What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Booking Holdings CFO Ewout Steenbergen says even the companies spending heavily to build large language models cannot yet predict their return on investment with confidence. Booking is still putting AI to work, but its clearest reported gains so far are internal—alongside careful cost tracking—while customer-facing results remain early and uncertain.
What Booking’s CFO says about AI returns
“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 told Fortune. He said companies may have assumptions and hypotheses, but also do not want to fall behind.
As an Amazon Associate I earn from qualifying purchases.
That is Steenbergen’s assessment of the difficulty of forecasting returns, not evidence that AI cannot pay off. His distinction is between spending on model development and redesigning work around AI. “The returns will be there for those processes that are being redesigned end to end,” he said.
Where Booking says it is seeing early gains
Customer service and engineering
Fortune reports Steenbergen said customer-service costs were slightly lower while bookings grew at a high-single-digit rate. He also said Booking was producing about 30% more code in production, with the comparison counting merge requests that pass tests and quality control.
#1 Best Overall
These are company-reported operating measures, not independent proof that AI alone caused the changes. The figures indicate where Booking says it is seeing progress, but do not isolate AI’s contribution from other factors.
Customer-facing booking changes
Steenbergen described improvements in booking time, conversion and cancellations as modest and based on little data. “It’s very early stage, so it’s not a lot of data,” he said. Fortune also reported that AI referrals accounted for under 1% of total room nights, a figure Steenbergen gave on Booking Holdings’ Q2 call in August 2026; it includes paid and unpaid referrals from large language models.
How Booking is trying to manage AI costs
Rather than send every task to the most capable and costly model, Booking says it routes simpler work to basic or open-source models and more complex tasks to more expensive ones. That approach treats model choice as a cost-versus-capability decision for each task, rather than assuming one model should handle everything.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Fix the driver behind crashes, sound loss and screen glitches3Repair Windows errors before they cause bigger problemsFor engineering, Steenbergen said the company tracks total IT cost per merge request, including both human and token costs. Counting only model usage would miss the labor involved in producing and reviewing code; measuring against accepted work ties spending to an output the company can assess. Fortune’s account does not provide the underlying calculation or a time series, so the metric is best understood as Booking’s stated management approach, not a public benchmark.
Building AI models versus using them
Steenbergen’s remarks draw a useful distinction: hyperscalers are investing in building large language models, while Booking is using models within travel and internal workflows. The spending scale he described—“hundreds of billions”—is his rounded characterization, not an independently established industry total in Fortune’s account.
For a company using models, the practical question is whether a task is completed better or more cheaply after including all relevant costs. Booking’s reported cost-per-merge-request measure is one example of that framing. The broader point is not that every AI project needs the same metric, but that model capability by itself does not show whether a business process is delivering value.
Rank #4
Why Booking is pursuing connected trips
Booking’s longer-term opportunity is not limited to making an individual hotel or flight search faster. Steenbergen described a connected-trip goal: helping travelers coordinate bookings across more than one travel vertical and respond when plans are disrupted. Booking.com reported low-double-digit growth in connected-trip transactions in Q2 2026, as covered by Fortune. A connected trip involves booking in multiple travel categories for the same journey.
The business rationale combines repeat use and loyalty with the possibility of serving more of a traveler’s needs. Steenbergen said travelers visit about five platforms on average before booking; Fortune’s report does not identify a separate underlying study for that figure, so it should be read as his estimate, not a universal measurement.
Other figures in Fortune’s account help explain Booking’s existing distribution model, but do not by themselves demonstrate AI returns. Steenbergen said about one-third of customers arrive through paid channels such as search, social media and metasearch, at an annual cost of $8 billion to $9 billion. Fortune also reported that the merchant model represented about 73% of gross bookings in Q2 2026, and that Level 2 and Level 3 Genius members together made up more than 30% of active customers. These are reported company figures, not independent evidence that AI caused a change in acquisition, bookings or loyalty.
What Steenbergen means by his “AI coach”
Steenbergen said he uses an AI coach and two agents: one as a strategic thought partner for board presentations and plans, and another to prepare for earnings calls through equity research. Fortune did not name the products or providers. His example is about learning and personal productivity, not a claim that a specific AI tool has delivered a measured financial return.
“If I have to learn and I have a coach, it’s very normal. Everyone has to learn,” he said. The example reflects a practical way to approach unfamiliar tools: use them to support specific work while continuing to evaluate what they contribute.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
