The last two years of travel-tech strategy have organized around one question: can we match the market leaders on AI-powered search and booking? For most platforms, the honest answer by mid-2026 is yes, roughly. The chatbots exist. The recommendation engines exist. The AI trip-planning features are live, or close enough that the roadmap slide reads as complete.

That is the wrong finish line. Winning that race gets a platform to parity on the 15 to 20 minutes a traveler spends deciding what to book. It says nothing about the days or weeks the traveler then spends in the destination, making decisions, and making them with a platform that is not the one that booked the trip.

The booking moment was never where the largest pool of uncaptured revenue lived. It is a real pool, and matching the leaders on AI-assisted search was a legitimate defensive move; falling behind there would have been an existential problem. But the in-destination economy sits entirely on the other side of that transaction, in the part of the journey every platform’s current AI investment stops short of touching. Parity on search recovers none of the in-destination spend that leaves once the traveler lands. It just means the platform is now even on the smaller of the two problems.

Where the incumbents have concentrated is instructive. Their AI investment, like most of the industry’s, has been optimized for conversion at the point of booking, because that is where their revenue model concentrates. Booking.com and Expedia built for the transaction, not for holding a continuously validated, hyper-local picture of what is open, bookable, and worth bundling in a destination this week. That is a genuinely different infrastructure problem, and scale alone does not solve it.

This is the question if the in-destination layer is so valuable, why has a platform with far more engineering capacity not simply built it?

Because the hard part is not the code.

Agentic AI works wherever cooperative systems already exist, the pre-negotiated technical, commercial, legal, and data rails that let software act on a real transaction. Those rails are the slow part. Structured supplier relationships, redemption and settlement integrity, certification, and multi-jurisdiction data handling take years to assemble, and they cannot be bought with model access.

This is the distinction a 2026 reader is most likely to miss, because they have watched their own engineering costs fall. AI compressed the cost of writing software. It did not compress the time required to negotiate supply, establish redemption integrity, or clear certification and data-handling requirements across jurisdictions. A platform that assumes the in-destination layer got cheaper because code got cheaper is mispricing the one part of it that did not move. That is also why the post-check-in position is winnable without out-engineering anyone: what stands between a platform and this layer is negotiated supply, redemption integrity, and certification, not raw engineering horsepower.

The reason this does not stay open is that the market underneath it keeps growing while the rails stay slow to build. Experience travel services are projected to grow from $138B in 2024 to $372.93B by 2034, a 10.5% compound annual growth rate. The leak widens with the market, and the years it takes to assemble the rails do not compress to meet it. A platform deciding this cycle is really deciding one thing: whether to spend those years building the rails, or to integrate a layer where they already exist.

Parity bought nothing durable. The rails take years. And the choice, build them or integrate them, is in front of the reader this cycle, not next. Matching the leaders bought parity. It didn’t buy the next advantage.

Jeff Kischuk

CEO and Founder Tripian