Booking CFO airs AI return concerns as benchmarks link discipline to real gains.
CFOs demand ROI as AI bills mount.
Chief financial officers are increasingly insisting on hard proof of returns from AI initiatives as spending on the technology accelerates. In a new survey of 270 finance leaders released this week, CFOs reported that their teams now devote 26% of the workweek to verifying or correcting the output of AI tools ([1]) ([2]), time that chips away at the hoped-for efficiency gains. Almost one-third of CFOs also said their companies exceeded their AI budgets by over 10% in the past year ([3]). These findings underline a growing "verification burden" on finance departments, as the drive to adopt AI is not yet translating proportionally into cost savings or productivity improvements.
Finance chiefs are also voicing concern that they lack visibility into how much value AI is actually delivering for the money spent. A separate late-September poll of 172 CFOs, finance directors and procurement heads found 57% are not confident they’re paying a fair price for AI, and only 51% say their AI investments are achieving measurable ROI ([4]) ([5]). Nearly half admitted they went over budget on AI projects last year ([6]). Meanwhile, a majority of CFOs in a recent Deloitte survey cited lack of cost transparency as their top internal worry around AI initiatives ([7]). Taken together, these reports paint a picture of finance leaders doubling down on governance and cost control. Many are pushing for detailed AI spending audits and insisting that new AI proposals come with clear metrics to track business value. As one finance VP noted, “You can’t negotiate a price you can’t see, and you can’t consolidate a stack you’ve never fully inventoried” ([8]).
Booking.com sees internal gains, preaches caution.
Even as many firms struggle to quantify returns on their AI investments, some companies are pulling ahead by measuring impact in specific operational areas. At the recent Fortune AIQ Summit in New York, Booking Holdings CFO Ewout Steenbergen revealed that AI-driven automation has already helped the online travel group reduce customer service costs per booking “by a lot” while improving customer satisfaction ([1]). Internally, the gains are more measurable: the company’s roughly 9,000 engineers are delivering about 30% more code into production thanks to AI coding assistants, according to Steenbergen ([2]). To keep machine-learning expenses in check, Booking uses what he calls “effective model cost routing,” assigning cheaper basic models to routine tasks and reserving more expensive large language models for complex problems ([3]). Engineering teams are now being evaluated on total IT cost per software release, combining cloud expenses for AI models with human developer time, to ensure that productivity wins aren’t cancelled out by surging usage fees ([4]).
Despite these efficiency improvements, Steenbergen cautioned that even the world’s biggest AI investors “don’t really know” their return on developing advanced models ([5]). In the consumer business, he noted, referrals from generative AI currently account for under 1% of Booking’s total hotel nights ([6]), a signal that AI’s top-line impact remains marginal so far. The CFO’s message for peers was twofold: first, the real payoffs will come from redesigning whole processes end-to-end with AI rather than tinkering at the margins ([7]); and second, companies must prevent AI costs from running out of control even as they adopt the technology, or any potential return will quickly evaporate ([8]).
Disciplined AI bets yield measurable results.
New industry benchmarks suggest that the key to unlocking AI’s value is a disciplined, enterprise-wide approach. The Hackett Group this week released its “AI World Class” benchmarks, which quantify how leading organisations change their economics through focused AI investments. According to the research, companies that systematically modernise their IT and business processes with AI achieve dramatic efficiency gains: they reduce IT operating process costs by 27%, 35% and improve project return on investment by 1.5× to 2.5×, on average ([1]) ([2]). These benefits extend beyond IT. For example, top-quartile adopters in finance are redesigning order-to-cash activities to cut processing costs by more than half while speeding up cash collection ([3]). In procurement functions, AI leaders are obtaining as much as 2.7× higher ROI compared to typical firms ([4]). The common thread is an emphasis on end-to-end process reengineering and rigorous prioritisation of use cases by business value, rather than scattershot experimentation ([5]) ([6]).
One real-world illustration comes from US healthcare giant CommonSpirit Health. Its CFO Michael Browning told investors this month that the hospital system has “captured $200 million in financial value from artificial intelligence,” by tracking each AI initiative from pilot to post-implementation audit ([7]) ([8]). Notably, CommonSpirit has kept direct AI spending to under 0.75% of its IT budget, using small-scale experiments and strict governance to ensure ROI-positive results ([9]). The lesson for other organisations is that artificial intelligence can deliver material savings, but only when projects start with clear, outcome-based goals and disciplined cost management. In the current climate of heightened scrutiny, the era of open-ended “AI for AI’s sake” appears to be over.