The pace at which finance chiefs are becoming responsible for AI oversight resembles the steep upward curve in companies’ overall use of the new technologies, a new report suggests.
In a Deloitte survey of CFOs at large North American companies, more than two-thirds of those who oversee AI trust, capital allocation or cost management said they took on an owner or approver role within the last three years.
Nearly 80% of the 200 respondents, all from companies with at least $1 billion in annual revenue, said finance has a formal role in AI-related decisions, with 38% defining or administering an evaluation framework and 41% serving as required approvers.
That left 20% who said they provide cost, budget or accounting guidance but do not shape AI decisions.
About half (53%) of survey participants indicated that the primary way they review AI-related costs is retrospectively, through vendor bills, reports, dashboards or manual analysis. Somewhat fewer (47%) are more proactive, by monitoring AI usage costs in or close to real time and actively managing deviations using alerts, dashboards, spending-anomaly reviews or budget forecasts, Deloitte wrote.
Relatively few of the CFOs reported using advanced AI-specific controls to manage consumption costs, while almost a quarter are using general controls only.
What happens when AI cost-control efforts fail? Almost three-quarters of those polled said addressing the issue takes at least several business days, while a third said the delay can be up to several weeks. Just 7% said the organization could typically pause or shut down the related AI usage automatically or near real time, while 19% said they do so the same business day.
Survey participants were also asked to identify the factors limiting their ability to monitor AI spending. The leading response, cited by 40% of the finance chiefs, was a lack of cost governance, processes or policies. Next, at 38% was “limited timely insight into token usage and unit costs,” followed by 36% for “limited standards for models, prompts, data retrieval, workflow design or usage standards.”
Deloitte said it was interesting that 27% of respondents pointed to contracts as a limiting factor, with inflexible terms for pricing, commitments or data egress.
“Some AI vendors are actively introducing restrictions, gating mechanisms and tollgating fees, which are charges imposed for accessing data that enterprises previously considered their own,” the report said. “CFOs, therefore, may want to consider how to leverage their relationships with external partners to negotiate more favorable contract terms.”
