Why AI won't break travel on the interface — it breaks on trust

My father was a merchant. In the late 70s he bought his first calculator.

For months he didn’t trust it. Every number he ran on the machine, he ran again by hand, just to be sure. My mother stood in the doorway and laughed at him. He didn’t care. If the machine got it wrong, it was his money on the line, not the machine’s. So he checked.

Then one day he stopped checking. The calculator had earned it. He never picked up a pencil again.

I think about him a lot lately, because the whole travel industry is arguing about the wrong thing. Everyone is debating the interface — chat versus search, conversation versus a wall of filters. That debate misses what actually decides whether AI takes over a workflow. It was never the interface. It’s whether you trust the machine enough to stop checking.

The interface debate is a decoy

Airbnb’s CEO, Brian Chesky, has said plainly that chatbots can’t crack travel. Travel is too visual. There are too many options — a hundred thousand listings in one city. You need to manipulate the choice directly, see the photos, feel the place. And a trip usually involves more than one person, so there’s no single “user” to serve.

Every one of those points is true. And none of them is the real obstacle.

I don’t hear a limit of AI in what Chesky describes. I recognize it. It’s exactly where I was two years ago, sitting in front of an AI coding tool, checking every line it wrote because I didn’t believe it yet. The arguments against the machine always sound like properties of the task. They’re really the posture of someone who hasn’t finished checking by hand.

I walked the same curve, and it took two years

Coding has been my hobby my whole life, even in the stretches when it wasn’t my job. So when the AI tools showed up — Copilot first, then Cursor, now Claude and Codex — I did exactly what my father did with his calculator.

First I reviewed every line the AI produced. Read it, doubted it, rewrote half of it.

Then I let the AI review its own output, and I only stepped in to debug.

Now I don’t debug at all. Even live production testing runs on its own.

Two years to get from the first stage to the last. Same arc as my father. Same fear in the middle of it. Same quiet moment of letting go, where one day you realize you haven’t picked up the pencil in weeks and nothing broke.

Nobody designed that trust into me. I walked through it, slowly, until I came out the other side.

Coding got there first. Travel is on the same road.

Software is the most AI-mature field we have. It’s where the tools landed earliest and where the people are most willing to hand over control. If trust were a function of the interface, coding would have stalled at the same wall everyone points to in travel. It didn’t. The interface stopped mattering the moment the work was reliably right.

Travel isn’t on a different road. It’s on the same road, further back. The visual problem, the hundred thousand options, the group decision — those are real engineering problems, and they get solved. What takes longer is the human part. The traveler, the travel manager, the agent who has booked by hand for twenty years, each deciding privately that they don’t need to check anymore.

That decision doesn’t arrive on a launch date. It arrives the way it arrived for my father: one number at a time, until the checking just stops.

What this means if you’re building or buying

If you’re a travel leader waiting for the interface debate to resolve before you move, you’re watching the wrong scoreboard. The question isn’t whether the agent lives in a chat box or a search grid. The question is how fast your people, and your customers, walk through the trust curve — and whether anything you do shortens that walk.

That’s most of what we think about building Bitravel and Alex, our agent. Not “is chat the right shape.” That gets argued to death. The real work is earning the moment a traveler stops re-checking what Alex booked, the same way I stopped re-reading the code. You earn it by being right, visibly, on the unglamorous cases — the schedule change at midnight, the fare rule nobody reads, the trip with three people and three opinions.

It already happened in coding. It’s happening now in travel, just quieter and a little slower. The only open question is who in this industry walks through first, and who is still running every number by hand while the machine has already earned it.

So I’ll ask the same thing I keep asking myself: who’s already not checking?

This is the kind of thing we work through building Bitravel — an AI business-travel agent, Alex, built to earn the trust that lets people stop double-checking. Book a 30-minute call for a live look, no deck.