The look-to-book multiplier: what AI agents really do to travel distribution

Look-to-book nearly drowned travel distribution years before anyone said the word “agent.”

I spent years on the tech front line of two large B2B travel suppliers — billions of search requests a day. Almost none of them came from a person actually deciding where to fly.

How the number got that big

Someone opens a meta-search to compare prices. That one query fans out to every OTA. Each OTA queries every B2B supplier it works with. And the suppliers were wired into each other — A asks B, B asks C, C asks back into A, the same request looping through the network until the TTL kills it around eight seconds later.

One human click. Thousands of hits on the same provider.

That traffic costs real money, and it lands in the price the traveler pays. So we watched look-to-book per partner obsessively. Past a million searches per booking, we’d act — thresholds, penalties, sometimes disconnection. I once had to cut off a partner I was close with; instead I asked for access to his system, found what was inflating his ratio, and fixed it myself so I wouldn’t have to.

Then the agents arrived

A single AI agent recently pulled 881,076 fares out of one airline in a single session — just to find the best flight. If you’ve ever managed distribution costs, you recognize it instantly.

The agent doesn’t pass your one query through. You say “Berlin, July 22.” It checks the 22nd, three days either side, the nearby airports, the alternate routings — because that’s how it builds you a better answer.

And then there’s the part nobody priced in: the bot never gets tired of you. Ask a human travel agent twice and you already feel like a nuisance. Fill in a meta-search form once and you stop. The bot has infinite patience and infinite attention, so travelers throw it everything they’d never dare ask a person — and it goes and searches all of it.

Same old look-to-book problem. A tireless new multiplier on top, now from both ends.

Why this time is different

Here’s what I keep coming back to. We could never fix the old loops — the protocols were legacy and frozen, nobody could touch them. Right now the whole industry is building the new agentic protocols from scratch. For the first time in fifteen years, the plumbing is open.

If you’re shaping this new agentic travel world, we already know where the loops form and where the cost hides. Don’t carry those problems into the new system. Build the fix into the protocol itself — a flag that ends the loop before it starts.

What it means if you buy or sell travel

For corporate travel teams, this is why an agent that simply “searches more” is a liability, not a feature. The right agent is the one that’s accountable for cost and policy as it searches — not one that fans out 800,000 fares on your account.

For airlines, OTAs and loyalty partners, the agents are coming to your inventory whether or not you design for them. Better to expose a clean, governed interface than to absorb the loop.

We spent a decade containing this. We finally have the chance to design it out instead. Let’s not waste it.

This is the kind of thing we think about while building Bitravel — an AI travel agent that’s fast like AI and accountable like a person. Book a 30-minute call for a live look, no deck.