Who Should Own AI at Work? I’m Still Thinking About the Answer

A couple of weeks ago, I ran a LinkedIn poll asking who should own AI within an organisation.

The options were HR, IT, Legal and Compliance, or shared ownership across the business.

Shared ownership won with 79% of the vote. IT received the remaining 21%. Nobody chose HR or Legal and Compliance.

My first reaction was that the result made complete sense.

AI touches too many parts of an organisation to belong neatly to one function. There are questions about technology and data, of course, but there are also questions about people,

risk, cost and how work will change.

Shared ownership felt like the obvious answer.

Then I started thinking about what that might actually look like.

Shared ownership sounds simpler than it probably is

Different functions will naturally look at AI through different lenses.

IT may be thinking about security and whether a tool will work with existing systems. Legal and Compliance may be considering what could go wrong. HR may be thinking about employees, skills and how AI could affect workplace decisions.

The department proposing the tool may be focused on a problem it needs to solve.

None of those perspectives is wrong. They may still lead to very different views about whether the organisation should proceed.

That is the part I keep returning to.

If one department opposes a tool that could genuinely benefit another, how is the overall decision reached?

I do not think shared ownership means every function must agree with every decision. That would probably make progress very difficult. But it also cannot mean that legitimate concerns are pushed aside because the potential benefit is exciting.

There needs to be a way to consider both.

I am not sure there is one model that will work for every organisation. Size, structure and the proposed use of AI will all make a difference.

My instinct, though, is that shared ownership still needs some form of clear decision-making. Otherwise, responsibility could become so widely distributed that nobody feels able to make the final call.

Could an AI Advisory and Ethics team help?

One of the comments underneath the poll suggested creating an AI Advisory and Ethics team.

I can see the value in that.

Bringing together people from different functions could help an organisation examine a proposed use more fully. IT might identify a technical concern that the operational team had not considered. HR might see a potential effect on employees. Legal or Data Protection might suggest controls that make the proposal safer.

It could lead to a better decision than any one function would make alone.

I can also imagine the difficulties.

Would the group advise or approve? Would every proposed use need to go through it? What would happen if its members could not agree?

There is a risk of creating another committee that has useful conversations but little authority to act.

Perhaps the answer is an advisory group with a named senior decision-maker who considers its recommendations and takes responsibility for the final decision.

That feels workable to me, although I would be interested to hear from organisations already doing this. The reality may be rather different once competing priorities, budgets and deadlines enter the conversation.

When does caution become limiting?

I have also been thinking about the role of risk and compliance.

Their involvement is essential. AI can create genuine issues around personal data, fairness, accuracy and accountability. Organisations need people who will challenge proposals and notice risks that others may overlook.

At the same time, I wonder whether the presence of risk can sometimes become the end of the conversation.

Very few new ways of working are entirely risk-free. There may be occasions when the risk can be reduced through a limited trial, fictional information or closer human review.

There is a difference between proceeding carelessly and testing an idea within sensible boundaries.

Doing nothing has consequences as well. Employees may continue using slow processes, or start experimenting with unapproved tools because the organisation has not provided a practical alternative.

I do not have a simple answer to where the line should sit. I suspect it will depend heavily on what the tool is doing and who could be affected if it gets something wrong.

A tool helping someone organise non-confidential notes is very different from one influencing recruitment or performance decisions. Treating every use in exactly the same way may not be particularly helpful.

Are we overlooking the people expected to use AI?

The ownership discussion tends to concentrate on who approves the technology and manages the risk.

I wonder whether we are spending enough time thinking about the people who will actually use it.

A tool can be technically secure and formally approved, but that does not mean employees will automatically know how to use it well.

Someone can use an approved system and still enter information they should have protected. A manager can accept an inaccurate answer because it sounds convincing. A team can gradually start using a simple productivity tool for tasks it was never intended to support.

Policies and controls matter, but training may be one of the most practical ways to reduce these risks.

Employees need to understand the boundaries. They also need the confidence to check what AI produces, question it and recognise when a task requires human judgement.

That learning needs to be connected to real work. A general explanation of AI will only take people so far. They need examples that reflect the decisions and information they handle in their own roles.

This may be where HR has a clearer part to play

HR received no votes in the poll.

I do not take that to mean people see no role for HR. It probably reflects the feeling that AI is too broad for any one department to own.

Even so, I think HR has something distinct to contribute.

AI will change some of the tasks people complete and the skills they need. Employees may need to become better at assessing information, recognising missing context and explaining how they reached a decision.

Managers may need support to understand where AI can help and where their own judgement remains essential.

Some roles may change gradually. Others could need more deliberate redesign. People will need opportunities to learn rather than simply being given access to a tool and expected to work everything out for themselves.

HR already works across learning, job design, workforce planning and organisational change. That seems like a useful perspective to bring into AI discussions early, even if HR does not own the wider agenda.

Perhaps this is the part that was missing from my original poll.

It asked who should own AI, but ownership is only one part of the picture. Someone also needs to help the workforce understand it, use it responsibly and develop the skills that changing work will require.

Where my thinking has landed

I still think shared ownership was the most sensible poll answer.

I am just less certain that saying “shared ownership” gets us very far on its own.

There may need to be a group that brings together the different perspectives. There probably also needs to be someone who can make the final decision when those perspectives do not align.

And alongside the policies, approvals and risk assessments, organisations may need to give just as much attention to workforce capability.

That is the part where I think HR can make a meaningful contribution.

I do not have the perfect structure for how all of this should work. I am still thinking it through, which is partly why I wanted to continue the conversation.

How is your organisation approaching AI ownership, and who is thinking about the skills and training people will need alongside it?