When a room of senior business leaders was asked what is getting in the way of measuring AI success in their organisations, the most common answer was data. The second was mindset.
“Data” feels obvious, but “mindset” is more revealing.
In this context, mindset does not mean scepticism about AI or resistance to change in the abstract. It means the willingness of the person in the middle of an organisation to change how they work, and specifically to give something up. That is a structural issue rather than a training or technology one, and most organisations have never been designed to make it possible.
The loudest AI conversation in most organisations happens at two levels. At the top there is strategy, investment, and the board-level narrative about what AI will mean for the business; at the bottom there is individual experimentation, tool access, training completions, and pilot results. Both matter. But the bit in the middle, where work actually happens and where AI has to change behaviour rather than just capability, is where it consistently stalls.
That gap has a name: the accountability gap. No single function was designed to own the commercial outcome of AI adoption, so nobody does.
Ask your IT lead how AI adoption is going and you will likely hear about licence utilisation and rollout timelines. Ask HR and you will hear about completion rates and engagement scores. Both are real measures of activity, however the question worth sitting with is whether either connects to a change in business performance.
One of the panellists, the CIO of a global law firm, put it directly: the operating model shift is owned by the commercial owner. If it is not, it will not work.
That framing matters because it reframes the problem. The mindset shifts from, “Which function should lead AI adoption?” to “Has anyone been made accountable for the outcome across both HR and IT?” When that accountability sits with a P&L owner, the right questions get asked: Did costs move? Did the business write more, charge more, or deliver faster? Did the commercial model shift? When accountability sits within a single function, the questions tend to stay functional too. What gets measured is what that function controls, not what the business ultimately needs to know.
Carly Lamb, AI Adoption and Workstream Lead at North Highland, named the specific moment where most organisations stall. Before you can make an operating model shift, she said, you have to answer two questions: what is AI going to own, not just help with? And who is giving up responsibility for it?
Those are different questions from the ones most organisations ask. Most organisations ask, “Where can AI add value?” and stop there. When the answer is everywhere (and it will be) it gives you a long list of opportunities and no mechanism for deciding which ones to actually change. Great for kicking off a brainstorming session, less effective for building a concrete plan.
Now let’s reverse that. When you ask things like what AI is going to own, and who is handing over the process they currently hold, the conversation becomes specific. It forces a decision rather than a discussion. And it creates the conditions for measuring whether anything actually changed.
Most leaders want the upside without that decision. They want AI to help, not to replace a process they are currently responsible for. So the old process keeps running alongside the new one, the organisation pays for both, and accountability for the AI version stays unclear.
When asked what was getting in the way of measuring AI success, the third most common answer turned out to be defining it. (Source: live poll, North Highland executive breakfast, June 2026.) But here’s what we’ve learned: Data, mindset, and defining success are really one problem showing up in three different ways.
The data problem is real, but data problems are usually symptoms. If you do not know what you are measuring, data quality cannot help you. If your success metrics are activity-based, clean data will give you accurate activity metrics, which do not tell you whether anything has changed in the business.
The defining success problem is structural. Before you can define what success looks like for an AI investment, you need to know what you are trying to change, in commercial terms. That requires the person accountable for commercial outcomes to be in the room at the start, not brought in after the fact to validate a rollout that has already happened.
The mindset problem is the consequence of not solving the other two. People in the middle of organisations are not resistant to change for no reason. They are resistant to ambiguous change with unclear upside and clear personal cost. Give them a specific commercial outcome to own, a clear definition of what success looks like, and clear accountability for the result, and the conditions for behaviour change are in place.
Before assessing tools, building a training catalogue, and launching a strategy session, you need to start here: Are the people responsible for the technology and the people responsible for the workforce being held to the same commercial outcome?
If the answer is no, or not yet, that is where the work starts. Not with the tool. Not with the training catalogue. With the structure underneath both.
The organisations seeing real returns from AI are not the ones spending the most or moving the fastest. They are the ones that built deliberately: designing the operating model, culture, and workforce together rather than in sequence, and putting someone with commercial accountability in charge of the outcome.
If you are wondering where your organisation stands on that, the answer is usually visible in the conversation that happens when someone asks: who is accountable if this does not work?
That conversation is exactly what we are built for. See where your organisation stands with our four-question AI diagnostic.
Drawn from a panel discussion at North Highland's executive breakfast, The Missing Half of Your AI Strategy, June 2026. Panellists: Andrew Allen, Chief Digital and Technology Officer for Government, Microsoft; Andrew Brammer, Global CIO and SCC Director, A&O Shearman; Steve Brennan, Market Enablement Director, QA; Carly Lamb, AI Adoption and Workstream Lead, North Highland. Facilitated by Laura Cameron, North Highland.