AI has a habit of spreading.
It might begin with a technology team experimenting with models.
Then a sales team starts using an assistant. Marketing adopts another tool. Customer service explores automation. Finance finds a way to accelerate analysis. HR starts testing AI in recruitment. Individual employees discover their own uses. A governance group begins thinking about risk. IT worries about integration. Legal worries about data. Someone in operations starts redesigning a workflow.
Each decision can be perfectly reasonable.
But eventually a different problem appears.
Who is responsible for making sure all of this adds up to something?
That question becomes more important as AI moves beyond isolated experiments and into the way an organisation actually works.
A business can have people responsible for AI technology, AI risk, individual AI projects and AI adoption without anybody being clearly accountable for whether AI is creating meaningful business value.
And that creates a dangerous gap.
Owning AI technology isn’t the same as owning AI outcomes.
AI crosses too many boundaries to belong neatly to one function#
Most technologies have an obvious organisational home.
Finance systems sit naturally with finance and technology teams. Marketing platforms have marketing owners. HR systems have HR owners. Infrastructure has technology owners.
AI is more awkward.
The model may be technology.
The data may belong to several functions.
The workflow may cross departments.
The employee behaviour required to make it useful may be a management problem.
The risk may involve legal, security and compliance.
The commercial outcome may belong to a business-unit leader.
The budget may sit somewhere else entirely.
This is one reason AI implementation can become fragmented so quickly. Everyone owns a legitimate piece of the problem.
But owning a piece is not the same as owning the whole.
In The Distance Between AI Capability and AI Value Is an Implementation Problem, I argued that increasingly capable models don’t automatically produce increasingly valuable businesses. Organisations still have to connect technological capability to processes, systems, people and economics.
Once you look at AI that way, ownership becomes difficult to avoid.
Someone has to connect the pieces.
Distributed experimentation eventually needs direction#
Early experimentation benefits from being distributed.
You want people across the organisation discovering what AI can do.
A salesperson will see opportunities a central technology team won’t. A finance employee will understand problems specific to finance. An operations team will know where its processes are slow. A customer-service employee will recognise repetitive work that outsiders may never notice.
That local knowledge matters.
But distributed discovery creates a second-order problem: an organisation can generate far more AI opportunities than it can sensibly pursue.
I wrote earlier this year that you can’t pursue every AI opportunity.
That isn’t just a prioritisation problem. It is an ownership problem.
Who decides which opportunities matter?
Who compares a customer-service opportunity with one in finance?
Who decides whether scarce technical capacity should support an efficiency project or a revenue opportunity?
Who stops a technically interesting initiative that has weak economics?
Who says no when a senior executive wants their project prioritised?
Who makes sure several teams aren’t solving versions of the same problem independently?
Without clear ownership, prioritisation can become a negotiation between functions rather than a decision about business value.
Ownership should begin with business priorities#
This is why I remain wary of AI strategies built primarily from lists of use cases.
In Stop Starting With AI Use Cases, I argued that businesses should begin with the problems and outcomes that matter rather than the technology itself.
Ownership should work the same way.
The person responsible for AI shouldn’t merely ask:
What AI projects are we running?
They should be able to answer:
Which business priorities are we trying to improve, why do we believe AI can materially affect them, and what evidence would tell us whether it has?
That changes the job.
The objective is no longer to maximise AI activity.
It is to allocate attention, capital and organisational effort towards the places where AI can create the greatest value.
That means understanding enough about the technology to know what is becoming possible, but also enough about the business to know what is worth doing.
Implementation crosses the organisation#
Even after the right opportunity has been selected, ownership matters because implementation rarely stays inside one department.
Last month I wrote that AI strategy depends on people using it. The month before that, I argued that AI won’t transform work unless the workflow changes.
Put those arguments together and the management challenge becomes clearer.
Imagine an AI initiative designed to improve how a business handles complex customer requests.
Technology may need to integrate the model.
Data teams may need to make customer information available.
Operations may need to redesign the workflow.
Legal and security may need to define appropriate controls.
Managers may need to change responsibilities.
Employees may need role-specific training.
Performance measures may need to change.
Someone may need to decide which cases AI can handle and which require human judgement.
The business may then need to determine what happens to any capacity released.
No single specialist necessarily has authority over all of those things.
Without someone accountable for the end-to-end outcome, each function can successfully complete its part while the initiative as a whole underperforms.
Technology can say the system works.
Security can say it is compliant.
HR can say training was delivered.
Operations can say the process changed.
Employees can say they have access.
And the CFO can still reasonably ask:
What did the business get?
Governance ownership is not the same as value ownership#
There is already evidence that senior ownership matters.
McKinsey’s March 2025 State of AI research found that CEO oversight of AI governance was among the organisational attributes most correlated with higher self-reported bottom-line impact from generative AI.
Twenty-eight per cent of respondents whose organisations used AI said the CEO was responsible for overseeing AI governance, while 17% reported board oversight. Respondents reported an average of two leaders responsible for AI governance.
That doesn’t mean CEO oversight causes higher returns. The research is correlational, and governance is only one part of creating value.
But it does reinforce a broader point: as AI becomes consequential, responsibility moves upwards.
There is an important distinction here, though.
Governance asks whether AI is being used appropriately.
Value ownership asks whether the organisation is using AI to produce worthwhile outcomes.
