CFO surveys reveal rising AI spend, murky returns.
Most AI projects’ ROI unclear, Gartner survey finds.
A fresh survey by Gartner, released on 12 October, suggests that most finance leaders are struggling to discern any clear return on investment from their AI projects. In a poll of 160 senior finance-function executives, 55% reported positive overall returns from their 2025 AI initiatives, but evaluation at the use-case level told a different story ([1]). The majority, 57%, of the finance AI use cases surveyed had “unclear returns”, with just 18% rated as having a positive ROI and 19% actually in the red ([2]). Even among the minority of cases where returns were measurable, outcomes were evenly split between successes and failures ([3]). Gartner noted that a small number of transformational AI wins appear to offset a higher volume of initiatives that fail to exceed their costs ([4]).
This imbalance puts CFOs in a tricky position: they are expected to fund innovation while knowing much of the spend may never pay back ([5]). Gartner’s recommendation is for finance leaders to demand more discipline around AI investments, setting clear strategic objectives and metrics in advance, and establishing performance targets and timelines to decide when to scale a project or shut it down if it underperforms ([6]). The survey’s findings underscore the need for CFOs to shift the AI conversation from vague promises to tangible, measured business outcomes.
Surprise costs drive CFOs to tighten AI budgets.
Rising and unpredictable operating costs are forcing many companies to re-examine how they fund AI. In the third quarter, 83% of organisations said their consumption-based AI expenses, such as pay-per-use cloud model fees, exceeded initial forecasts ([1]). A U.S. Bank study in September further found more than half of finance leaders went over budget on AI spending in the past year ([2]). Deloitte’s own research warns that 60% of CFOs expect AI costs and complexity to rise substantially through 2027 ([3]), and one in four firms has already delayed or cancelled AI projects due to unexpected expenses ([4]).
In response, CFOs are pulling AI initiatives under tighter financial control. Organisations known for aggressive AI adoption, including Uber, Amazon and Cisco, have imposed new spending caps or restrictions to prevent runaway costs, with some budgets exhausted months ahead of schedule ([5]). Finance leaders are also introducing “stage-gate” funding processes that release money in phases only when projects meet defined performance milestones ([6]). Speaking at a Fortune event this week, Block CFO Amrita Ahuja predicted companies may even start adding AI “agents” to their official organisational charts as digital workers, to clarify accountability and justify the technology’s costs ([7]). The overriding aim is to ensure AI spending delivers real financial value, and to cut off projects that don’t.
Investments still rising, but proof of ROI now required.
Despite the challenges, finance chiefs are not giving up on AI, they are just becoming more exacting. A new survey by cloud cost platform CloudZero indicates that 87% of senior finance leaders plan to boost AI budgets for the next fiscal year, with nearly half expecting to increase spending by over 20% ([1]). Yet in the same study 61% of respondents admitted their current AI spending is running ahead of any demonstrable results ([2]). Boards are accordingly tightening oversight: only 26% of corporate boards still advocate a “blank cheque” approach to AI, while two-thirds now make new funding conditional on proven returns ([3]). In fact, 22% of boards won’t approve any additional AI investment until they see real ROI, leaving many finance teams having to answer tough questions they currently can’t ([4]).
If anything, the gulf between expectations and outcomes is motivating a sharper focus on value. The OneStream “Forward Finance” index, released on 13 October, found that while 74% of CFOs surveyed are already using AI to reduce manual work, only 14% believe their AI initiatives have delivered “enterprise-level” value so far ([5]). In practice this means that barely one in seven finance chiefs can point to AI driving significant business performance gains. Those finance teams that do master ROI measurement are pulling ahead: a related study found three-quarters of CFOs unable to demonstrate AI returns have held back investment, whereas companies that can quantify gains are continuing to invest heavily in automation and analytics ([6]). In 2023 and 2024 many CFOs funded AI based on faith in its promise; in late 2026, that era is decisively over. Finance leaders are still willing to invest in AI, but only if they can see its impact on costs, revenue or productivity in black and white.