Rolling AI out to the whole team, not just the keen ones
Every organisation has a few people who adopted AI on their own. The hard part is everyone else. Company-wide adoption depends on role-specific training, clear rules, and measuring use rather than licences.
In most teams, AI adoption starts with a handful of enthusiasts. They find the tools themselves, build their own prompts and quietly get faster. Leadership sees the results and buys licences for everyone. Three months later, usage reports show the same handful of people doing most of the work, and a long tail of accounts that were opened once.
That pattern is not a sign that the rest of the team is resistant. It is a sign that the rollout was designed for people who were already convinced.
Why the early group does not scale
Early adopters learn by experiment. They tolerate wrong answers, enjoy trying things and make time to explore. Most people have none of those conditions. They need to see AI work on their own tasks, in their own systems, with a clear sense of what they are allowed to do and how to tell when the output is wrong. A general demonstration of what the tool can do answers none of those questions.
Licences measure what you bought. Usage measures what changed.
Train by role, on real work
Replace the single all-hands session with short, role-specific sessions built on the team's own documents. Finance works through a variance commentary. HR drafts a job description against a real policy. Operations summarises a real incident log. Each session should end with two or three tasks people will do with AI that week.
Teach checking alongside using. The most valuable thing a new user can learn is how their tool fails on their work: where it invents detail, where it misses context, and which tasks it should not be given at all.
Publish the rules before the rollout
People hold back when they are unsure what is allowed. Before launch, publish a short page: which tools are approved, which data can go into them, what must be reviewed before it leaves the team, and who to ask. Clarity does more for adoption than encouragement.
Use the early adopters as coaches
Your enthusiasts are the best resource you have, but only if their time is formally allocated. Give each team a named go-to person with a few hours a month to help colleagues, collect prompts that work and report what does not.
Measure use, not licences
Licence counts and logins say little. Track the share of people who used AI on a real task in the last two weeks, the specific tasks that changed, and the time those tasks now take. Review it by team at thirty and ninety days, and put follow-up training where usage is flat.
- Drawn from Biruni Applied AI engagements and training programmes, not from published research.
- —Rollouts designed for early adopters leave most of the team behind.
- —Train by role on the team's own documents, and teach checking alongside using.
- —Publish approved tools and data rules before launch.
- —Measure active use on real tasks, not licences.