AI Adoption Isn’t an IT Decision

That framing is the first mistake, and it’s an expensive one. The evidence now accumulating says the thing that decides whether AI pays off is almost never the technology. It’s the organisation around it.

The tell is in the disappointment

Start with the returns, because they’re blunt. In one 2026 enterprise survey — from the AI firm Writer, so read it with a vendor’s incentives in mind — fewer than a third of executives said they’d seen significant return from generative AI, fewer still from AI agents, and nearly half called their AI adoption so far a ‘massive disappointment.’

Sit with how strange that is. The tools work; anyone who has used a current model knows they are capable. Yet the returns keep coming in thin. When capability is high and payoff is low, the bottleneck isn’t the tool — it’s everything the tool is dropped into. And that is not an IT problem.

The constraint is organisational, not technical

Microsoft’s 2026 Work Trend Index put numbers on exactly this. Analysing what actually drives the value organisations get from AI, it found that organisational factors — culture, manager support, how work and talent are structured — explained more than twice as much of the impact as individual factors did. Its own summary is blunt: the main obstacle to AI’s impact is not the technology, and not the people, but the organisational systems around them.

McKinsey’s research points the same way from a different angle. Agents are being deployed widely, but only around a quarter of organisations have actually scaled one into real operation. Deploying is easy. Adopting — changing how the work is done so the capability compounds — is the hard part, and it is an organisational act, not a procurement one.

Which makes it a leadership decision, not a rollout

Once you see the constraint as organisational, the decision changes hands. Where AI goes first and how fast; whose roles get rebuilt around it; what gets reskilled and what gets retired; what the customer experience becomes; how much ground a competitor takes while you deliberate — none of these are configuration settings. They are strategic, cross-functional calls about the shape of the business, and they belong to the people accountable for that shape, not to whoever owns the software.

The Writer survey, for all its vendor framing, catches the human edge of this: 95% of executives said roles and team structures were already changing because of AI, and — where the change was handled badly — 54% said it was ‘tearing their company apart.’ Roles change either way. Whether that lands as new capability or as damage is a leadership decision about pace and sequence, made well or badly.

The pace problem

Pace is where it bites hardest, because both directions carry a cost. Move too fast and you break things that were working — quality, trust, the confidence of the people whose jobs just shifted under them. Move too slow and a competitor who judged it better takes ground you don’t get back. There is no safe default, no setting marked ‘correct.’ It is a genuine trade-off, weighed under uncertainty, with real consequences on either side — which is the defining shape of a leadership decision, not a technical one.

Rehearse the decision before you live it

Here is the awkward part. Most organisations make this decision exactly once, live, with the real business — and learn what they got wrong from the wreckage, or from the window they missed. The stakes are at their highest and the practice is nil.

That is precisely the gap a business simulation is built to fill. Put a cross-functional leadership team around a board, hand them the adoption decision — where to deploy, how fast, what to protect, what to let change — and let them argue it out, commit, and watch the consequences land, before any of it is real. The value isn’t a right answer handed down; it’s the team building the judgement to make the call together, having felt the trade-offs bite somewhere a wrong move costs nothing but a rethink.

And because the decision is fundamentally social — cross-functional, contested, owned by no single person — it wants the format that forces people to reason out loud, in the same room, looking at the same board. That is the thing a board-based simulation does that a briefing deck cannot.

The durable point

Treating AI adoption as an IT decision quietly assumes the hard part is the technology. The evidence says the hard part is the organisation — its structure, its pace, its people, and its judgement about all three. Those don’t get resolved by a rollout plan. They get resolved by leadership making a series of difficult, contested calls, well — and the organisations that make them well will mostly be the ones that practised.