“Half my team uses AI daily and the other half never touches it”

Written By Rafaela Heming

An AI skills gap within a team is normal, it widens on its own, and it doesn't close by asking the second half to try harder. The daily users get faster every week; the non-users have one bad experience and stop. The way to close it is to make the first group teach the second on shared, real tasks, with no one made to feel behind.

I'm Rafaela Buena, and in the AI workshops I run for teams in Amsterdam, this split shows up in the intake forms every single time. Three people who've built their own Custom GPTs. Four who tried it once in 2024. Two who quietly think the whole thing is nonsense. Same team, same manager, same tools.

The split is real and it's wide

WRITER's 2026 enterprise survey found 70% of employees now use AI for at least 30 minutes a day. That sounds high until you look at who's left out. McKinsey's 2025 research, as summarised by Shibumi's aggregation of AI fatigue data, found only 12% of US employees had integrated AI into daily work while 49% reported never using it.

Lenny's Newsletter and Noam Segal's 2026 Tech Worker Sentiment Survey, drawing on thousands of responses, described the workforce as split almost exactly in half: one half thriving with AI, the other shaken. They grouped people into four emotional types: the Energized, the Conflicted, the Disoriented, and the Resentful.

You probably recognise all four at your next standup.

Why the AI skills gap within a team grows by itself

Compounding. The daily user learns something new every day, because they're using it every day. Their prompts get better, their setups get richer, their output gets faster. The non-user's last experience was a bad one, so there's no reason to go back.

MIT's Project NANDA research on enterprise AI called the core problem a "learning gap." The tool doesn't teach itself. Somebody has to, and in most teams nobody's been asked to.

There's also a quieter reason. The Glassdoor 2026 data found workers increasingly describe their employers as "AI pushers." If the second half feels judged for not using it, they'll avoid it more, not less. Shame is a terrible onboarding strategy.

What closes the gap

Not a mandate. Not a course. Three things.

  • Pair, don't preach. Fast Company reported in 2026 that Accenture, Target and Unilever run reverse-mentoring programmes where younger or more AI-fluent staff teach senior ones, with both groups reporting improved trust. You don't need a programme. You need one hour where a daily user sits with a non-user and does the non-user's task, together.

  • Start from the non-user's task. Not from a demo. The person who "never touches it" needs to see it work on their own Tuesday problem, in their own words, with their own material.

  • Make "I don't get this" a normal sentence. The Conflicted and the Disoriented in that sentiment survey aren't resisting. They're waiting for someone to make it safe to ask.

This is the whole design of a team workshop as I run it. Everyone brings their own task. Nobody watches a slide deck. The people who are already fluent get pulled into helping, which also stops them being bored. By the end, the gap hasn't disappeared, but the second half has a working habit and the first half has stopped being a separate species.

One thing the fluent half needs to hear

Being fast with AI isn't the same as being right with it. The business.com 2026 survey found the most disliked AI behaviours were all shortcuts: unchecked output, unedited text, outsourced thinking. Your daily users are the ones most at risk of that. The gap works in both directions.

A quick diagnostic

Ask everyone on the team, anonymously: "In the last week, what's one task you did with AI, and one you tried and gave up on?" You'll get a map. The tasks people gave up on are your workshop agenda.

FAQ

Why do some employees use AI and others don't?

Mostly experience, not ability. Early users had a good first result and kept going. Non-users had a bad one and stopped. MIT's research frames it as a learning gap, not a talent gap.

How do you close an AI skills gap in a team?

Pair fluent users with non-users on the non-user's real tasks, start from problems rather than demos, and make it safe to struggle. Companies using reverse mentoring report better trust alongside better skills.

Should managers require everyone to use AI?

Mandates tend to produce compliance without value. Give people a chosen first task, show it working on their own work, and let the habit form. Requirements can come later, once there's something real to require.

If your team has split into the fluent and the stuck, a team workshop built on both halves' real tasks is the fastest way I know to get them back in the same room, doing the same work, at roughly the same speed.

Sources

Rafaela Heming
Next

“My colleagues keep sending me unedited AI text”