The $9 Billion Question: How Booking Holdings Is Betting AI Can Rewire Trip Planning
(SeaPRwire) – By: Robert Kensington
Ewout Steenbergen walked into the AIQ Summit on October 1st carrying a heavy burden. His company spends nearly nine billion dollars a year chasing travelers. That is not a side figure. That is the central operating cost of a business built on visibility. Every dollar spent on search ads and metasearch platforms is a dollar extracted from margin. The question now is whether artificial intelligence can actually shrink that number, or whether it simply becomes another expensive line item on top of the existing ones.
The official story from Booking Holdings is elegant. They envision a “connected trip” where flights, hotels, restaurants, rental cars, and activities are stitched together by AI. Rain threatens your Paris itinerary and the Louvre reservation shifts indoors automatically. Your flight delays and the dinner booking moves forward without you lifting a finger. This is the promise. It is also the kind of promise that sounds impressive in a keynote but requires a complete restructuring of how customer service, engineering, and acquisition channels operate to actually deliver it. The subtext is sharper. Booking does not want to build a better chatbot. They want to rebuild the entire customer acquisition funnel around a system that converts faster, cancels less often, and keeps travelers inside their platform rather than bouncing between six different apps.
The numbers matter here. Booking pulls roughly one-third of its consumer traffic from paid channels. The remaining two-thirds arrive directly. That means about eight billion dollars in annual spend buying the first third. Steenbergen acknowledged that even hyperscalers spending hundreds of billions on large language models do not really know what the return on investment looks like yet. They operate on assumptions and hypotheses. But the strategic logic is brutal. You do not fall behind. That is the real answer. Meanwhile, early signals from Booking’s own AI tools show booking occurring slightly faster, conversion rates ticking upward slightly, and cancellation rates dipping slightly. The word “slightly” does a lot of heavy lifting. It also reveals a company in the messy middle phase where pilots are expanding beyond isolated experiments but the unit economics have not yet proven themselves at scale.
The internal engineering data tells a different kind of story. Booking has roughly nine thousand technologists. They are reporting about a thirty percent increase in merge requests pushed to production. Thirty percent is not trivial. It matters that this figure reflects code that passed quality controls, not raw generation output. On the customer service side, booking units are growing at a high single-digit rate while total costs declined slightly and satisfaction increased. That is the triangle that makes financial sense. More volume, lower cost, happier customers. The CFO’s approach is to monitor token costs and licensing fees closely. They are using what Steenbergen calls “effective model-cost routing,” reserving cheaper models for simpler tasks and expensive ones for complex decisions. One engineering metric they track combines human and AI costs into a single number: total IT cost per merge request. This is the mature way to think about AI investment. The question was never whether a model costs more in isolation. It is whether it lowers the all-in cost of producing a valuable outcome.
Booking is applying this internally with personal tools. Steenbergen uses an AI coach, a strategic thought partner for board materials, and a critical equity research agent to pressure-test earnings call preparation. He is not outsourcing judgment. He is stress-testing it. The financial model underlying all of this rests on channel diversification. As consumers shift toward AI agents for discovery, paid acquisition costs could spread across new surfaces. That would improve unit economics. But only if the connected trip actually works. If it does, loyalty deepens, trust compounds, and engagement frequency rises. If it does not, Booking remains exactly where it started. Burning eight or nine billion dollars a year to reach people who were going to book anyway, just through slightly different screens.
Author bio: Robert Kensington, an overseas entrepreneurial veteran with decades of experience in real-economy industrial investment and expansion.