For decades, commerce operated on two axes: you sold to consumers (B2C) or you sold to businesses (B2B). Then came a third axis nobody saw coming — selling to agents.
Y Combinator recently called it B2A — business-to-agent. The idea is simple: as AI agents become the primary way people interact with the internet, they will increasingly be the ones making purchases, booking services, and executing transactions. The human is still there, but one step removed. The agent is the customer.
Travel booking is going to be one of the first markets where this transition is complete. Not in five years. Now.
Why travel, and why first
Think about what a hotel booking actually requires. A destination. Dates. Budget. Maybe a few preferences. That's it. The task is bounded, well-defined, and produces a binary outcome: either a booking happens or it doesn't.
Compare that to buying a car, choosing a school, or selecting a healthcare provider. These decisions involve deep personal context, years of relationship-building, and emotional weight that agents aren't equipped to handle — yet. Travel is different. The parameters fit neatly inside a JSON object. The task is exactly the kind of bounded, repeatable, high-frequency job that agents excel at.
The numbers back this up. Over a billion people book travel online each year. The global OTA market is worth hundreds of billions. And crucially, users already accept near-total automation in adjacent tasks — flight check-ins, itinerary management, travel alerts. The leap to "my agent just booked the hotel" is smaller than it seems.
The infrastructure gap
Here is the problem: the travel industry was built for humans. Every OTA, every metasearch engine, every booking platform was designed around a browser, a screen, and a person making choices from a list.
Agents don't need lists. They need decisions. And they need infrastructure that speaks their language — structured outputs, confidence scores, single recommendations, validation before execution. None of the incumbents provide this. They were never designed to.
This is the gap that b2a Bluepillow fills. Not competing with OTAs for the booking itself, but sitting between the agent and the OTA as the intelligence layer — the entity that receives a query, aggregates supply across 16+ sources, applies proprietary ranking based on real conversion data, and returns a single answer.
What B2A looks like in practice
In a B2A travel stack, the flow changes fundamentally. The user tells their agent: "Book me somewhere nice in Rome for the weekend." The agent doesn't open a browser. It calls an API. That API — b2a Bluepillow — returns: hotel X, via OTA Y, at price Z, confidence 0.94. The agent executes. The user sees a confirmation.
No search. No comparison. No decision fatigue. The human delegated entirely, and the infrastructure handled the complexity underneath.
The neutrality advantage
b2a Bluepillow doesn't sell bookings. It doesn't compete with OTAs for margin. It doesn't have an advertising model that creates bias in its recommendations. Its only incentive is to make the best possible recommendation.
In the B2A era, neutrality isn't just an ethical choice. It's a competitive moat.
What comes next
Travel is the proving ground. But the B2A pattern — a neutral decision layer between agents and a fragmented supply market — will repeat across verticals. Restaurant reservations. Ground transportation. Event ticketing. Every category where supply is abundant and the decision is bounded enough for an agent to handle.
We're building the first one. If you're building agents that touch travel, you already know where this is going.