb is currently a noun for a house. He wants the unit to be the user — an authenticated identity, a rich preference library, a real-world social graph, eventually a membership program. He’s explicit that Facebook abandoned this and there’s a vacuum. “Proof of personhood is going to be really, really important in an age of AI.”
From three things to fifty. Industrialise the new-vertical pipeline so Airbnb is to travel/lifestyle what Amazon was to books-then-everything. Homes, experiences, services — eventually flights and beyond.
Self-disrupt the app before someone else does. He concedes the innovator’s-dilemma trap: $100B/year flows through Airbnb’s app, hosts depend on it, public-company guidance discipline kills experimentation. So he’s running “little sandboxes” — possibly separate apps — to ask what a radically different Airbnb looks like.

He thinks apps are dead in the long run. “I don’t think there will be apps in the future. I think there’ll be agents.” Which is a pretty striking thing for the CEO of an app-first company to say on a podcast.

What endures

Patrick asks the obvious follow-up: if AI is making creation cheap and ephemeral, what’s left to build that lasts? Chesky’s answer is interesting because it’s tonally not a CEO answer:

Software is extremely ephemeral. Hardware has medium endurance. Environments and physical worlds have huge endurance… No matter how good our app is now, in 10 years I’ll look at it and think it looks like crap.

So what does he think will last for Airbnb? Not the software. The community, the brand, the principles, the logo, the voice, the mission. He’s been telling the company they’re building a community, not an app. Hard to know whether this is real conviction or a way to make peace with the fact that the thing he obsesses over (the app) won’t matter in a decade.

Hiring as the actual job

The most operationally useful section of the interview. Chesky’s claim: every CEO chooses between time on hiring and time on managing. They trade off one-for-one. He didn’t believe this when Sam Altman told him in 2008. He believes it now.

His hiring system, in his words:

Don’t do searches. Build pipelines. A search firm gives you 50 profiles, you whittle to a hire, you discover 12 months later they’re bad. Wrong shape.
Start with results, work backwards to people. Don’t hire someone from Nike because Nike has good marketing. Find an ad you like, find who made it.
Build referral mafias. Every meeting is to get the next meeting. Ask every good person who else is good. Map the design mafia at Apple, the ops mafia at Uber.
Be co-hiring-manager for the top 200, not just executives. “My executives should hire people so good they would never work for them without my help.”
He spends 2-3 hours a day on it. First and last call of the day is to the recruiting team.

The motivation reset

The most personal arc: Chesky says he confused success with love. The IPO made him miserable the next morning because the agilation drug had stopped working.

Agilation is like a cup with a hole at the bottom. You keep filling it in thinking it’s love, except it just keeps coming out at the bottom.

He cites Rick Rubin: an artist is an artist when they make it for themselves. The pivot was to stop chasing being someone (CEO of a hot company) and pursue making something. Obama allegedly told him: “If I focused on who I want to be — president — and I wasn’t, my life would have been a failure. I focused on what I wanted to do — help people — so I didn’t need to be president.”

His four heroes — da Vinci, Van Gogh, Walt Disney, Steve Jobs — all worked until the day they died. He thinks of himself as a designer who stumbled into one of the largest design canvases in history.

Key Takeaways

Crawl, walk, run, fly. Sequence for any new product or function — fix what’s broken before adding features, reframe before rethinking, rethink before reinventing.
One to ten to many. Pilot in one market, scale to ten, then industrialise. Airbnb’s services launch failed at 100-city scope; he attributes this to violating the rule.
Hawaii teams. ~10-12 person elite squads, treated like internal startups, owned by a single problem. Reportedly drove ~600 bps of Airbnb’s $134B in GMV.
The bedrock shift. Airbnb’s atomic unit moves from home → person. The bet is on identity + preference graph + membership, not on listings.
Hiring/management trade-off is one-to-one. Time on recruiting subtracts directly from time on managing. He spends 2-3 hours/day on it. First and last call of the day.
Pipeline > search. Don’t hire when you need to. Continuously meet people, ask everyone who else is good, build referral mafias.
Results-first hiring. Start from work you admire, find who made it, then approach. Don’t start from resumes.
Co-hiring-manager for the top 200. Founders shouldn’t only hire their direct reports.
No pure people managers in the AI era. If you only do one-on-ones and manage emotions, your job is gone. Every leader has to also do the craft.
Fewer management layers. Catholic Church does it in 4. Most companies have 7-9. He wants Airbnb closer to 4.
The 11-star exercise. Imagine experiences beyond reality (Beatles arrival, ride to space) to make six-star experiences feel obviously achievable.
Sandbox apps. Public-company discipline kills experimentation, so he’s running separate-app experiments asking what an AI-native Airbnb looks like.
“There won’t be apps, there’ll be agents.” Notable concession from the CEO of an app-first business.
Apps don’t endure; communities, brands, principles do. What he’s optimising for survival is the network and the values, not the interface.

Claude’s Take

Patrick O’Shaughnessy interviews like a fan. There are no challenge questions, no “show me the data,” no pushback when Chesky claims his 10-person Hawaii team drives 600 bps of $134B GMV (that’s $8B+ in attributed revenue from one team — possible, but the kind of claim that deserves at least one follow-up). The conversation is shaped to flatter, and Chesky knows how to play it — the “I had a dream and the bureaucrat was me the whole time” beat is well-rehearsed.

That said, the operating ideas hold up if you separate them from the founder-mode mythology. Three of them are worth taking seriously:

The hiring-vs-management trade-off is the most actionable claim in the episode and almost certainly correct. The mistake most people running anything past 30 employees make is treating recruiting as something they delegate. Chesky’s “first and last call of the day is the recruiter” is the kind of thing that actually shows up in calendars, and you can audit whether someone’s doing it.
One to ten to many is a useful corrective to the typical Silicon Valley instinct to launch globally. Airbnb’s own services failure is the proof case he points to.
Hawaii teams is plausible because it’s just a renaming of the small-strike-team pattern that actually works — what kills it usually isn’t the model but the org’s antibodies. Chesky’s claim is that founder-mode authority is what keeps the antibodies away. That’s testable, kind of.

The vision stuff is much weaker. “Atomic unit shifts from home to person” sounds great in a podcast but is the kind of thing every consumer company has been trying for 15 years. Facebook tried, Google tried, Amazon tried; the actual asset would be hundreds of millions of preference profiles people care enough to maintain, and no one has shown that’s earnable for a brand whose users currently engage with it 1-3 times a year. Services has been the public answer to “where’s the second hit?” for two years now. The earnings call language (“project Hawaii,” “600 bps”) is shareholder-deck language, not operating reality.

The “no apps, only agents” line is the most interesting throwaway, because if Chesky believes it, the whole identity-graph play makes more sense — Airbnb wants to be the verified-identity layer that an agent talks to on your behalf. That’s a much sharper bet than “we’re going to add 50 verticals.” The conversation never gets there because Patrick doesn’t push.

Net: useful for the management craft (hiring system, sequence rules, simplicity discipline), light on substance for what Airbnb actually becomes. The stock is flat for a reason — the second hit hasn’t shipped — and Chesky says as much. 7/10. Reach for the operating ideas; discount the vision.