CFOs chase AI results through cost cuts as ROI doubts mount.
UK CFOs betting on AI for cost cuts.
A prominent new survey suggests that many finance chiefs see headcount reduction as the most tangible short-term payoff from AI. Deloitte’s latest UK CFO Survey, released on 14 October, found 52% of CFOs at large British companies expect to reduce their workforce in the next 12 months as a consequence of deploying AI ([1]). The poll of 66 finance leaders indicates that while artificial intelligence is increasingly being adopted, it is still in its early stages at most firms ([2]). More than two-thirds (67%) of respondents said their organisations have so far achieved only about a quarter of the AI-driven changes they ultimately anticipate ([3]).
These findings highlight that even as CFOs talk up AI’s potential, many have yet to see significant bottom-line impact from these investments. Only around one-third of the finance chiefs surveyed believe AI will substantially boost their profit margins in the next year ([4]). In fact, more CFOs expect AI to reduce staff counts than expect it to raise margins, suggesting the technology is viewed chiefly as a cost-cutting tool rather than a proven profit driver ([5]). This underscores that, for now, the clearest "ROI" from AI projects is being sought in operational efficiencies and labour cost savings, while revenue and profit uplifts remain largely aspirational.
Finance leaders clamp down on AI spending.
New data from corporate banking and software surveys reveal that unanticipated AI costs are pushing CFOs to reassert control over budgets. In the United States, a U.S. Bank report based on 1,000 finance leaders in August found 51% said their spending on AI tools had exceeded what was budgeted in the past year ([1]). Separately, Pigment’s Q3 2026 global CFO Index reported that 83% of organisations saw consumption-based AI costs come in higher than planned, with average overrun of 27%, and only 13% staying on budget ([2]) ([3]). A key culprit is the unpredictability of usage-based pricing for generative AI models: in one global survey, 20% of companies said these AI operating costs (such as token fees) were already constraining their use of the technology ([4]).
In response, finance leaders are becoming more strict about how AI initiatives are funded. Companies are moving AI budgets under central oversight and introducing “stage-gate” funding that releases money only when projects meet set performance targets ([5]). Earlier this year, Uber’s CTO revealed the company had burnt through its entire 2026 budget for AI coding tools by April ([6]). Similarly, Amazon disclosed one internal AI project that ran more than 800% over its cost estimate before it was caught ([7]). Those and other high-profile overruns have led CFOs and CIOs at firms from Cisco to major banks to impose usage caps and cut access to non-essential AI tools ([8]). The goal is to prevent cost surprises and ensure each new AI project demonstrates real business value before further capital is committed.
Trillions in tech investment hang on AI payoffs.
The ROI question looms not just for individual companies but for the tech sector as a whole. On 14 October, Gartner warned that more than $3.5 trillion of planned global AI infrastructure spending by 2030 could be “exposed to impairment”, in other words, at risk of being written down, if industry players fail to earn adequate returns on these massive investments ([1]). The analyst firm calculates that to justify the capital outlays on new AI-optimised data centres and chips, providers would need to generate roughly $8.5 trillion in cumulative revenue by 2030, far above what current forecasts predict ([2]). Without dramatically higher returns, Gartner expects the bulk of smaller “neocloud” AI infrastructure firms will fold or be absorbed by larger players by the decade’s end, as even today’s hyperscalers face pressure to deliver a 12% return on invested capital amid rapidly depreciating AI assets ([3]).
For enterprise technology buyers and investors, the implication is that the era of carefree AI spending is drawing to a close. The prospect of write-offs at the industry level is likely to translate into more scrutiny on value delivered at the company level. Boards and finance chiefs are now asking tougher questions about which AI projects truly move the dial on revenue, margin or efficiency, and pulling back funding when the answers are unconvincing.