How to Get Your Team to Actually Use AI

The part that is not about tools

You can pick the right use case, build an honest business case, write a clean one-page policy, and still watch the whole thing fade out by week six. Adoption fails for human reasons far more often than technical ones, and those reasons are all in the room with you, mostly unspoken.

There are usually three: someone who thinks it is nonsense, someone who thinks it is magic, and everyone quietly wondering whether they are training their replacement.

The sceptic

The sceptic tried an AI tool once, got a confidently wrong answer, and drew a conclusion. Or they have watched three transformation programmes arrive and leave, and they are pacing themselves.

Do not argue. Ask what they tried and what went wrong. Often they gave it one vague line and got exactly the output such a line deserves. Sometimes the task genuinely was not suitable, in which case they have just done part of your audit for you.

Then make a narrow, personal offer: pick the single task they most dislike, sit with them for fifteen minutes, and make that one thing better. A sceptic who has personally saved forty minutes becomes your most credible advocate, because everyone knows they were not predisposed to like it.

The over-enthusiast

Harder to manage, and more damaging, because the problem arrives disguised as progress. This is the person generating volume: long documents, elaborate summaries, plans nobody asked for, all produced fast and none of it checked. They are cheerful, visible, and slowly filling your team with slop.

Do not dampen the enthusiasm, because if you squash it you will not get it back. Redirect it. Make them the person who verifies output before anything goes out, or ask them to build the team's prompt library from what actually works. Then be direct about the standard: the measure is not how much you produced, it is how much of it survived review unchanged.

The fear nobody says out loud

Here is the one that decides whether any of this works. Some of your team are worried about their jobs. They will not raise it, because raising it looks like insecurity and invites exactly the attention they are trying to avoid.

Pretending nobody is worried does not work. It is visible, it is a little insulting, and it means the conversation happens without you, in messages you are not part of, with worse information than you could supply.

Say it plainly instead, in your own words, and then answer three questions honestly.

Do I have plans to reduce the team? Answer with the truth. If the answer is no, say no clearly. If you do not know, say you do not know, which is uncomfortable and still better than a reassurance that turns out to be false.

What happens to the time we save? Name the actual work it goes into.

Will you tell us if it changes? Yes, and then you have to mean it.

I manage a team of [9]. We are starting a small AI trial on [two tasks].
I want to address the worry about job security honestly.
My actual situation: [no redundancy plans / plans I cannot discuss yet /
genuinely do not know].
Give me:

Three sentences I can say that are honest given that situation
Three questions my team is likely to ask but may not ask out loud
For each, an honest answer, including how to say I do not know when I do not
Do not give me reassuring language that promises things I cannot control.

Make it safe to be bad at it

A quieter blocker sits underneath all three types. Trying a new tool in front of colleagues means being visibly incompetent for a while, which is why the most experienced person on your team is often the last to try anything.

Fix it by going first and being bad at it publicly. Show the prompt that produced rubbish. Say what you changed. Five minutes of you failing in front of the team beats any training session, because it moves the standard from get this right to have a go.

Design a 15 minute team session where nine people each try one AI prompt on
their own real work. Constraints:

No slides, no theory, no tool demo
Everyone leaves having produced one thing they might actually use
Include the exact instruction I read aloud to start it
Include one question I ask at the end that surfaces what did not work
Assume mixed confidence, including two people who have never used these tools.

Let it spread sideways

You are not the most persuasive person on your team about this. You are the boss, which means people agree with you in the room and do as they like afterwards. The person who genuinely shifts behaviour is the colleague who says this saved me an hour on the Thursday report, look.

So make the wins visible and let other people own them. Two minutes at the start of a team meeting, a named person, one specific thing, what it saved. Not a slide. Not you presenting it.

Turn this into three plain lines for a team meeting:
what the task was, what changed, how much time it saved.
No superlatives, no marketing language, no exclamation marks.
Keep the person's own words where you can.
WHAT THEY TOLD ME:
[paste it]

Six weeks of small visible wins will do more than any launch. Adoption is not an announcement, it is a habit, and habits spread by demonstration.