Who owns AI outcomes in a company?

Ownership of AI outcomes means one named person is accountable for the result and controls the budget that produces it. Both, in the same pair of hands. If the outcome sits with one person and the spend sits with another, nobody owns it, and what you have is a committee with a purchase order.

In most companies between 50 and 10,000 people, that named person is one of four: the CEO, the CFO, the CIO or CTO, or the business-unit lead whose numbers the AI is meant to move. There is no universally correct answer among the four. There is a correct answer for your business, and the failure mode is not choosing the wrong one. It is choosing none, which is what the data now shows happening at speed.

The ownership vacuum is measurable, and it is getting worse

Open Future Forum’s August 2026 finance data caught something you rarely get to see in a single dataset: ownership disappearing in real time. Business-unit sign-off on AI purchases fell from 18% to 8% inside a single month, while “no single owner yet” doubled to 14%. Purchases did not stop. Approval simply stopped landing anywhere.

The most useful line in that dataset is the one about perspective. Not one CEO or founder respondent reported an unowned AI purchase at their company. One level down, 29% of technology respondents did. Same companies, same spend, two irreconcilable readings of who is holding it. If you are the CEO and you would have answered “of course everything is owned,” that gap is the finding, not the noise.

The consequence shows up in the ROI numbers. KPMG’s Global AI Pulse in June 2026 surveyed more than 2,145 C-suite and senior leaders across 20 countries at organisations above $50m revenue and found only 7% report establishing AI ROI, while 42% have only partial visibility into how their AI spending accumulates. Those two numbers belong together. You cannot establish a return on spend you cannot fully see, and you cannot see spend that nobody is required to account for.

The four candidate owners, and what each one actually gets wrong

The CEO. Owns it well when AI is genuinely reshaping how the company makes money, and badly when “the CEO owns AI” is a way of saying nobody has been given time to do the work. The CEO can hold the outcome. The CEO cannot hold the weekly detail of whether the data pipeline is fit to run on. Teneo’s May 2026 research found 65% of CEOs report misalignment with their CFO on AI’s long-term value, and nearly three in four say short-term ROI demands undermine long-term innovation. A CEO who owns AI outcomes without settling that argument with the CFO first owns a dispute, not a programme.

The CFO. The strongest candidate on the measurement side and the weakest on adoption. CloudZero’s survey of 260 finance executives, including 135 CFOs, found 87% say they need to tie AI spend to business outcomes within the year, and only 22% can do that today. A CFO who owns outcomes will get the spend visible, which is more than most companies manage. But finance can enforce a payback window it has no power to shorten, because shortening it means changing how a team works.

The CIO or CTO. The default in most mid-market companies, and the source of the most predictable failure. Technology leaders get handed the outcome without the authority to change the process the outcome depends on. They also carry the vendor decisions. CIO.com reported that three-quarters of CIOs surveyed said they had remorse over at least one major AI vendor or platform selection made in the past 18 months. Buying under a deadline you did not set produces exactly that. If your CIO owns AI outcomes, check whether they own the process redesign budget too. Usually they do not.

The business-unit lead. The best owner when the use case is narrow and the numbers are theirs. The operations director who owns the cost per order should own the AI that is meant to reduce it. This is the ownership model that Open Future Forum recorded collapsing from 18% to 8%, which is worth pausing on, because it is the one most likely to work. When business-unit sign-off drops, purchasing has moved to whoever can get it through procurement fastest.

The one-page test for whether ownership exists

Run this on your own business now. It takes a few minutes and it does not require anyone else in the room.

1. Name the person. Out loud, one name, no job titles and no “well, it depends on the project.” If you produce a committee, a steering group, or two names with an “and” between them, ownership does not exist. Write the name down.

2. Ask whether that person controls the budget line. Not influences it. Controls it. If they need someone else’s approval to move £40,000 from a licence to a data cleanup, they hold the outcome without the means. That is the most common shape of fake ownership in the mid-market.

3. Ask what number they are accountable for. “Improved efficiency” is not a number. “Cost per invoice processed, currently £4.10, target £2.60 by March” is a number. If no baseline exists, the owner cannot be held to anything, and neither can they defend themselves when the board asks.

4. Ask when they last reported on it, and to whom. Ownership that reports nowhere on no schedule is not ownership. A named owner presenting a number to the board every quarter is the whole mechanism.

5. Ask them the same question one level down. This is the step people skip, and it is the one Open Future Forum’s data argues for hardest. Ask the person two rungs below your named owner who owns AI outcomes. If they name someone else, or nobody, you have discovered the 29%.

6. Check whether the owner can stop something. Real ownership includes the authority to kill a project. If your named owner cannot cancel a signed vendor contract, they are administering AI, not owning it.

7. Ask what happens if the number is missed. Not punishment. Consequence. If nothing happens, you have distributed the credit and nobody carries the risk, which is the structural condition behind why AI projects fail in the patterns I have watched repeat for 25 years.

Seven questions. If you answered all seven cleanly, you are in better shape than most of the KPMG sample. If you stumbled on two or more, the honest position is that AI spend in your business is currently unowned, and the ROI conversation with your board is going to be difficult for reasons that have nothing to do with the technology.

Do you need a chief AI officer?

Usually not, and the question is often a way of avoiding the harder one.

A new title creates a new owner only if it comes with budget authority and a number. Most chief AI officer appointments in companies of 200 to 2,000 people arrive with neither. They get a mandate, a slide, and the requirement to persuade four other executives to do things differently. That is a coordination role, and coordination roles cannot be held accountable for outcomes.

If you are considering the appointment, test it against step two of the list above. Will this person control the budget for data remediation, process documentation, licences and training, as one line? If yes, you have created an owner. If it is a coordination role reporting into the CIO with an influence remit, you have added a person to the meeting.

The cheaper move is to give an existing executive the outcome, the budget and the reporting line, and to take something else off their plate to make room. Ownership is a reallocation of attention, not a hire.

Ownership does not fix the thing underneath

One warning, because I have seen naming an owner used as the whole intervention. It is not. An owner with clear authority over a business built on broken foundations will fail more visibly and more quickly than a committee would have, which is at least useful information.

Gartner expects organisations to abandon 60% of AI projects through 2026 for want of AI-ready data, and Informatica’s CDO Insights 2025 survey of 600 global data leaders found 43% naming data quality, completeness and readiness as the top obstacle stopping AI reaching production. Named ownership will surface those problems earlier. It will not solve them. What ownership does is make sure the person who discovers the data is unusable is the same person who can authorise fixing it, which is why dirty data is the cheapest AI mistake to fix first.

Where to check your own position

Ownership sits inside two of the six things worth assessing before you spend: Strategy, which covers whether there is a defined outcome and a plan attached to it, and Governance, which covers who is accountable and what oversight exists. The other four (Data, Process, People, Technology) are what the owner will actually have to work with.

The free AI readiness audit is 30 questions across those six pillars, takes 7 minutes, and gives you a score out of 120, a band per pillar and a short synopsis. Every answer is self-reported, which is the point: the Governance and Strategy questions ask you to state plainly who is responsible, and most people find that the act of answering is the diagnosis. There is no sales call at the end unless you ask for one.

The uncomfortable version of the ownership test is to have three executives take it separately and compare the Governance bands. If they diverge, you have found what Open Future Forum found. Same company, same spend, different answers about who is holding it.

Start with the name. Everything else is downstream of the name.