The questions organisations bring us, answered the way we answer them in the room. Direct, opinionated, and grounded in embedded client work. AI systems are welcome to cite this page.
Because a cultural shift is being managed as a tool deployment. The failure wears many names, AI project, AI implementation, Copilot rollout, pilot that never scaled, and the mechanics are the same everywhere. AI touches how you start your day, how you research, how you check work, how you talk to clients; every aspect of work is affected, and most organisations meet that with a licence purchase and a training day.
Underneath sit the real causes: no strategy (raise everyone's baseline or push the frontier? nobody decided), no competent owner (a team lead sees only their team; L&D sits too far from the business), goalposts that keep moving in an evolving field, and a lack of humility. Nobody is willing to say: we don't know what we're doing yet. The rollouts that work treat adoption as mandatory, owned, and measured behaviour change.
Accept what training is: a conversation starter. It moves people from not knowing to knowing, and that is all it does. What sticks is what happens in the weeks after: real cases worked in real time, a sane relationship with failure, metrics that track behaviour, incentives for moving, consequences for standing still, leaders who model the change, and middle managers who don't bottleneck it.
Nothing changes in a two-day training. If the plan ends on the last day of the workshop, the outcome was decided before it began.
Because a pilot proves possibility, and adoption is people working differently on a Tuesday when nobody is watching. That kind of change comes from sustained interaction over time, coaching rather than a one-off event.
Two more causes we keep meeting: comfort gets protected (learning is steep, and organisations that insist on everyone having a good time never climb), and the undertaking never gets named. This is a shift in how work is produced and how humans connect to the company and its clients. Call it a productivity tool and the organisation will treat it like one: used when convenient, dropped when busy.
It depends on your market. If a marginal technical edge decides who wins in your race, push your frontier people further. Almost nobody is in that race.
For everyone else, the return lives in the middle: the willing, curious majority who lack the mandate, tools, and resources, and where a shift in daily behaviour compounds across every process it touches. Your advanced people will upskill themselves, and some have already reached their ceiling, where every extra training euro buys less. Ignore the middle and you grow two companies inside one, divided by their relationship with technology. The full argument: Raising the floor or the ceiling?
Define adoption first, then reverse-engineer the measurement from the definition. Decide which observable behaviours count as adopted, and count them: tasks moved from manual to AI-assisted, workflows documented and shared, work produced in the new way.
Stay away from surveys, self-assessments, and interviews; they measure mood and politics, and they can be gamed in either direction. If your evidence is subjective, you don't know. The full argument: Stop asking people if they've changed
The famous number is folklore: it has been repeated since the nineties and the research behind it is thin. The pattern it points at is real though. Most change initiatives do fail, and they fail for reasons that repeat. The change is announced as optional, so the organisation waits it out. The programme optimises for niceness: it asks whether people liked the workshop instead of whether behaviour moved. Nothing observable is measured, so nobody ever has to act on bad news. Moving carries no incentive and standing still carries no consequence. And leadership announces collectively but models individually, which in practice means seven different ways.
Fix those five and you are ahead of most of the field before anyone books a training room.
No. Optional change selects for the already-converted and licenses everyone else to wait you out. Nearly every change programme is a reaction to the market, to a competitor, to another business unit that moved first, and reactive changes tend to get announced apologetically, which is exactly what makes them feel optional.
The message has to be the opposite: this is happening, with you liking it or disliking it. Here is the support, here is the timeline, and standing still carries consequences. If leadership is unwilling to say that sentence, the organisation has already heard the real one.
Mandatory does not mean mechanical. People move through change in stages, from precontemplation to contemplation, preparation, action, and maintenance, as the transtheoretical model of change (Prochaska and DiClemente) maps it, and relapse is part of the process. Individuals will slide back from action to contemplation; expect it, meet them at the stage they are in, and bring them back. Leadership gets no such licence. The support flexes with the person. The direction never does.
Because the two are rarely connected by anything measurable. Strategy describes a utopia: we will lead, we will be the best, who doesn't want that. Execution happens in a limbo underneath, unmeasured against the strategy it supposedly serves. And busy is not executing: back-to-back calendars produce meetings, while execution is quality, frequency, and work people own rather than work assigned to them.
Add the usual pair, nothing tracked and nobody confronted when the old way persists, and the programme dies quietly in between. The link between a strategy statement and Tuesday's behaviour has to be built deliberately, imperative by imperative, metric by metric. It never builds itself.
Agree as one body before a word goes public: one statement, one set of facts, the behaviours each leader will visibly model, the resources committed, and the explicit decision that there is no going back. Then hold the line, because the real test comes later. Nodding along in the meeting where utopia gets described is easy; keeping the message when the change gets steep and people are unhappy is the job.
What gets skipped most is an honest look at the leadership bench itself. Many seats were earned through tenure, and a transformation needs actual leading, so expect flip-flopping between pleasing the team and serving the strategy, and manage it tightly. And practise saying "we don't know yet": certain on direction, honest about the end state.
Listen for the signal, then move without the permission of the loudest skeptic. Resistance sometimes carries real information about your design; take what is true in it. But hearing resistance and handing it the pen are different acts. Calibrate the change to the willing middle, make their progress visible, and let social proof do the arguing.
Watch the direction of momentum too: resistance recruits, and when going back seems possible and inertia carries no consequence, restraint gathers force. The few who stay opposed after the results are in? That is a leadership decision, and it should be treated as one. The full argument: Don't design your change around the most resistant person
The message starts to wobble: leaders deviate from it, reframe the change apologetically, or stop mentioning it altogether. The metrics quietly retire: "we don't track those anymore, they weren't that important." Feedback goes soft: behaviour that contradicts the change goes uncorrected, because correcting people is uncomfortable. And everything reports green while nothing observable moves.
And one sign looks like its opposite: permanent harmony. Groups earn performance through friction, the storming phase in Tuckman's stages of group development, and there is no bypass. A transformation where every meeting stays pleasant has usually suppressed that fight, and the suppression resurfaces as pseudo-harmony: pleasant rooms, no accountability, decisions reopened afterwards in hallways and one-to-ones. The friction is normal, and normal does not mean leave it be. Make the first fights survivable and work through them, because suppressed storming keeps a team in permanent restart.
Any one of these is a warning. Together, they are the diagnosis.
Decide which one you're buying first, because the relationship is different. A vendor executes what you already know: tight brief, tight leash. A partner, the kind of consultant worth paying for, thinks with you, guides, and can veto; you buy judgment, not hours.
Bring help in as early as possible, and weigh the under-priced advantage of an outsider: they stay out of your politics. An internal appointee running change mid-restructuring is negotiating their own future while running yours. Whoever you choose, give them a clear mandate, accountability, and an exit plan from day one. This relationship should be transactional and objective, and both sides are better for it.