Who actually owns AI in your organization?

Who actually owns AI in your organization?

I started asking this question at the LGPro VIC (Local Government Professionals) conference in June of this year.

Two weeks ago I asked it in Bengaluru and Mumbai, to an audience of senior technology, security and risk leaders, as part of a data and AI security roadshow with Veeam Software Securiti AI.

Two weeks from now I’ll ask it to the same type of audience in Melbourne, Hong Kong and Singapore.

At each event, the answer comes back fast – and remarkably consistent.

The audience agrees – the business owns it and IT supports it.

That’s a perfectly sensible answer – but it’s a much smaller question than the one I asked.

The question a room of tech professionals hear

If you’ve spent your career running technology rollouts, the question, “who owns it” is a no brainer. Who buys the software? Who implements it? Who trains people to use it? You pick the tool, you configure it, you train the users and you hand it over.

IT rollouts used to be long, slow, expensive – and confined.

But #AI has changed that world.

So I push these audiences a little further. Who’s responsible for the data cleanup, the slow, unglamorous work of getting your data to a point where you’d trust an AI to act on it?

Who’s responsible for bringing the humans along, explaining (or better yet, consulting with) the humans whose jobs will be impacted by the rollout?

Who manages the agents once they’re running? And what happens if they go rogue. Does your staff know how and to whom to report this?

And are any of those decisions actually being made, or is everyone quietly assuming someone else has made them?

What I’m hearing in these event rooms and beyond

When I talk to people one on one about these issues I hear the same issues come up repeatedly.

The first is that nobody is responsible for data cleanup. When I ask who owns it, I don’t get a debate about which team. I get no name at all. It’s the job everyone agrees matters and nobody has been handed.

The second is that the line between IT and the business is blurring, and mostly in one direction. IT teams are being asked to come up with the use cases for the tools they’ve rolled out. Some are being asked to weigh in on how roles get redesigned. Sit with that for a moment. The people brought in to make the technology work are increasingly being asked to decide what work people do. That isn’t a criticism of IT, who are often the most AI-fluent people in the building. But it isn’t the job they were given, and they don’t hold the levers that make that kind of change stick: reporting lines, communications, and how roles are designed and rewarded.

And the second leads us to the third. Workflow design, or more accurately, redesign. When IT pushes back and tells the business that it’s not responsible for role design HR is tasked with this. And yet, they don’t know what people do on a day to day basis much more than IT.

And almost nobody mentions communications, meaning what gets said to the people whose work is about to change, when it gets said, and in what words.

Veeam’s own commissioned research, a survey of 300 technology and business leaders, found just 7% have a dedicated AI governance leader. I’m not suggesting the fix is one more title on the org chart. But in my experience, “it’s shared” has a habit of quietly becoming “it’s nobody’s.”

Why this keeps happening

None of this comes from carelessness. It comes from a mental model that served organizations well for decades.

We’re treating AI implementation the way we treated every technology implementation before it. One big decision, made once. Expensive, so it gets examined carefully up front. Then a go-live, a round of training and a handover to business as usual.

AI doesn’t hold still long enough for that model to work. The tools change underneath you every few months. Agents don’t sit politely waiting to be used; once they’re live, they act, and someone has to watch what they do. Data you cleaned last year drifts again. A workflow redesigned around one version of a tool needs another look when the next version arrives. And every one of those changes lands on people who need to understand what’s happening to their work, and why.

So ownership of AI can’t be a decision you make once at the start. It has to be a set of jobs someone keeps doing for as long as the technology is running.

A sharper version of the question

“The business owns it, IT supports” is a fine place to start. It just needs breaking into the pieces that actually do the work.

Pick one AI initiative that’s running in your organization right now, and try to write a person’s name, not a department, next to each of these:

  • Getting the data clean enough to trust, and keeping it that way
  • Redesigning the workflow around the tool, rather than bolting the tool onto the old one
  • Deciding what changes for the people whose roles it touches
  • What gets said to those people, when, and in what words
  • Monitoring the agents once they’re live
  • Making the call, and owning the explanation, when something goes wrong

If you can put a name next to all six, you’re ahead of nearly everyone I talk to. If you find “IT” written next to four of them, or a question mark next to the data, that’s a conversation worth having this week, while it’s still a planning question rather than an incident review.

I’ve got more cities ahead on this roadshow, and I’ll keep asking. I’d genuinely like to hear how it lands in your world. Of those six jobs, which one has nobody’s name next to it where you work, and what’s stopping someone from picking it up?

Scroll to Top

SEE Kim Seeling Smith ON STAGE

Where AI strategy meets human reality.
Live, bespoke AI demonstrations on stage.