Rebecca Greenfield/Fortune
Not even the companies pouring the most money into artificial intelligence can say with confidence what they’re getting back for it, according to Booking Holdings CFO Ewout Steenbergen, who argued that uncertainty extends all the way to the top of the industry.
Speaking at the inaugural Fortune AIQ Summit in New York on October 1, held in partnership with ServiceNow, Steenbergen said hyperscalers spending hundreds of billions of dollars developing large language models “don’t really know what is going to be the ROI” on that spending, a striking admission given how much capital has flowed into frontier AI development industry-wide.
At Booking itself, the payoff looks uneven depending on where you look. Referrals driven directly by large language models currently account for under 1% of the company’s total room nights, a modest number that undercuts some of the louder claims about AI reshaping consumer search and booking behavior overnight. Internally, though, the gains have been more tangible: Booking’s roughly 9,000 engineers are getting about 30% more code into production, a figure Steenbergen said counts only merge requests that actually pass testing and quality control, not just raw output.
That split informs how Steenbergen is approaching AI investment more broadly. “Returns will be there for those processes that are being redesigned end to end,” he told Fortune, but he was equally clear that spending needs limits: companies have to make sure “that those costs on the other hand are not going out of control.”
Steenbergen also offered a candid note on how he personally keeps up with a technology moving this fast, describing himself as working with what he called an “AI coach” to stay current. “If I have to learn and I have a coach, it’s very normal. Everyone has to learn,” he said, framing continuous learning as a basic requirement for executives navigating AI rather than something unusual for someone at his level.
The comments land as finance chiefs across industries face growing pressure to show hard numbers behind AI spending that has, in many cases, been justified more by competitive necessity than proven returns. Steenbergen’s message was less a dismissal of AI’s potential than a caution against assuming the returns are automatic: at Booking, the technology is delivering measurable gains in engineering output while still barely moving the needle on the customer-facing side, a split that may be more common across the industry than hyperscalers’ massive capital commitments would suggest.
