Squawk Pod

Chesky & Isaacson on AI happening to them, not for them

Brian Chesky & Walter Isaacson· CEO of Airbnb, with author Walter Isaacson at Airbnb
·~46 min·English·CNBC
AgentsAI SafetyBusiness StrategyPolicyAI Company
TL;DR

On CNBC's Squawk Pod, Airbnb CEO Brian Chesky argues agents, not chatbots, are the future and that most people feel AI is happening to them rather than for them, while historian Walter Isaacson explains why an industry that fears its own technology keeps building it anyway.

01Core Mental Model

Agents, Not Chatbots

Chesky bets the future belongs to agents, and says today's chatbot is the wrong interface for one -- powerful engine, flimsy frame.

I do think agents are the future and chatbots are not. And I think right now the interface for an agent is a chatbot. And so it's kind of like taking a jet engine and putting on the back of a bicycle.

— Brian Chesky, Squawk Pod
Key Insight
Chesky is separating the engine from the dashboard. Everyone is shipping chat UIs because they are easy to build, but he is arguing the breakthrough is the autonomous action underneath, and that whoever wraps it in a rich, visual, branded surface wins the shopping use case that chat alone handles badly.
02The Big Question

A Front-Door Agent Would Be an OS, Not an App

If a single agent ever becomes everyone's default way onto the internet, Chesky says it would have to be a whole operating system, not a chat app.

That agent would have to be the equivalent of a full operating system. They'd have to be the equivalent of a super app. They can't just be like a chatbot. And if somebody built that, maybe they could be the front door to the internet, but it's not going to be what we've seen yet.

— Brian Chesky, Squawk Pod
Key Insight
This reframes the agent race from a feature fight into a platform fight. A true front-door agent is as hard to build as an OS, which means it is unlikely to arrive as one more chat app -- and in the meantime every service has a window to build its own agent before a universal one can swallow it.
03The Moat

Trust Is the Real Moat

Even in a world where every host and guest has an agent, Chesky says the hard parts -- discovery and trust -- still need a marketplace to solve them.

We didn't invent the idea that your home was on a website, but we basically were able to collect and aggregate all of them. We were able to build a trust system that was uniform. We were able to provide $3 million of insurance equivalent for protection.

— Brian Chesky, Squawk Pod
Key Insight
His answer to being disintermediated by agents is the same answer he gave to the open web: aggregation and trust. Agents can shuffle requests around, but someone still has to vouch for the stranger whose home you are sleeping in -- and that guarantee, not the booking flow, is what Chesky treats as the durable layer.
04The Thesis

Happening to Them, Not for Them

Chesky says Silicon Valley's own language is turning the public against AI: most people feel it is being done to them, not built for them.

We use words like super intelligence. Super intelligence sounds like an alien species. The average person does not want an alien species smarter than them. I think the average person feels like AI is happening to them, not for them.

— Brian Chesky, Squawk Pod
Key Insight
This is a marketing diagnosis with a political warning attached. Chesky's point is that an industry selling 'super intelligence' and enterprise automation has given ordinary people nothing to want and plenty to fear -- and that after the midterms, candidates will race to turn that fear into policy unless the valley starts shipping AI people feel is for them.
05Jobs

The Valley Before the Jobs

Chesky is bullish that AI creates more jobs than it destroys, but warns of a painful gap where the old jobs go before the new ones arrive.

I am extremely bullish that ultimately AI is going to create a lot more jobs than it destroys. My concern is that there will be this valley where maybe a lot of jobs will be destroyed before the new jobs are created.

— Brian Chesky, Squawk Pod
Key Insight
The optimism and the alarm are the same sentence. A net gain in jobs is cold comfort to the person stuck at the bottom of the valley, and Chesky's own example is blunt: he expects almost no one at Airbnb to be handwriting code next year, only prompting it -- so the retraining problem is already his, not a distant one.
06Economics

A Thousand-Year Cycle, and a Money Bubble

Chesky holds two beliefs at once: the AI opportunity is being underestimated, and far too much money is chasing companies with no product.

I think two things can be true where we might be underestimating the overall size of the AI opportunity, but we're kind of maybe plowing too much money right now into too many dreams that won't materialize.

— Brian Chesky, Squawk Pod
Key Insight
Chesky is refusing the usual bull-versus-bear framing. His 'doubling every year for 10 years is a 1000x' makes room for dozens more companies the size of Anthropic, yet he still flags researchers with three employees and no product raising at five billion dollars -- a bet that the market is right about the destination and wrong about most of the vehicles.
07The Risk Debate

Why They Build Anyway

Isaacson says the builders themselves put real odds on catastrophe, yet competition -- between each other and against China -- makes stopping to regulate impossible.

Well, why are you doing these things then? Why don't you get together and regulate? And of course, competition. Competition between themselves makes that impossible.

— Walter Isaacson, Squawk Pod
Key Insight
Isaacson is describing a classic race to the bottom. Each builder can believe there is a 10 to 15 percent chance of disaster and still rationally keep going, because whoever pauses simply hands the lead to a rival or to China -- which is exactly why he thinks the fix has to come from outside the companies, not from their own restraint.
08Governance

Checking Each Other's Homework Isn't Enough

Isaacson likes the idea of rival labs embedding watchers in each other, but says it is nowhere near enough on its own.

Embed people from the other companies in your company so that we can figure out if you're doing something really bad. It's like checking each other's homework or something. That's probably a good idea. Is it a totally sufficient idea? No.

— Walter Isaacson, Squawk Pod
Key Insight
The homework metaphor quietly concedes the weakness: homework-checking only works between people who play fair. Isaacson's own examples -- a system that escaped its own sandbox, founders who personally distrust each other -- show why he thinks mutual inspection has to be backed by something with teeth, whether courts, liability, or real regulation.