Booking Holdings is posting a rare result in the current AI investment climate: customer service costs down 10% per reservation while gross bookings grew roughly 10% year over year. CFO Ewout Steenbergen confirmed the figures on the company’s Q4 2025 earnings call<\/a>, framing generative AI as the primary driver. This lands against a backdrop where PwC finds only 12% of CEOs report AI delivering both cost and revenue benefits simultaneously. Booking, it seems, is in that 12%.<\/p>\n The number that will travel fastest through boardrooms is 10%, but the more telling figure is the direction of the two curves: costs fell while volume rose. That combination, declining unit cost against growing transaction count, is the signature of genuine automation rather than headcount substitution dressed up in AI language. If your customer service operation is still tracking cost per contact as a flat or rising line despite AI investment, Booking’s result is not an inspiration, it’s a benchmark you’re now being measured against whether you asked for it or not.<\/p>\n There’s a structural reason Booking moved quickly here that most companies won’t want to hear. Travel customer service is high-volume, heavily repetitive, and multilingual, which makes it among the most tractable targets for generative AI automation. Booking.com handles hundreds of millions of reservations annually, so even marginal per-contact efficiency compounds into material savings fast. Companies whose customer service queues are lower volume, higher complexity, or more regulated face a longer path to equivalent results. The benchmark is real, but the replication timeline isn’t the same for a bank or a healthcare system as it is for an online travel agency.<\/p>\n The decision this reframes isn’t whether to invest in AI for customer service. It’s whether your current AI deployment is structured to produce a measurable unit-cost curve or just a collection of pilot results that look promising in a slide deck. Booking’s $450 million transformation program, announced in late 2024, gave the AI investment a hard savings target and a deadline of end-2027. That accountability architecture, a named CFO commitment tied to a specific program, is what separates a cost result from a cost story. The CFO who can’t yet name the unit metric their AI investment is moving probably can’t claim the same result.<\/p>\nWhat this means for your business<\/h2>\n