AI spending is accelerating faster than enterprises can prove financial value. The missing costs include correction, rework, verification, and the human judgment required to make AI output usable.

AI return on investment is becoming harder to defend because spending is accelerating faster than enterprises can prove financial value. The problem is not that AI lacks utility. The problem is that many ROI models count adoption, output volume, generation speed, and vendor-promised productivity while excluding correction, rework, verification, infrastructure, and the human judgment required to make AI output usable.

The CFO problem with AI ROI is now three problems converging in the same financial model:

  1. Expected savings are underperforming.
  2. AI costs are becoming harder to forecast and control.
  3. Human verification labor remains missing from the calculation.

AI ROI is being squeezed from both sides before the human governance layer is even priced.

AI Spending Is Rising Faster Than Verified Value

Finance leaders are not retreating from AI. They are being asked to approve larger investments while the evidence supporting those investments remains incomplete.

Bain & Company reported in 2026 that 56% of CFOs planned to increase enterprise AI investment by more than 15% during the year. Over the following two years, 83% anticipated increases above 15%, and 42% expected AI investment to rise by more than 30%.

Deployment maturity is not advancing at the same rate. Bain found that only 15% to 25% of surveyed companies had fully scaled AI within finance.

The capital commitment is therefore moving ahead of the operating evidence.

CFOs are being asked to fund infrastructure, models, agents, integrations, data preparation, and expanding usage before the enterprise can consistently distinguish between AI deployment, employee adoption, productivity, cost reduction, and attributable financial return.

Those are not interchangeable measures.

The Savings Story Is Not Arriving Cleanly

A Bain survey previously reported by Bloomberg found that 40% of companies were showing AI cost reductions of 10% or less, while only 4% globally had reached savings above 30%.

More recent finance research shows the same measurement problem. Boston Consulting Group found that only 45% of surveyed finance executives could quantify the return from their AI investments. Median reported ROI was 10%, compared with the 20% many organizations were targeting.

The missing return is not necessarily proof that AI creates no value. It indicates that companies are having difficulty converting AI activity into measurable financial outcomes.

If AI generates more output but requires more review, the gross speed gain is not the net savings. If AI reduces drafting time but increases senior verification time, the labor did not disappear. It moved into a more expensive layer.

Productivity can rise while margins remain unchanged.

Wall Street May Begin Rewarding AI Spending Discipline

For several years, companies were rewarded for announcing larger AI investments. The market signal was expansion: more infrastructure, more compute, more models, more agents, and more enterprise deployment.

That incentive may be reaching its limit.

As AI capital expenditure rises, investors are asking where the return appears in revenue, operating margins, free cash flow, and durable cost reduction. Spending that once demonstrated strategic ambition may eventually be interpreted as weak capital discipline when the financial return remains unverified.

The first major company to constrain AI spending may not be punished for falling behind. It may be rewarded for demanding evidence before approving the next tranche of investment.

This is not an argument against AI. It is the ordinary discipline applied to every other material capital commitment:

What return was promised?
What return was produced?
What was the full cost of producing it?
What evidence supports the calculation?

AI should not become exempt from those questions because its future potential is substantial.

Unpredictable AI Consumption Is Only One Cost

The Information reported that Anthropic customers, including ServiceNow, had difficulty predicting what they would pay for AI usage. ServiceNow disclosed that it had consumed its full-year budget for Anthropic tools before the year was over.

Companies encouraged employees to use AI as proof of adoption, and some treated high token consumption as a success signal. The bill then became harder to forecast.

Token volume is not ROI.

Deloitte has described AI as a structural change in enterprise cost behavior because token consumption introduces volatility into operating expenses, margins, forecasts, and capital planning. Unlike conventional software licenses, AI costs can change with model selection, prompt length, data retrieval, output volume, agent behavior, and employee usage.

More usage does not automatically mean more savings. More complex usage can mean more expense, more monitoring, more review, and more correction.

But token spending is only the visible expense. The larger CFO problem is the cost that appears elsewhere.

The Hidden Cost of AI ROI: Pricing Correction and Verification

One vendor analysis of AI-assisted software development estimated that only 18 cents of each dollar spent on AI tokens reached shipped production output. The remaining expenditure was attributed to fixing AI-generated defects, rewriting code that lacked sufficient context, and absorbing review delays.

The estimate is specific to AI-assisted software development and should not be treated as a universal enterprise ROI figure. Its cost structure is nevertheless instructive.

The visible AI charge may only be the first cost.

If a tool produces a draft in two minutes but a senior employee spends twenty minutes verifying, correcting, and contextualizing it, the ROI calculation cannot stop at the two-minute output. AI externalizes execution while internalizing judgment. If finance counts the first without measuring the second, the return calculation is incomplete.

This is where the Power User Trap℠ becomes a finance issue.

The power user is the employee through whom AI becomes usable inside real work: learning the failure modes, supplying institutional context, catching plausible errors, and translating machine output into something the organization can rely on.

That work is frequently absorbed into payroll without being attributed to the AI investment that created it.

The CFO Question Has Changed

The CFO question is no longer simply: How much are we spending on AI?

It is also:

What did AI actually remove?
What work did it relocate?
What new work did it create?
What did the organization have to correct before the output became usable?
Did faster production become lower cost, higher revenue, stronger margins, or durable operating capacity?

The full AI cost model includes licenses, tokens, infrastructure, vendor fees, integration, data preparation, usage monitoring, deployment labor, internal review, correction, rework, compliance controls, and the power users whose judgment stabilizes the system without appearing in a dedicated budget line.

The financial question is not whether AI can produce output faster. The financial question is whether the organization can prove that faster output became durable operating value after the full human and technical cost was counted.

The Materiality Question Set

Before the next board meeting or budget cycle, the organization should be able to answer:

  • Which AI investments are being evaluated through adoption or output volume rather than verified operating value?
  • What portion of reported productivity became revenue, margin improvement, cost reduction, or additional operating capacity?
  • If verification, correction, and informal training are priced in, does the AI investment still clear the approval threshold?
  • Where are productivity gains being attributed to AI when the output is being stabilized by unmeasured human judgment?
  • What evidence substantiates AI efficiency claims made to the board or investors beyond output volume, token consumption, or adoption rates?
  • What conditions would cause the organization to reduce, pause, or reallocate AI spending?

CFO AI Investment Advisory

Lozen Advisory’s CFO AI Investment Advisory helps finance leaders evaluate whether AI spending should be funded, restructured, reduced, or paused when the full cost of implementation is measured.