B2A: the agent distribution channel most travel brands aren't in yet

Your next customer probably won’t open your website. It will query your API. If that sentence sounds like a slide from some futurist deck, fair enough. But I’ve watched this shift start happening inside real workflows, and it’s not a forecast anymore.

For my whole career travel distribution had two channels. B2B, selling to the trade. B2C, selling to the traveler direct. Both were built around a human doing the buying — someone clicking, comparing, deciding. A third channel is now forming, fast, and most travel brands don’t realize they’re already missing from it.

I call it B2A. Business to agent. The buyer is an AI.

What B2A actually looks like

Picture a corporate traveler typing one line: “Singapore Thursday, stopover in London for meetings, New York by Friday, preferred carriers, keep it within policy.”

A human travel arranger would take that and spend forty minutes on it. Multi-city routing. Cabin rules per leg. Carrier compliance. Fare logic. Policy validation. Back and forth on the options.

An agent resolves the whole thing in seconds. No arranger in the loop. The traveler never opens six tabs.

That’s the part people miss when they hear “AI in travel” and picture a chatbot bolted onto a booking page. This isn’t a nicer front end. The thing doing the searching, comparing and deciding is no longer a person. The customer changed shape. And a customer that reads APIs does not behave like one that reads marketing pages.

Who actually wins from this

The value here is not theoretical, and it’s not evenly split.

For the traveler: instant amendments, availability at 3am, a system that already knows your seat preference and your frequent flyer numbers, that compares hundreds of options at once and rebooks you before a delay becomes your problem.

For the TMC — and this is the one I care about most, because it’s who we build for — the economics change at the root. Cost stops tracking headcount. A team handling a thousand bookings can handle a hundred thousand on the same infrastructure. No overtime in peak season. No manual entry errors at 2am. Quality stops depending on which agent caught the ticket.

For distribution itself: API-direct relationships, cleaner demand signals, fewer manual exceptions, lower cost at every layer.

That’s the upside. Here’s the catch nobody wants to hear.

B2A only works if your data is ready

I spent years processing supplier content at Tourico and later at Hotelbeds. Billions of requests a day, hundreds of thousands of price lines in a single response. You learn quickly what separates the suppliers that show up in results from the ones that quietly don’t.

Context tagging. Look-to-book ratios. Response time. Caching. Price accuracy. Outage resilience. Boring words. They decide everything.

A human buyer forgives you. Stale price, slow page, a fare that’s off by a few dollars — they sigh and keep going, because they already came to your site. The machine does not forgive. It pulls fresh data, checks it, and if yours is slow or wrong it moves on to the next supplier in the same millisecond. There’s no second chance, no brand loyalty, no “let me refresh.” You were either in the result or you weren’t.

This is the part that flips the usual conversation. Most travel brands are still investing in how they look to a human. B2A doesn’t care how you look. It cares whether your data is clean, fast and current enough to be trusted by something that compares you against everyone else, instantly, with no sentiment.

This is already in production

Tripadvisor is live inside Claude, Alexa+, OpenAI and Perplexity. Skyscanner launched with OpenAI’s Operator. These aren’t experiments tucked away in an innovation lab. They’re the same early bet: be present where the agent looks, before the agent stops looking for you.

And it’s happening at the company level too. One company we work with put their travel agency behind Bitravel’s agentic layer. Their people already live in Claude. Now the whole pre-trip flow — researching who they’re meeting, prepping the agenda, checking the schedule — runs in one conversation. When it’s time to book, nothing breaks. Claude talks to the agency’s agent directly. The arrangements get handled. The traveler never left the thread they were already working in.

Preparation and booking stopped being two separate jobs. They became one continuous thread. That’s B2A working, not B2A in theory.

Brands that don’t adapt to this aren’t falling behind in some gradual way you can fix next quarter. They’re being removed from the set the agent even considers. You don’t lose the comparison. You’re not in it.

So the question I’d ask, if you run a travel brand: when an agent goes looking on your customer’s behalf, does it find you — and does your data hold up when it does?

This is the kind of thing we work through building Bitravel — putting a real travel agency behind an agentic layer so the booking happens inside the agent’s flow, not bolted onto the side of it. Book a 30-minute call for a live look, no deck.