Owned by Finance: CFOs Now Hold the AI Budget, and Sellers Have to Prove the Return

Every revenue team selling AI this year has learned to find the champion, the technical evaluator and the executive sponsor. The research published over the past month says the person who now decides whether an AI product is bought, expanded or quietly throttled sits in none of those seats. It is the CFO, who owns the AI budget, has already watched it overrun, and still cannot tell the board what it returned. That is a different buyer from the one most AI sales motions were built for, and it wants something most pitches do not offer: a way to control the spend, and evidence it can verify on its own.

Finance Took the Budget, Then Found the Overrun

Datarails' 2026 CFO Sentiments Survey, fielded in July among 270 CFOs and finance leaders at U.S. companies with more than 1,000 employees and at least $100 million in revenue, found 84% now own their organization's AI budget, while 13% say no one clearly does. Ownership arrived alongside overspend: 32% said their organization exceeded its AI budget by at least 10% in the past 12 months. Appetite has not faded. 53% plan to expand AI licenses over the next year, and only 7% plan to cut or consolidate. What has changed is who signs, and what they have already seen.

Deloitte's third-quarter 2026 CFO Signals survey of 200 CFOs at North American companies with at least $1 billion in revenue shows how formal that role has become. Nearly 80% say finance has a formal role in AI decisions: 41% serve as required approvers and 38% define or administer the evaluation framework. Only 20% limit themselves to cost and budget guidance. For a seller, that means the evaluation criteria for an AI deal are increasingly written by finance, not by the team that ran the pilot.

The Controls Lag Behind the Commitments

The same Deloitte data explains why CFOs arrive at AI deals cautious. 53% review AI usage costs primarily in hindsight, through vendor bills, periodic reports or manual analysis, and only 20% use advanced AI-specific controls such as workload-level cost attribution or AI gateways. When usage runs past approved limits, 74% typically need several business days to several weeks to pause or shut it down. Just 19% act the same business day.

Asked what stands in the way, 40% of the CFOs cite a lack of AI cost governance and policy, 38% point to limited timely insight into token usage and unit costs, and 27% say contract terms, including pricing, commitments or data egress, limit their flexibility. Each of those is something a vendor shapes. A product that shows the buyer what it costs while it is running, and a contract that lets finance adjust when it overruns, removes a reason to say no. One that delivers a surprise invoice at month end hands finance a reason to cap the rollout.

Trust in the output is the other half of the problem. Datarails found 75% of CFOs cite a lack of auditability and 71% cite accuracy and hallucinations as reasons they hesitate to fully trust AI. A CFO who cannot audit the work is unlikely to accept a vendor's estimate of what that work is worth.

Spending More, Proving Less

The pressure runs all the way to the top. Gartner's 2027 CEO Survey of 203 CEOs and senior executives, fielded in April and May and released in late October, found 44% struggle to demonstrate AI-driven financial returns to stakeholders, even as 75% plan to increase technology investment. “The AI value problem will not be solved by simply adding more AI to existing ways of working,” said David Furlonger, Distinguished VP Analyst at Gartner.

That gap lands on the CFO's desk, and from there on the vendor's. IDC reported on October 26 that CFOs or their delegates approve 78% of technology purchases, and that they want evidence that is independent of the vendor. Buyers are also building their own math. In G2's 2026 Buyer Behavior Report, comparing total cost of ownership tied for the top use buyers name for AI agents in software purchasing, at 51%, and 87% said they are more likely to buy from a vendor offering transparent AI than from a cheaper black-box competitor. The ROI slide built by the seller is now checked against a model built by the buyer.

The CRO Playbook: Sell to the Budget Owner

• Qualify finance in, not around. With 41% of large-company CFOs acting as required approvers, identify who in finance signs the AI line and what framework they use before the pilot starts.
• Sell the cost controls with the product. Usage dashboards, spend alerts and caps answer the 38% of CFOs who lack timely insight into unit costs. Put them in the demo, not the appendix.
• Write contracts finance can adjust. 27% of CFOs say contract terms limit their flexibility. Clear consumption tiers and adjustment rights in the quote remove an objection before procurement raises it.
• Bring evidence the buyer can verify. Finance approves 78% of technology purchases and wants proof that does not come from the vendor. Lead with customer references, third-party benchmarks and a cost model the buyer can rerun.
• Agree on the return metric at signature. With 44% of CEOs unable to show AI returns, set one measurable financial outcome with the buyer and report against it before the next budget cycle.
• Teach reps to speak finance. Unit cost, payback period and budget variance are now the language of the AI deal. Reps who cannot discuss them will lose the room to someone who can.

None of this signals a retreat from AI spending. CFOs plan to buy more, and CEOs plan to invest more. But the budget has a new owner, and that owner has already seen it run over. The vendors that win the next round of AI deals will be the ones that make finance's job easier: spend it can see, terms it can change, and a return it can prove without taking the seller's word.