AI spend has no owner
the 79% overrun paradox

The most expensive sentence in enterprise AI right now is “we are jointly accountable for AI spend.” A DoiT-commissioned survey by Sapio Research of 500 US and UK finance leaders produces a counterintuitive result: 79% of organizations overspent on AI in the past 12 months, and the organizations that rate their FinOps practices most mature overran the most — 89% over budget, with a mean overspend of 30.9%. Governance has not failed. Governance is working: mature organizations simply see overruns earlier than immature ones. What has actually failed is ownership — technology and finance each own half, which in practice means nobody owns it.

79% overspent — and the most mature overran the most

DoiT commissioned independent research agency Sapio Research to survey 500 finance leaders at US and UK organizations with 1,000 or more employees (February 2026, plus or minus 4.4pp): 79% experienced AI-related cost overruns in the past 12 months (source: DoiT — https://www.doit.com/blog/ai-spending-survey ). The sharpest numbers sit inside the maturity split: organizations self-assessing as “very mature or leading edge” on FinOps overran at a 89% rate with a 30.9% mean overspend, while early-stage FinOps organizations show a 69% overrun rate and a 16.1% mean.

Read quickly, that looks like an argument against investing in governance. It is the opposite. Mature organizations run larger, more complex AI programs and have the instrumentation to surface overruns that less mature organizations simply never catch. The overruns at less mature organizations are not smaller because spending is better controlled; they are smaller on paper because much of the spending is not being measured. FinOps maturity surfaces problems. It does not prevent them.

One company, two maturities: C-suite 93%, managers 60%

The most structural finding in the survey is the perception gap. C-suite respondents rate their organization’s FinOps maturity at 93% mature or better; manager-level respondents put that figure at 60% — a 33-point gap, both describing the same organization. This is not a disagreement about facts. It is a structural visibility problem. Leadership sees investment ambition, governance intent, and the frameworks presented at the board level. Operations sees the projects without named cost owners, the attribution systems that were scoped but never built, and the overruns that land on real budgets rather than in strategy decks.

The ownership vacuum: Technology 55%, Finance 53%

Combine the spending gap with the trust gap and the root cause is ownership. The survey found accountability for AI spend split almost evenly between technology leadership at 55% and finance at 53%, with no clear single owner at the operational level. When asked who holds final authority in a spending conflict, C-suite respondents named the CEO three times more often than managers did. At the level where spending actually happens, the question of who controls the AI budget often has no settled answer.

Shared ownership, in practice, tends to mean no ownership. And AI spend is currently one of the most expensive places for that ambiguity to live — which is why the sentence “we are jointly accountable” costs so much.

The 12-month ROI clock

83% of finance leaders expect clear, quantifiable AI returns within 12 months; 81% are already adjusting their AI spend or plan to within the year. But only 15% can calculate AI ROI without significant bottlenecks. The leading barriers: the pace of technological change at 40%, finance and engineering defining success differently at 37%, and a lack of clear financial attribution at 36%.

The definition gap deserves to be separated from the other two: it is the one barrier that does not require new tooling to fix. Finance and engineering must agree on what AI success means before any measurement system gets built. C-suite respondents feel this barrier most acutely (43% naming it a significant obstacle versus 33% of managers). Whoever defines the metric first holds the budget conversation.

This is a pattern, not an outlier: IDC FERS corroboration

IDC’s FERS Survey Wave 4 (July 2026) measured the same pattern: 61.8% of organizations rate their own cost governance as “defined” or “optimizing,” yet only 45.4% have real-time dashboards tracking token consumption and cost by workflow; the two thirds that self-rate governance as mature are the same two thirds that missed budget (source: IDC — https://www.idc.com/resource-center/blog/the-agent-economy-is-scaling-faster-than-it-can-be-metered/ ). We covered that research in depth on August 31 (see “Agent Economy Outruns Its Meters: IDC on Cost Governance”).

Two independent surveys, different respondent bases (finance leaders versus IT decision-makers), published months apart, arriving at the same conclusion: self-assessment of AI cost governance and actual outcomes have stopped tracking each other. This is not an anomaly. It is an industry-level pattern.

Our judgment

First, the correlation between governance maturity and overruns is not evidence that governance failed; it is evidence that visibility is working. The case for buying governance is “see overruns sooner,” not “never overrun” — the data does not support selling governance as prevention. In the boardroom, the honest framing is: governance lets you see the problem while you can still act.

Second, ownership before tooling. Before buying another FinOps or governance platform, answer one question: which role is accountable for each AI workflow’s spend? IDC’s own framework starts with naming a per-workflow accountable owner — naming an owner is cheaper than buying tools, and it is the precondition for any dashboard to mean anything.

Third, the 12-month clock is an opportunity, not a threat. The finance-engineering definition gap (37%) is the only barrier fixable without new tooling; 83% demanding returns against 15% who can measure them means the side that defines the metric first is deciding which projects survive the next budget cycle.

Fourth, for OOMeta, the ownership vacuum and the visibility gap point at the same thing: enterprises do not lack more budget, they lack a per-workflow, independently verifiable spend ledger. Measurement precedes governance; ownership precedes tooling.

Action list

1) Name a single accountable owner for agent spend, per workflow, before the next budget cycle — not after the next overrun.

2) Replace the monthly invoice review with real-time, per-workflow cost visibility. A once-a-month snapshot cannot manage a cost surface that moves by the hour.

3) Have finance and engineering agree on what AI success means before purchasing measurement tooling — the cheapest cut at the 37% barrier.

4) Cap any single vendor’s share of inference spend, generally below 40% (IDC’s suggestion), and keep a second option warm as pricing and terms keep shifting quarter to quarter.