Four surveys, four answers

Different questions

Ask a FinOps lead whether AI returns are being tracked, and you may hear that it is handled. The FinOps Foundation's State of FinOps survey, published in February, found that 98% of its 1,192 respondents now manage AI spend. [1]

Managing the spend and knowing whether it paid off are separate questions. On the second one, this year's surveys do not agree.

KPMG found 7% of senior leaders saying their companies have reached established returns from AI. Bain, surveying more than 100 CFOs, found 31% satisfied with AI outcomes. Protiviti found 35% of finance groups calling themselves highly or moderately effective at measuring AI return. Gartner found that high performers that regularly track AI return report positive returns on 81% of their AI initiatives. [2] [3] [4] [5]

Those are four different bars. Established return, satisfaction, self-rated effectiveness at measurement and a share of initiatives cannot be averaged into one number, and they will keep disagreeing however good the data gets. What they share is underneath. Each one depends on a company being able to say where an AI dollar went and what came back. In the same Gartner survey, low performers did not know the return on 29% of their initiatives, and about 11% of organizations did not know what their own function spent on AI last year. [5]

The missing record

Deloitte's Finance Trends 2027 survey, of 1,434 finance leaders, shows how funding decisions get made. Two-thirds use a process that stresses measurement, such as stage gates or capital approval with a quantified return. But 23% said executive or board mandates drive most of their AI and technology investment decisions, with no measurement process to gauge success, and in North America that was 28%. Asked how they judge AI investments, 36% named time and productivity and 23% named revenue enablement. [6]

Finance is being asked to answer for this. In that survey, 48% of finance leaders said they lead oversight of AI and technology spending and cost controls, outside finance's traditional scope. In a separate Deloitte survey of 200 North American CFOs, the most common answer, at 33%, was that the CISO owns AI governance; 19% of CFOs said they themselves hold the greatest responsibility for it. The FinOps Foundation counts 78% of FinOps teams reporting to the CTO or CIO. The question lands in finance. Much of the data that would answer it may sit in other departments. [6] [7] [1]

Deloitte's own advice points the right way. In October it told CIOs, CFOs and business leaders to keep a value register linking each AI use case to its expected and realized value, and to ask whether each one has "earned the next dollar" (Deloitte, 8 Oct 2026). [8]

Keep the value register

Here is what that register needs, using a hypothetical. A claims team gets approval for a drafting assistant on the condition that it costs no more than $1.10 a claim. The record needs the month's dollars for that team and workflow, taken from the invoice rather than an estimate. It needs the $1.10, written down before the work began, so the bar cannot move later. And it needs the month's actual cost per claim from the claims system, confirmed by someone in finance. If that comes in at $1.34, the bet is losing money, and the record says so before the next budget round.

In many companies, those pieces can sit in different places. The bill is with procurement or the cloud team. The business case is in a slide deck. The outcome is in the claims or ticketing system. Joining them is the work.

One figure in KPMG's global data stands out. Among companies with strong visibility into AI cost, 15% report established returns. Without it, 3%. That is a correlation, and companies that see their costs clearly may differ in other ways too. It is still a reason to think the record is worth keeping. [2]

A practical test before the next budget cycle: take the three largest AI lines and ask whether each had a cost ceiling written down before the work began. Start with the ones that did not.