NewsStocksBooking Holdings CFO Says Even AI Hyperscalers Don't Know Their ROI—and He's Learning From an 'AI Coach'

Booking Holdings CFO Says Even AI Hyperscalers Don't Know Their ROI—and He's Learning From an 'AI Coach'

Author: Fortune Crypto·

Key Takeaways

  • •Booking Holdings CFO Ewout Steenbergen said at the Fortune AIQ Summit that hyperscalers investing hundreds of billions of dollars in large language models do not truly know what returns their spending will produce.
  • •Internal AI deployment has lowered customer service costs while bookings grow at a high-single-digit rate, and Booking's roughly 9,000 engineers are delivering about 30% more code that passes testing and quality control.
  • •Referrals from large language models still account for less than 1% of room nights, while Booking spends $8 billion to $9 billion annually on paid channels that bring in about one-third of its customers.
  • •Booking controls AI expenses through model cost routing, assigning simple tasks to cheaper basic and open-source models while reserving expensive models for complex work and measuring teams on total IT cost per merge request.
  • •Steenbergen said AI could turn Booking into a more high-frequency business by proactively adjusting trips, with connected-trip transactions growing at a low-double-digit rate, merchant bookings reaching about 73% of gross bookings, and top Genius loyalty tiers exceeding 30% of active customers.
Booking Holdings CFO Says Even AI Hyperscalers Don't Know Their ROI—and He's Learning From an 'AI Coach'

Booking Holdings' chief financial officer believes the companies spending the most on artificial intelligence are essentially guessing at their returns. Speaking Thursday at the Fortune AIQ Summit, held at the New York Stock Exchange, CFO Ewout Steenbergen said even the AI industry's biggest spenders lack a clear picture of what their massive investments will deliver.

“I even think that the hyperscalers that spend hundreds of billions on the development of their large language models, they don't really know what is going to be the ROI,” Steenbergen said. “They have maybe some assumptions and some hypothesis, but the whole point is, you don't want to fall behind.”

That ambiguity, he suggested, is good news for everyone else. Companies that use AI models rather than build them pay far less—provided they keep a close watch on the bill. “The returns will be there for those processes that are being redesigned end to end,” he told Fortune senior writer Sheryl Estrada, “but you also have to make sure that those costs on the other hand are not going out of control.”

Booking Holdings, the parent company of Booking.com, Agoda, Priceline, and OpenTable, has told investors that AI will make travel easier for customers and marketing cheaper for the company. Customer-facing results, however, remain modest. On the company's second-quarter earnings call in August, Steenbergen said referrals from large language models, both paid and unpaid, accounted for less than 1% of total room nights. The clearer gains, he told Estrada, are emerging inside the company.

Watching the Token Bill

The past two years were spent experimenting, Steenbergen said. The next phase involves rebuilding processes “from a white sheet.”

Customer service went first. Bookings are rising at a high-single-digit rate while customer service costs are slightly down, he said, meaning the cost per booking has fallen “by a lot.” Customer satisfaction rose at the same time. Meanwhile, Booking's roughly 9,000 engineers are getting about 30% more code into production, counting only merge requests that pass testing and quality control.

To contain AI spending, Booking relies on what Steenbergen called “effective model cost routing.” Basic and open-source models handle simple tasks, while more expensive models take on the complex ones. Engineering groups are measured on total IT cost per merge request, which combines human and AI-token costs. Token spending can rise, he noted, as long as the cost of each merge request reaching production comes down.

'The Hotel Is Still the Hotel'

AI will not change how people travel, Steenbergen said—only how they plan and book. “I think fundamentally the way … travel will happen in the future is not going to change with AI, but what is going to change is the experience for travelers,” he said. “The hotel is still the hotel, and the airline is still the airline, and the rental car is still the rental car.”

Travelers visit about five platforms on average before booking, he said. AI could shorten that research phase and then manage disruptions—such as a delayed flight that upends a restaurant reservation. Booking refers to this vision as its “connected trip.”

Roughly one-third of Booking's customers arrive through paid channels such as search, social media, and metasearch sites, and the company spends $8 billion to $9 billion a year on those channels, he said. AI referrals, at less than 1% of room nights so far, supply only a sliver of the traffic those channels provide. Booking runs constant A/B tests and large optimization models to determine where that money goes. The remaining two-thirds of customers come directly, and the company is embedding AI tools in its own apps to keep them.

Booking's own AI tools are showing modest gains, he said. Customers who use them take slightly less time to book, convert at somewhat higher rates, and cancel a bit less often. “It's very early stage, so it's not a lot of data,” he said.

From Transactional to High-Frequency

Steenbergen framed the larger opportunity as growth rather than savings. Answering an audience question about trip bundles, he said Booking is “more transactional” today, and AI could turn it into “a much more high-frequency business.”

He offered an example: a traveler has a trip to Paris planned, and the forecast for Wednesday calls for bad weather. “We can proactively say, 'Hey, shall we change your schedule around on Wednesday? It's bad, so why don't you do the Louvre tour then and your walking tour on Thursday?'” If a sports-loving customer has an empty day, the system could offer to book kite surfing. That, he said, would build “more loyalty, more trust, and more brand value.”

Booking points to early evidence in its second-quarter numbers. Connected trip transactions, in which travelers book more than one travel vertical for the same trip, grew at a low-double-digit rate at Booking.com. Merchant bookings—transactions in which Booking processes the customer's payment—reached about 73% of gross bookings, and Level 2 and Level 3 Genius members, the upper tiers of Booking.com's loyalty program, made up more than 30% of active customers. Such figures recur in the company's quarterly reports, giving investors a running baseline for the connected-trip strategy he described.

Steenbergen is learning alongside the company. He said he has an “AI coach,” and he talks about it inside the company deliberately. He also has two AI agents working for him: one, which he calls a “strategic thought partner,” helps with board presentations and strategic plans; the other, a “critical equity research agent,” helps him prepare for earnings calls.

“No one knows this,” he said. “If I have to learn, and I have a coach, it's very normal. Everyone has to learn.”

Editor's note: For this story, Fortune journalists used generative AI as a research tool. An editor verified the accuracy of the information before publishing. This story was originally featured on Fortune.com.