When AI Starts to Scale, Who Owns the Business Outcome?
–
September 11, 2026
An AI steering meeting can look healthy. Use cases are growing. Adoption is improving. Teams can point to time saved, better service, faster decisions and new capability.
Then somebody asks which business outcomes matter enough to concentrate investment around, and which executive is personally accountable for delivering them.
The answer can be much less tidy than the dashboard.
Early in an AI programme, central coordination makes sense. Organisations need standards, platforms, data access, security, risk controls and a way to help teams experiment without creating chaos. An AI office, technology function or transformation team can provide much of that discipline.
The economics get harder to read once AI starts changing real operating work.
A central team cannot release capacity on behalf of a business unit. It cannot decide which customer journey should change, which process trade-off is acceptable, whether a role should be redesigned, how an incentive should change or whether a local productivity gain is worth scaling. Those decisions sit with leaders who own the operation.
McKinsey’s latest work is useful here.
Its 28 August 2026 article, The new management playbook for AI: How to move faster and create more value, examines 20 companies that McKinsey says have consistently created significant economic value from AI-enabled business transformation.
The technology itself is not the main explanation McKinsey offers. The tools are broadly available. The advantage comes from how organisations apply them to real business problems at scale and from the capabilities they build around that work.
The economics in the sample are significant. McKinsey reports that the 20 companies improved steady-state EBITDA by 20 percent on average after three years, measured at company level or at the relevant business-unit level where the transformation was narrower.
Just as important is where they concentrated their attention. Two-thirds focused on three business domains or fewer. McKinsey describes those domains as economic leverage points, places where relatively small improvements could create material financial impact.
It challenges the way AI portfolios are often discussed.
Breadth tells you how widely AI is being used. It does not tell you whether investment is concentrated on the parts of the business where better decisions, lower cost, more capacity, improved conversion or reduced risk can materially change the economics.
McKinsey’s management findings make the ownership question even clearer. It describes a C-suite that understands AI as the most significant driver of success in its sample. It also says its successful examples consistently have a senior business leader who integrates the business, technology and change dimensions and is accountable for the business outcome. In some cases that can be a paired business-and-technology model, but the business outcome still has an identifiable owner.
Sponsorship and ownership are different jobs.
An executive can sponsor an AI programme without owning the operating decisions that determine whether value appears. A central AI team can create capability without having authority over the process, people, resources and commercial choices that turn that capability into an outcome.
McKinsey’s examples point to a more practical model. Technology and data capability sit inside a broader management system. Business leaders decide where AI matters, reshape the surrounding work, carry accountability for the result and keep improving the capability as conditions change.
Scaling is not simply a rollout exercise.
McKinsey gives the example of Freeport, where a senior operational leader with AI experience and credibility in the business became effectively the domain owner for a critical part of the operation. He led the AI system into the operation and remained accountable for the business outcome. Its broader examples make the same point in different ways: adoption, process design, data, technology and management capability have to reinforce one another.
McKinsey selected 20 successful companies, so the study does not prove that every organisation should concentrate on exactly three domains, copy the same operating model or expect the same financial result. It is evidence about a high-performing sample, not a universal formula.
Apply the same test to any material AI initiative. Identify the economic leverage point it is meant to improve, the business leader who owns the outcome and the operating decisions that person can actually make. Then establish what evidence would show that the business is changing, rather than simply generating more activity, and what would justify more investment, a change of course or a stop decision.
If those answers are clear, the organisation may already have the ownership discipline it needs. If they are not, another dashboard will not solve the problem. The missing issue is likely to be the connection between AI investment and the management decisions that produce the business outcome.
The most useful part of McKinsey’s playbook is the management discipline behind the technology. AI value has to survive inside the operating decisions of the business; it cannot be delegated to the technology function simply because the capability is technical.
A useful test for an executive team is this:
If your AI investment doubled tomorrow, which business leader would know where to place it and what business result would justify that choice?
Sources: McKinsey & Company, “The new management playbook for AI: How to move faster and create more value”, 28 August 2026.