Booking Holdings spends nearly $9 billion a year to reach travelers. Its CFO explains how AI is changing trip planning

Booking Holdings CFO Ewout Steenbergen said early data show customers using the company’s AI tools book somewhat faster, convert at somewhat higher rates, and cancel less often, while noting the evidence is preliminary. Booking’s booking units grew at a high single-digit rate as total customer-service costs declined slightly, and AI tools helped roughly 9,000 technologists increase tested, quality-cleared production merge requests by about 30%. The company is evaluating AI using all-in unit economics, including token, licensing, human, and IT costs.

Analysis

The investable question is who owns the traveler relationship when AI becomes the discovery layer. If Booking can turn its existing customer base into repeat, multi-service trips, higher attachment and fewer service interventions could improve contribution per customer while reducing dependence on paid search. But discovery migrating to third-party agents is not automatically a CAC benefit: those agents may charge for placement, steer users to direct supplier inventory, or retain the customer data and loyalty relationship. Hotels, airlines, and restaurants could gain bargaining power if AI makes their offers easier to compare.

Management’s early conversion and service-cost signals are encouraging, not yet proof of durable operating leverage. Higher engineering throughput may also fund faster product expansion rather than lower expense. The key verification points are paid-acquisition cost per booking, direct/repeat mix, cross-service attachment, cancellation and service cost per booking, and whether AI-originated traffic carries new distribution fees.

Near term, the interview alone is not a sufficient catalyst for repricing. Over 1–3 months, watch earnings commentary for measurable unit economics and evidence that AI traffic converts without a costly intermediary. Over 6–18 months, the structural outcome hinges on whether Booking’s proprietary inventory, customer data, and service recovery make its own interface valuable even when an external agent initiates the journey. The thesis weakens if AI-led bookings grow but paid distribution costs or third-party fees rise, or if customer retention and attachment fail to improve.

Key Decisions for Investors

• No immediate directional trade on the interview: treat the operating claims as management-reported and await independently visible booking economics.
• Set a conditional long-BKNG watch for earnings or guidance evidence of falling acquisition cost per booking alongside stable or improving conversion and repeat/direct mix; avoid treating AI feature adoption alone as confirmation.
• Monitor Expedia and Airbnb as competitive read-throughs, and hotel and airline direct-channel commentary for signs that AI agents are shifting customer ownership toward suppliers rather than OTAs.
• Falsify the positive thesis if AI-originated volume rises without better attachment or retention, while paid-channel costs, agent distribution fees, or service cost per booking increase.