The First AI Win at Booking Is Cost

Booking Holdings, Inc. reported a measurable gain from AI on its most recent earnings call. The company said customer-service cost per booking fell about 10 percent year over year, even as bookings rose roughly 10 percent. CFO Ewout Steenbergen attributed the decline to generative AI tools deployed in support operations. He said the result was visible in the company’s operating expense line and contributed to improved margins.

The disclosure comes as large language models are increasingly used for travel planning and research. In an interview published a day before the earnings call by McKinsey & Company, CEO Glenn Fogel acknowledged the risk that transactions could eventually occur within LLM ecosystems rather than on Booking’s own platforms. “There’s certainly always a risk,” he said.

Booking has shown AI can improve its cost structure. The harder question is whether it can also protect its position in distribution.

Where AI Gets Measurable

On the call, Steenbergen said customer-service costs declined on a per-booking basis by roughly 10 percent year over year. He linked that improvement to generative AI deployed in support functions. Management also pointed to margin expansion supported by technology-driven efficiencies, particularly in customer service.

This use of AI is operational. It focuses on handling inquiries, resolving issues faster, and reducing manual intervention. The impact appears in operating expenses rather than in new product features.

Scale magnifies the effect. Booking processes hundreds of billions of dollars in travel annually. In 2024, it reported $166 billion in gross bookings and more than $8 billion in adjusted earnings.

Competitors have discussed AI in customer support and trip planning. Expedia Group, Inc. has highlighted automation and conversational tools, including generative AI integrations.

Airbnb, Inc. has described AI-assisted search and support improvements in product updates and earnings commentary.

Neither has publicly cited a specific percentage reduction in cost per booking tied directly to generative AI.

The earnings call established that at least one AI initiative has moved beyond experimentation and into measurable financial impact.

In the McKinsey interview, Fogel described the “connected trip” as a system in which flights, hotels, ground transport, and attractions are coordinated through technology. If a flight is delayed or weather shifts, plans adjust. The goal is to use AI and machine learning to anticipate problems and respond quickly.

Fogel pointed to the industry’s regulatory and payment framework as sources of complexity. Executing travel transactions, he argued, requires trust, compliance, and infrastructure that extend beyond recommendations. He said it could take a long time before large language models can manage the full combination of execution and trust required in travel.

Large language models are becoming another interface for travel discovery. If they capture more of the search and planning process, online travel agencies could face pressure on traffic and distribution.

Booking’s position rests on scale, cross-vertical coverage, and direct customer relationships. Fogel said roughly the mid-60 percent of customers come directly to the company’s platforms.

So far, the quantified AI impact Booking has disclosed relates to operating efficiency. The company has not provided specific metrics showing higher conversion rates, increased cross-sell, or measurable revenue lift tied directly to AI-driven personalization.