Access Isn’t Adoption
Organisations have thrown open the doors. Worker access to approved AI tools jumped 50% in a single year, yet among the people who have access, fewer than 60% use it in their daily work, a figure that hasn’t moved at all (Deloitte, 2026). Handing someone a tool changes nothing. Changing how the work happens is what changes everything.
That flat line is the important bit. Access has surged and usage has stayed exactly where it was, which tells you availability was never the constraint. Nobody wakes up wanting to use a tool. They want the report finished, the query answered, the backlog cleared. Until the tool is the obvious route to one of those things, at a specific moment, in a specific task, it stays an icon on a screen that people mean to explore when things quieten down.
People Are More Willing Than You’d Guess
The instinct is to blame resistance, but the data says otherwise. Just 4% of workers actively distrust and avoid AI. More than two-thirds are enthusiastic or at least open to it (Deloitte, 2026). The barrier was never attitude. It’s that nobody showed them where the tool fits into a Tuesday.
There’s a catch in the detail, though. Only around one in eight is actively enthusiastic. The rest are willing but waiting, and rollouts are almost always designed for the enthusiasts, because they’re the ones who volunteer, turn up to the demo and give encouraging feedback. Build your plan around them and you’ll mistake their energy for progress while the willing majority, the people who’d use it happily if someone sat with them for twenty minutes, quietly carry on as before.
Trust Breaks on the First Mistake
Here’s the psychology that catches organisations out: people forgive a colleague’s error in seconds, but one bad AI answer and they quietly write off the whole system (Current Opinion in Psychology, 2024). So design for it. Introduce AI as a drafting partner rather than an oracle, show where human input shapes the output, and let people adjust things themselves. Control is what turns a sceptic into a user.
The reason is oddly specific. People assume a colleague who gets something wrong will learn from it, and assume a system won’t. They also treat one tool’s failure as evidence about all of them, in a way they’d never do with a person. Which means expectations matter enormously at the start. Promise a machine that’s always right and the first wobble ends it. Introduce something that drafts well and needs a human eye, and the same wobble is simply the process working as described.
Increase AI Adoption by Design
The organisations getting this right share two habits: they give people protected time to learn, and they involve them in reshaping the work itself rather than presenting it as finished. The reward is dramatic. Those organisations increase AI adoption: 50% more staff use it every single day (BCG, 2025).
Adoption was never a personality trait of a workforce. It’s a design outcome, and it’s entirely within your control. Get the design right and the tool stops being the new thing. It just becomes how good work gets done here, and the day nobody mentions it any more is the day it worked.
