AI Strategy & Transformation8 min read·By , Independent AI Advisor & Fractional AI Lead·

The AI team should not own AI value

A central AI team can own the platform, the standards and the risk controls. The value claim needs a business owner whose own numbers move.

BCG's July 2026 piece on how CEOs scale AI value reports that only 14% of CEOs clearly define the P&L impact of all their AI initiatives, and that ownership often stays spread across functions. A September 2026 BCG release adds that more than 80% of AI spending now sits outside enterprise IT. BCG does not say when either survey was fielded, so read these as directional signals, not a snapshot of today.

Put together, they describe a familiar shape. The money is moving into the business, while the definition of what it should return often is not. In that gap, the central AI team tends to become the default owner of "AI value", simply because it is the most visible group with AI in its name.

I think that is a mistake. Not because AI teams are weak, but because they are the wrong place to hold the value claim.

Two different things are called ownership

Most bundle two kinds of accountability.

Capability and risk accountability covers the platform, methods, model and vendor choices, security, data access and standards. It is cross-cutting work and benefits from being done once, by specialists, for everyone.

Value accountability is a promise that a particular workflow will produce more revenue, lower cost, faster cycle times or better quality than today, and that someone will stand behind the number. It is local work. It depends on the incentives, staffing, and customer behavior of one part of the business.

An AI team can deliver the first group flawlessly and still have no power over the second. It cannot change a sales team's targets, reassign a support queue, rewrite a policy, or decide that saved hours become a smaller team rather than a longer to-do list. Yet when value is reported to the executive committee, the central team is often the one presenting it.

This is a different question from the one in my ten-minute business case test, which asks whether one use case has a path to captured value. This essay asks how a whole portfolio should assign the claim, and what the central team is left accountable for.

What the evidence supports, and what it does not

No research I found tests whether business ownership causes better AI results. The surveys point in a direction, and I want to state it carefully.

McKinsey's August 2026 survey, fielded between 4 May and 8 June 2026 with 1,719 respondents, found that 37% attribute some EBIT impact to AI, essentially unchanged from the year before. About 6% are high performers, defined as attributing at least 5% of EBIT to AI and reporting significant value, which is only 92 respondents. Nearly three-quarters of them say they fundamentally redesigned workflows because of AI, against about a quarter of other respondents. The data are self-reported and correlational, and the high-performer group is small.

Even so, is not something a central team can do on a function's behalf. Whoever runs the workflow has to change it. BCG's CEO piece points the same way: its highest-impact companies are described as far more likely to redesign workflows end to end and twice as likely to restructure accountability, and it recommends measuring CXOs and P&L leaders on outcomes rather than activity. That is a consultancy's reading of its own survey, with no sample size published.

My claim is narrower than either source. If an initiative's value depends on changing how a business unit works, the person who controls that unit should own the claim that it will pay off. Nothing in these surveys contradicts that.

A two-name test

This test is my own proposal. For each AI initiative on your list, write down two names.

The first name belongs to the person whose budget, headcount plan or target will change if the initiative works. Not the sponsor who liked the demo, but the person who will be asked next quarter to commit to the new number.

The second name belongs to the person who absorbs the cost if it fails: the wasted integration effort, the disrupted team, the customers who got a worse answer in the meantime.

If both names belong to the AI team, you do not have a business case with an owner. You have a technology project reporting to itself. If the first name is a business leader and the second is not, that leader holds an option with no downside, and their enthusiasm will be hard to calibrate. The test is deliberately crude. It needs no framework, only a willingness to leave a cell empty and see who objects.

An illustration, not a case: a support organization an AI drafting tool. The AI team reports a 20% drop in handling time. The head of support has not changed staffing plans or service targets. Whose number moved? Until the head of support can say what they will do with the freed capacity, the 20% is a capability result, not a value claim.

What the central team should own instead

Moving the claim does not make the AI team less important. It makes the job clearer, and harder to hide behind.

I would hold the central team accountable for four things: evaluation that is honest and repeatable, a platform and controls that are safe to build on, reusable components that actually get reused, and a baseline for every initiative before it starts. The baseline matters most. It turns "the pilot worked" into a statement someone can later be held to, and it is cheap to collect before launch and often impossible afterwards.

The team should also be allowed to say that a claimed value is not supported by the data. That is only possible if the claim belongs to someone else. When one team owns both the system and the result, it is grading its own work.

A value-ownership ledger

Also my proposal: one line per initiative, reviewed on a fixed cadence.

FieldQuestion it answers
Value ownerWhich business leader signs the claim?
BaselineWhat does the current process produce today, measured how?
ClaimWhat change, by when, against what cost?
Capability ownerWho is accountable for the system working as specified?
Stop conditionWhat evidence ends the initiative, and who decides?

Its value is plain: two roles that are usually blurred appear in separate columns.

The best objection

The strongest counter-argument is that business units are bad at this. Many lack the technical judgment to evaluate models, and decentralized ownership can produce duplicated tools, uneven controls and unmanaged risk. McKinsey's March 2025 report, based on a survey fielded in July 2024, found that risk and compliance and data governance were often fully centralized, while tech talent and adoption of AI solutions most often followed a hybrid or partially centralized model. That is older evidence about structure, not about results, but it supports the instinct that some things should not be left to each function.

I agree, and the proposal does not decentralize those things. Risk, standards and the platform stay central. Only the value claim moves. A business owner who depends on a central team for evaluation and controls is a feature of this design.

A second objection is practical: in some companies nobody in the business wants the number. That is useful information. An initiative that no business leader will put their name against has probably not yet found the problem it solves.

What I would do on Monday

Take the current AI portfolio and fill in the first column of the ledger. Count the initiatives whose value owner is a business leader who has agreed to it, as opposed to a sponsor who has been informed. Then count the ones where the only available name is the AI team's.

The second number is your exposure. Those initiatives need a business owner, a smaller scope that one will accept, or an honest relabeling as capability-building work that is not yet expected to return anything.

Where this connects

Sources

How CEOs scale AI value

Boston Consulting Group · 2026-07-22

AI is starting to pay off: companies generate value (press release)

Boston Consulting Group · 2026-09-30

The state of AI in 2026: On the road to ROI

McKinsey & Company · 2026-08-25

The state of AI: How organizations are rewiring to capture value

McKinsey & Company · 2025-03-12

The two-name test, the ledger, the support illustration and the argument for separating the value claim from capability accountability are my own synthesis, not findings from these sources. The surveys are self-reported, and BCG does not publish sample size or fielding dates for the CEO survey cited here.

Get new essays as they publish.

One email per essay. No noise between.

Berk Bayri

Independent AI Advisor & Fractional AI Lead

I help leadership teams decide where AI belongs, test it before it scales and build the teams that run it, drawing on 25+ years of building digital products, experiences and capabilities.