Who Your AI Travel Agent Actually Works For

From the Travel Tech Podcast newsletter for episode 34, Trust as Infrastructure: Who Your AI Travel Agent Actually Works For.

Read the small print on the big booking sites and you'll find the offer at the top may not be the one in your best interest, just the one that pays best. Travel has quietly accepted that for decades, and the six people in this episode think AI agents are about to make it everyone's problem. None of them were pitching an agent. For those of you across aviation, travel tech, and AI, I wanted that perspective on the show.

One guest spent 15 years at IATA and put the barcode on your boarding pass. Another left the OECD to build a flight search tool alone in Paris, another launched one of the first online travel agency stacks in 1996, and the rest built the AI behind Meta's Ray-Ban glasses, work on protecting travellers from fraud, and have spent careers in airline distribution plumbing. Listen on Apple Podcasts, also streaming on Spotify and YouTube.

They disagree about who wins the agentic booking race. They agree completely that nobody is accountable when an agent gets it wrong.

The IATA nobody has built for agents

That tension has a structural cause. Travel's intermediary started as an airline or railway ticket office, then grew to sell every supplier and every mode of transport, ending up as one agent advising both the traveler and the supplier at once. The plumbing reflects it: most bed banks still can't be reached directly by an agent, and fake hotel sites with full image libraries already outrank real properties on their own names.

AI adds pressure at a scale distribution was never designed for. The GDS model was built for hundreds or thousands of agencies querying inventory, and Jim Hetzel points out it could soon be facing billions of bots. Personalisation is working (an agent that knows you have kids or support Tottenham can route you to the right supplier), but unsourced recommendations still send users clicking through to check whether the answer is even legitimate.

In this episode, Eric Leopold, Christopher Olivares of Elyo, Mike Putman of Custom Travel Solutions, Nancy Li Smith of Boop, Fred Bean of Hotelport, and Jim Hetzel cover the buyer's agent versus seller's agent split, the hidden cost of free booking platforms, brands getting routed around by personal agents, social trust versus model quality, and what an IATA style trust layer for agents would actually need.

Key Takeaways

• The subscription realigns the agent: A traveler paying directly for an assistant creates a buyer's agent that travel distribution has lacked since the ticket office went multi-supplier.
• OTA commissions are a traveler cost: Hotels pass some or most of the distribution fee into the room rate, so "free" platforms are paid for indirectly.
• Brand loyalty won't survive a good agent: Once an agent knows the traveler well, it will route to whichever supplier or channel delivers the best outcome, regardless of the OTA logo.
• Provenance is the product: Users trust recommendations anchored to a person with a reputation at stake more than a model's unsourced output.
• Trust is five functions, not one: Agent to agent commerce needs identity, authority, enforcement, settlement, and dispute resolution, none of which can be built in a sprint.

Alex Brooker, Host of the Travel Tech Podcast, Founder of Airside Labs