Ask a leadership team who owns AI and you will usually get a name within five seconds. Ask what that person is allowed to decide without checking first, and the room goes quiet. The distance between the name and the authority is where most AI strategies stall.
Who should own AI strategy?
Your leadership team owns AI strategy and cannot give it away. What it can give away is delivery. What it should spread rather than concentrate is expertise. Most organizations do the opposite: they appoint one person, hand them all three jobs at once, give them no time and no authority, and call that ownership.
This is not a criticism of the people appointed. It is a criticism of the appointment. An AI lead can run a rollout well. An AI lead cannot decide what your organization will refuse to use AI for, and cannot be the person who answers when a parent, a governor or an inspector asks who approved it.
Why "who owns AI?" is really three questions
The phrase "owning AI" bundles three jobs that have different owners, different skills, and different consequences when they go wrong. Separate them and the question answers itself.
The decision. What AI is for here, what we will not use it for, what we are stopping to make room, and who answers when it produces something wrong. This is the work of the leadership team, and it is the work I wrote about in which decisions AI should never make.
The delivery. The rollout. Which tools are approved, who is trained and when, which pilots run, what gets measured, what happens on the date the pilot ends. This needs one name, a calendar, and the authority to say yes.
The expertise. Knowing what the tools actually do this month. It dates fastest, matters least in the boardroom, and is the one job you should never let sit with a single person.
| The job | Who should own it | What it looks like when all three land on one person |
|---|---|---|
| The decision | The leadership team, with one named member accountable | The AI lead is asked to write the AI policy, so the organization's position on AI becomes whatever one person believed in October |
| The delivery | One named person, with something taken off their plate and the authority to approve | They hold the responsibility, carry it on top of a full job, and have to ask permission for the decisions they were appointed to make |
| The expertise | Spread across teams, topped up from outside | Everyone waits for one person to have read the release notes, and the organization stays as current as that person's worst month |
The Ownership Test
The test is whether a team can say what it is handing over, what it has removed to make room, and who still answers for the result.
The Ownership Test is three questions I ask before a leadership team appoints anyone to lead AI: Which of the three jobs are we handing over, the decision, the delivery or the expertise? What has come off that person's plate to make room? And when this goes wrong, whose name is on it? A team that can answer all three has delegated. A team that cannot has appointed someone to be blamed.
The third question is the one that changes the room. Teams answer the first two quickly and then discover they have no answer to the third, because the honest answer is the head, the principal or the chief executive, and that person has just spent twenty minutes explaining why AI belongs to someone else. This is my suggested way of thinking about the decision rather than a tested method, but I have yet to see a team answer all three and still appoint the way they were planning to.
What goes wrong when one person is given all three
The failure is quiet, which is why it takes a year to notice. Nothing collapses. The rollout simply narrows to whatever the appointed person can personally carry.
I have held a version of this job. As Director of Digital Strategy in a further education college I was the person the technology was handed to, and the lesson I took from it is that a title moves the work without moving the power. Across the leadership teams I work with now, the same pattern shows up in three ways.
The lead becomes a bottleneck. Every question routes to one inbox, so the pace of the whole organization becomes the pace of one person's diary.
The leadership team stops thinking about AI. Once a name exists, AI leaves the agenda. It reappears as an incident.
Nobody owns the consequences. When something goes wrong, the appointed person has responsibility without authority and the leadership team has authority without responsibility. Both are surprised.
What the evidence says about the ownership gap
The gap between how much AI is being used and how much of it anyone owns is now measurable.
IBM's study of AI in K-12 education, published on September 2, 2026 and conducted by Morning Consult with 1,019 educators and 1,029 parents in July 2026, found that 76% of middle school and 73% of high school classroom educators report AI being used in their classroom at least weekly, while only 20% of K-12 educators say they have received extensive AI training. Only 24% of educators, and 16% of parents, say the US education system is adapting very well to AI.
RAND's survey panels found the same shape a year earlier. In AI Use in Schools Is Quickly Increasing but Guidance Lags Behind (Doss and colleagues, September 30, 2025), 45% of principals reported having school or district policies or guidance on AI use, against 54% of students and 53% of English language arts, math and science teachers using AI for school.
Read those two findings together and the problem is not adoption. Adoption has already happened. The problem is that use is universal and ownership is optional.
The standards that govern AI risk everywhere else are unambiguous about where that lands. The US National Institute of Standards and Technology's AI Risk Management Framework (2023) states at GOVERN 2.1 that "Roles and responsibilities and lines of communication related to mapping, measuring, and managing AI risks are documented and are clear to individuals and teams throughout the organization," and at GOVERN 2.3 that "Executive leadership of the organization takes responsibility for decisions about risks associated with AI system development and deployment." Documented, clear, and executive. Not an enthusiast with a spare Tuesday.
