Everyone agreed.Nothing changed.
Compliance on paper, the old behaviour underneath.
The leadership team feels that something must be done with AI, so an AI initiative is started. The direction, governance, goals and the envisioned technology are perceived to be clear enough. The journey starts.
And no one argues with it. People attend, they try the tool, and after the initial excitement it is clear that making this technology work requires much more than what was anticipated. So, after a while the new way is abandoned and little has changed, other than growing scepticism towards the value of AI.
The middle layer is asked to champion a change that rewires the core whilst there is no full visibility on the nature of their job once the AI transformation is implemented.
Another communication round does not settle it. Neither does another pilot.
The question is how to make the mindset & behaviour side of AI transformation stick and create conditions for the organisation to lean into what now becomes possible.
How we work on it
A learning loop: decide, lead, verify, inside one real change.
- Decide. Your leadership team in one room, holding your current plans against the futures and deciding what changes. The first cycle builds those futures: a one-day AI Leadership Experience.
- Lead. Those decisions applied inside one real change, rather than in a parallel programme.
- Verify. A measurement of what people actually do, re-run after the attention has moved on.
Across all three, four questions do the work. The fourth is the only one that proves anything.
Why change?
There is a business reason underneath every real transformation, and it has to be said out loud. The test is whether standing still is still a viable option. If carrying on as before remains an acceptable outcome, it is the one the organisation will take.
Where to start?
At the top. Not because leadership needs convincing first, but because the top sets the boundary of what the rest of the organisation believes is allowed. A team does not take a risk its leadership has not visibly taken.
What defines the change?
The strategy, not the technology. A transformation that enables the strategy has something to hold on to when it gets uncomfortable, and it will get uncomfortable. One that runs beside the strategy is the first thing dropped when the quarter gets hard.
How do we know it is successful?
You run the next one without us. We are building an internal AI transformation capability rather than delivering a programme, so we come back after this change has stopped being interesting and look at what your leaders did with the one after it. Behaviour that held only while we were in the room was attendance, not capability.
What your leaders own after our one-day AI Leadership Experience
Five concrete deliverables that this day yields, ensuring a solid start to your AI leadership team journey.
“You only see it once you get it,” as Johan Cruijff put it. That shift is what the first day is for.
- A five minute film of the future your own room described, and it is yours to decide what to do with
- Which of the priorities already in your plan are robust across the set, and which are brittle
- A small set of plausible futures for your market, written by the room as stories rather than forecasts
- One early signal each, with a source and a name against it, across people, business and technology
- The strategic conversation the day makes unavoidable, and the meeting you already hold where it lands
That is the start of cycle one. Lead and verify follow, inside a change you are already making.
One factor at zero, result at zero.
People is one of three factors, and it is a multiplication rather than a sum. The tooling can be right and the process rebuilt, and if the behaviour does not move the result stays where it was. This is the factor that quietly decides whether the other two pay off.
McKinsey surveyed 1,719 organisations in 2026: 89% use AI in at least one business function, 44% have scaled it across the enterprise, and 6% say they can attribute 5% or more of their profit to it. The distance between scaling something and being paid for it is where this factor sits.
Nobody was ordered to.
A 950-person accountancy firm, where adoption was voluntary throughout.
869 of 950 people are active firm-wide, counted in the system rather than self-reported.
96% of the 105 volunteers stayed active for the full three-month pilot, voluntarily. 79% said the work had become more enjoyable. Participants self-reported freeing 162 minutes a week on recurring work, against a target of 45.
Those are behavioural numbers, and behaviour is what this work is judged on. Nobody stays in a voluntary pilot for three months because the strategy said so.
Jeroen Haverkorn van Rijsewijk
Leadership & Growth · founding partner
Facilitates leadership teams through the part of AI adoption no tool solves: building the trust a team needs to drive an AI transformation, and having an honest discussion on what AI does to the organisation and its licence to operate.
He graduated in evolutionary algorithms at Leiden University in 2003, a branch of AI two hype cycles before this one. Then fifteen years facilitating transformation at organisational, team and individual level, at Aberkyn, McKinsey's leadership arm, and now H3. Most AI decisions fall between two rooms, the one that understands the technology and the one that has to carry what it changes. He works in both.
Jeroen is associated with ForChiefs (opens in a new tab), an alliance of executive sidekicks who work with leaders in organisations on four pillars of transformation: purpose, vision & strategy, culture, and leadership.
(opens in a new tab)
Asked in every first conversation
Is this change management?
It overlaps, and the difference is where it starts. Change management usually begins after the decision, to get people through it. We start before it: why this has to change at all, whether standing still is still an option, which strategy the change is there to serve, and how you will know it worked.
What if people are afraid of losing their jobs?
Then that is the first thing to answer, honestly, before anything else lands. An unanswered version of this question does not stay quiet. It turns into hedging, slow adoption and workarounds. We help leadership decide what is true and say it, including when the answer is uncomfortable.
How do you measure whether behaviour actually changed?
By what people do, counted where possible rather than surveyed, and re-measured after the programme has stopped being interesting. A change that only holds while it is being watched has not held.
What about the middle layer?
They carry the heaviest version of it. The change they are asked to champion also rewrites their own job: what they decide and what they are measured on both move. Leadership has to describe that job before asking them to lead the change.
Do we need a programme for this?
No, and a programme is often how it gets avoided. This runs inside one real change with the people affected by it: one day to decide, then the work itself, then a measurement later.
What if the leadership team is not moving either?
Then that is the finding, and it is worth having early. In McKinsey's 2026 survey of 1,719 organisations, the ones that can point to AI in their profit were twice as likely to say their senior leaders genuinely own it. No amount of work lower down compensates for a leadership team that behaves as if the change applies to everyone else.
Start with what needs to change.
A one-day AI Leadership Experience: the people who have to carry the change in one room, and a decision written down at the end with owners against it.
See how we startStart with why it has to change.
Let's work out what the business reason is, what happens if nothing moves, and who has to carry it. If the problem turns out to be the process rather than the people, we will say so.