The two should connect, but they are not interchangeable.
A business can govern an AI initiative extremely well and still choose the wrong problem, implement it poorly or fail to capture any economic benefit.
Does every business need a Chief AI Officer?#
The emergence of the Chief AI Officer is one response to this ownership problem.
IBM Institute for Business Value research published in 2025 found that 26% of organisations in its global research had appointed a Chief AI Officer. Its underlying study surveyed more than 600 CAIOs or equivalents across 22 countries and 21 industries.
The growth of the role makes sense.
AI is strategically important, technically complicated and organisationally diffuse. Creating a senior position with explicit responsibility can provide focus.
But I don’t think the conclusion should be that every organisation needs a Chief AI Officer.
Titles don’t solve accountability problems by themselves.
A CAIO without authority over priorities, resources or implementation can become an AI evangelist rather than an owner.
Equally, a business may already have an executive capable of owning the agenda: a CEO, COO, CIO, CTO, transformation leader or another senior operator, depending on the organisation.
The important question isn’t what the title says.
It is whether somebody has enough authority, commercial understanding and organisational reach to connect AI activity to business outcomes.
Central ownership doesn’t mean centralising every decision#
There is an obvious risk here.
If one person owns AI, do all AI decisions have to go through them?
That would probably be a mistake.
AI is too broad and is changing too quickly for a central team to understand every useful application across an organisation.
Business functions still need autonomy.
Employees still need room to experiment.
Domain experts still need to shape how AI is used in their work.
Technology teams still need to make technical decisions. Security teams still need authority over security. Legal teams still need to define legal boundaries.
Ownership should not mean becoming the bottleneck through which every prompt, licence and experiment must pass.
AI ownership is less about centralising every decision and more about making sure distributed decisions add up to something.
A good owner creates direction and coherence.
They establish what the organisation is trying to achieve.
They make prioritisation possible.
They clarify decision rights.
They ensure dependencies between functions are resolved.
They make sure somebody remains accountable when an initiative crosses organisational boundaries.
And they create enough visibility to know whether the portfolio is producing value.
The owner needs permission to stop things#
One of the least glamorous parts of AI leadership may be stopping work.
AI creates an endless supply of interesting possibilities.
New models appear. Vendors launch products. Employees discover tools. Competitors announce initiatives. Executives return from conferences with ideas.
An organisation without strong ownership can accumulate pilots indefinitely.
Every project has a sponsor.
Every team can explain why its idea matters.
Nobody wants to be the person who says no to innovation.
But resources are finite.
Technical capacity is finite.
Management attention is finite.
The organisation’s ability to absorb change is finite.
So ownership has to include the authority to stop low-value activity and concentrate effort elsewhere.
That requires a commercial standard stronger than enthusiasm.
What problem are we solving?
How important is it?
Why is AI the right capability?
What has to change for the value to be captured?
How will we know whether it worked?
If those questions don’t have convincing answers, the initiative may not deserve to continue.
Accountability should follow the outcome#
This is perhaps the most important part.
AI initiatives are often described in technological terms.
Deploy the assistant.
Build the agent.
Automate the process.
Integrate the model.
Launch the platform.
Those are deliverables.
They aren’t outcomes.
If the original objective was to reduce customer-resolution time, then resolution time matters.
If it was to increase sales capacity, then sales capacity matters.
If it was to improve conversion, then conversion matters.
If it was to avoid additional hiring while the business grows, then that operational leverage matters.
If it was to improve decision quality, the organisation needs some way of understanding whether decisions improved.
The owner of an AI initiative should ultimately be accountable for the outcome the AI was supposed to improve, not merely the technology that was deployed.
That sounds obvious.
In practice, it is a significant shift.
It moves AI away from being something the technology organisation delivers to the business and towards something the business uses to change its performance.
Ownership becomes more important as AI becomes less visible#
This challenge may grow as AI becomes embedded into ordinary work.
Today, AI initiatives are often visible because they are new.
There is a project. A tool. A pilot. A launch.
Over time, AI may become a capability inside hundreds of processes, products and decisions.
At that point, the organisation may no longer have a neat list of things called AI projects.
AI will simply be part of how work happens.
That makes ownership more important, not less.
Someone still needs to understand where the organisation is investing, which capabilities are reusable, where risks are accumulating, what employees need, which workflows should change and whether value is actually being captured.
The role may eventually look less like managing a technology programme and more like managing an organisational capability.
Someone has to connect the chain#
Over the last few months, the AI implementation problem has become increasingly clear to me.
Capability alone isn’t enough.
AI needs the right organisational context.
Workflows may need to be redesigned.
People need to change how they work with the capability.
And all of that has to connect to an economic outcome.
Each part can have a different owner.
But the chain cannot have no owner.
A business priority becomes an AI opportunity.
The opportunity gets prioritised.
The technology gets implemented.
The workflow changes.
People adopt a new way of working.
The organisation captures — or fails to capture — an outcome.
At every stage, responsibility can fragment.
The leadership task is to make sure it doesn’t disappear.
That doesn’t require one person to make every decision. It doesn’t require every organisation to create a new C-suite role. And it doesn’t mean AI should become detached from the functions where the actual work happens.
It means somebody needs to remain accountable for whether the pieces connect.
AI doesn’t necessarily need another executive title. But at some point, someone has to own the outcome.