Education research is arriving at the same place from the other direction. DeMatthews, Reyes, Hart and James, writing in Educational Management Administration & Leadership on February 5, 2026, set out six domains of AI leadership for school principals, including ethical use, strategic planning and capacity building, and treat AI as a leadership imperative belonging to the principal rather than a technical function to be assigned elsewhere. They also argue for distributed leadership, which is precisely the split I am describing: shared work, named accountability.
Schools already know how to do this in one domain. Safeguarding solved the ownership problem by naming a role and writing down what that role must do. Keeping Children Safe in Education 2026, the statutory guidance for England dated September 2026, requires the designated safeguarding lead to take lead responsibility for safeguarding including online safety and understanding the filtering and monitoring systems in place, and now describes images that may be "digitally altered or wholly generated using artificial intelligence." A named role, a written duty, and the accountability still sitting with the governing body. Nobody thinks appointing a designated safeguarding lead means the head has stopped being responsible for safeguarding. AI has somehow acquired the opposite convention.
When a single named AI lead is the right answer
There is a real case for concentrating AI in one person, and it holds in narrower conditions than the people making it usually admit.
The case against is well made. John Winsor, Jen Stave and Ryan Kurt argued in Harvard Business Review on August 4, 2025 that appointing a single chief AI officer often fails, because the role is too broad and drifts away from what the organization actually needs, and that responsibility works better shared across executives and departments.
I agree with the direction, but a distributed model has a failure mode of its own, and I see it more often than the bottleneck: shared ownership becomes nobody's ownership, and the organization drifts for a year with everybody nodding. A single named lead is the right call when the organization is small enough that the lead sits in the room where decisions are made, when there is a hard external deadline such as a statutory policy date, or when the first job is simply to find out what is already happening. In each of those cases you are concentrating the delivery, not the decision, and you should say so out loud when you appoint.
The Leadership Question
The question to put to your own team is not who should lead on AI. It is this: if the person we are about to appoint left in six months, which of these things would stop?
Ask it before the appointment, not after. Anything that would stop was never delegated, because it was never anybody's job in the first place. It was resting on one person's enthusiasm, which is the most common AI operating model I encounter and the least durable. The things that should survive that person's departure are the decisions, the written position, and the accountability. The things that will legitimately pause are the rollout and the training calendar, and that is what a successor picks up.
Leadership teams that get this right tend to look slightly boring from the outside. They have four sentences written down somewhere: what AI is for here, what we will not use it for, who approves a new tool, and who answers when it gets something wrong. Those four sentences are the strategy. The rest is delivery.
What to name in writing this term
Four things need a name in the minutes, and all four fit inside one leadership meeting rather than an away day in the spring.
Name the accountable leader for AI decisions. One person on the leadership team, in the minutes, by name. Not a committee.
Name the delivery lead, and say what has come off their plate. If nothing has, you have not appointed anyone. You have added a line to a job description.
Write the four sentences. What AI is for here, what we will not use it for, who approves a tool, who answers when it goes wrong. Short enough to be read aloud.
Say where the expertise lives. Two or three people in different parts of the organization, not one, with a named external source they are allowed to ask.
Consider a hypothetical secondary school that appoints its assistant head as AI lead: to pass the Ownership Test, the leadership team would need to record that it has handed over delivery only, drop one of her existing responsibilities to make room, and minute that the head remains accountable for AI decisions. Same appointment, entirely different structure, and the difference takes ten minutes of honesty in a meeting.
If your leadership team can already show me who decided, who delivers and whose name is on it, you have the ownership question solved and the rest of the work is the visible practice of leading it well. If you cannot, no strategy document is going to help, because the ownership question is the one nobody outside your organization can answer for you.
Working through this with your leadership team
If your team is deciding who should own AI, and is not certain the answer would survive that person leaving, this is the kind of work I support through AI strategy sessions with school leadership teams: getting the decisions, the delivery and the accountability written down and owned before the next incident makes the decision for you.
Sources and further reading
- New IBM Study Finds AI Adoption Is Outpacing K-12 Readiness, IBM, September 2, 2026 (survey conducted by Morning Consult, July 2026)
- AI Use in Schools Is Quickly Increasing but Guidance Lags Behind: Findings from the RAND Survey Panels, Christopher Joseph Doss et al., RAND, September 30, 2025
- Artificial Intelligence Risk Management Framework (AI RMF 1.0), Core: GOVERN, National Institute of Standards and Technology, 2023
- Leadership for artificial intelligence use in schools: A six-domain framework for ethical, equitable, and effective integration, David DeMatthews, Pedro Reyes, Torri D Hart and Lebon James III, Educational Management Administration & Leadership, February 5, 2026
- Your AI Strategy Needs More Than a Single Leader, John Winsor, Jen Stave and Ryan Kurt, Harvard Business Review, August 4, 2025
- Keeping children safe in education, Department for Education, statutory guidance, 2026 edition
Dan Fitzpatrick is the founder of The AI Educator, a Forbes contributor and a bestselling author on AI in education, and works with school, trust and district leadership teams on AI strategy, readiness and governance. More about Dan.
