AI programmes often measure what is easy to count: adoption rates, hours saved, cost per case. What dashboards would often miss is a team that works differently. Why this is relevant: Change rarely starts at the top. It starts in one unit, around one problem, led by the person who runs it. This Practice Note shows what that looks like, and how the moves of this series come together into a first track on your own unit: one problem, worked through until the unit works differently.
The programme has done its part. The licences are out, the first training wave is through, a network of champions meets every other week, data and governance workstreams are humming along, and the adoption dashboard shows a healthy curve. Ask what has changed in the work, though, and the answers get vague. The reports look the same, the hand-offs run as before, the meetings have the same agenda. Somewhere in the organization there is usually one exception: a unit where the leader or a coalition of enthusiasts picked a problem that mattered to the team, changed who does what around it, and kept at it (solving data, integration, approvals) for a few months. That unit rarely shows up on the dashboard. It is where the transformation is actually happening – and where the capabilities your organization urgently needs are built.
This is an old pattern, and it has been studied. In 1990, Michael Beer, Russell Eisenstat and Bert Spector followed company-wide change efforts in large corporations and found that the programmes from the top rarely produced change (Harvard Business Review, 1990). They called the underlying assumption the fallacy of programmatic change: the belief that changing attitudes, through training, mission statements and campaigns, will change behaviour. In the companies that did change, it ran the other way round. Units started with a concrete business problem and changed roles, responsibilities and relationships to solve it. They called this task alignment. The new behaviour came from the new context, and the attitudes followed. The centre's job was to spread what worked, "without pushing it from the top".
AI programmes, unsurprisingly, repeat the fallacy with new tools. If one were being polemical, one could say that AI skills training and access are the attitude campaign of 2026. Both things are true: they are necessary, and they are not the change. The change is in the work: it is in who decides, how the hand-offs run, what the saved time is for, which story people tell each other.
AI transformation happens in the unit, and the unit leader's first track is the transformation, not a pilot of it. A pilot is designed to be evaluated, produce information for management and inputs to decisions. A track is designed to change how one unit works, for good, and to show the rest of the organization what is possible. The programme can create the conditions. It cannot do this part for you.
Working on this in your own organization? See how the MK Practice works →
It starts with one problem that matters to your team, not with "use AI more". When I started doing AI strategy work in early 2024, the first impactful use cases had exactly this character: a team using AI to tackle an urgent issue – maybe imperfectly, but in earnest. A volume that breaks the team, a backlog that never shrinks, a hand-off that keeps failing, a report that eats a week every month.
Then the questions of this series become your track, in whatever order your problem needs them. Who may overrule the tool, and who signs what. Who gets a say in redesigning the work around it, and on which step. What the saved time and the new work are for, and what counts as good enough. And which story your team is already telling, next to the official one. Not every unit needs all of them, and they are not the only questions that will come up.
Give it six to ten weeks and three review dates, and keep a short log of what you decided and what happened. In our work with leadership teams, the log is a terrific resource afterwards: it turns one unit's experience into something the next unit can use – which is exactly the spreading Beer and his colleagues described.
This is the work the MK Practice is built for: one unit, one problem, the working topics for each of these questions, and a practitioner who works through it with you and signs off each step. If you are about to start a track like this in your own unit, we would like to hear about it.
So before the next AI programme milestone, pick the one problem in your unit that you would most like to see solved by spring, and write down what you think it would take. That page is the start of your track. The one unit and the one problem are also where a leader's learning should start; personal journeys, curated by experts, shaped by AI and signed by a practitioner, are where Management Kits is heading.
You don't read a practice. You run it.
The MK Practice is a structured process you run on your own organization: a diagnostic that reads where your unit needs work first, working topics like the ones in this series, a work agenda that keeps the moves coming, and a practitioner who looks at your problem within 48 hours. Founding Member: USD 990 a year, 30-day money-back guarantee.
Want to talk it through first? Book 20 minutes with a Management Kits team member. Running it with a whole leadership team? See the MK Practice for organizations.