In 47% of companies, the CEO approves AI spend. In almost none of them does the CEO carry the burden of proving it worked. That job lands on finance, which reported the number as its top blocker to spending more.
The figure comes from The Executive AI Leverage Report, a survey of 421 executives that Murray Newlands ran across seven finance, security, growth, and founder events for the Open Future Forum (July 2026). It is a single self-published source, and the sample is event-based rather than random, so read the percentages as a strong signal from one room full of operators, not as market-wide truth. Even with that caveat, the shape of the finding is worth sitting with, because it describes a structural mismatch that most AI strategy decks never name.
Who Signs Is Not Who Answers For It
The survey asked who holds final approval authority on AI investment. The answers: 47% CEO, 26% CFO or finance, 21% CIO or CTO. Now ask a second question the report makes unavoidable. Who gets asked, six months later, whether the money produced anything?
That answer is finance, every time. And finance knows it. When the survey asked what blocks more AI spending, 53% of finance respondents named proving ROI as the primary obstacle. The pattern is a chief executive with the pen and a finance function with the burden of proof, and the two are usually different people reporting on different clocks.
This is not the familiar “measure your ROI” lecture that every vendor already delivers. The measurement problem is downstream of an ownership problem. When approval authority and return accountability sit in different chairs, the person who said yes has already moved on to the next initiative by the time the question of value comes due. Finance inherits a commitment it did not size and is then asked to defend it. The controls you actually need are not better dashboards. They are a decision record that ties each approval to a named owner and a return date before the check clears.
The Proof Window Is Six Months
Finance is not asking for patience. It is asking for evidence, fast. In the survey, 62% of finance respondents expect measurable return within six months, and 79% expect it within a year. That is the real budget cycle for AI now, and it is short.
Six months is not enough time to run a leisurely pilot, gather anecdotes, and reconvene next fiscal year. It is barely enough time to instrument a workflow, establish a baseline, ship a change, and read the delta against that baseline. Any AI initiative launched today without a measurement plan attached is already burning a third of its proof window on setup. The teams that win this cycle are the ones that treat the baseline as the first deliverable, not the retrospective.
The reporting clock also explains why so many pilots read as failures. A pilot with no pre-agreed baseline cannot produce a number finance will accept in month six. The work may have created value. Nobody can prove it in the window, so it counts as a miss.
One in Six Is Now Funding AI Out of Headcount
Here is the finding that changes the stakes. 17% of finance leaders, one in six, said they now fund AI at least partly from headcount budgets. That is not incremental innovation money. That is payroll being redirected into software on the bet that the software covers the work.
Substitution funding raises the accountability temperature sharply. When AI is paid for out of a discretionary innovation line, a disappointing return is a write-off. When it is paid for out of headcount, a disappointing return is a hole in the org chart, roles left unbackfilled against a tool that underdelivered. The proof burden stops being a finance reporting exercise and becomes an operational risk with names attached.
This is the surface where governance earns its keep. A company drawing AI spend from payroll needs to know, per initiative, what human capacity was traded away and whether the tool has closed that distance yet. If nobody is tracking the substitution explicitly, the shortfall shows up as attrition, overload, and missed work months after the funding decision, when it is hardest to trace back to the AI bet that caused it.
The Security Line Nobody Funded
The same authority-versus-accountability split appears in security, one function over. 56% of security leaders in the survey said securing AI agents and their access is now a top priority. Only 31% have a dedicated AI-security budget. More than half of the people responsible for the risk are working without a line item to address it.
The pattern is consistent across the report. The organization has assigned the worry to one group and the money to another, and the two have not been introduced. Approval sits upstream of the people who carry the consequence, whether that consequence is an unproven return or an unfunded exposure.
The Pricing Signal Underneath
One more data point reframes the rest. Among founders in the survey, 50% price on usage, 25% per seat, and 18% on outcomes. The market is drifting from selling access toward selling consumption and, increasingly, results. Vendors are starting to accept payment tied to what the software actually does.
That drift is a gift to any buyer willing to use it. If a growing share of vendors will price against outcomes, the accountability burden that finance carries can be pushed back toward the party best positioned to prove value, the vendor selling the capability. The buyers who negotiate outcome terms turn the proof problem into a shared one instead of shouldering it alone.
Do This Now
Before approving the next AI investment, write three things on the approval itself, not in a separate deck. The named owner accountable for the return, not the executive who signed. The baseline metric and the date it was captured, before any spend. The proof date, inside the six-month window finance already expects. If the money is coming out of a headcount budget, add a fourth line: the specific human capacity being traded and the checkpoint to confirm the tool has covered it.
That single discipline, attaching accountability to authority at the moment of approval, closes the distance the survey exposes. Everything else, the dashboards, the ROI models, the security budget, is downstream of getting that one record right.
This analysis synthesizes The Executive AI Leverage Report (Open Future Forum, July 2026).
Victorino Group helps enterprises tie AI approval authority to return accountability before the check clears. Let’s talk.
All articles on The Thinking Wire are written with the assistance of Anthropic's Opus LLM. Each piece goes through multi-agent research to verify facts and surface contradictions, followed by human review and approval before publication. If you find any inaccurate information or wish to contact our editorial team, please reach out at editorial@victorinollc.com . About The Thinking Wire →
If this resonates, let's talk
We help companies implement AI without losing control.
Schedule a Conversation