What Being a Teacher Taught Me About ‘AI Champions’

Naming someone a champion doesn’t make a team more capable. I’ve been the champion twice, once where it worked and once where it didn’t, and the difference had nothing to do with enthusiasm

The UK government’s AI Adoption Plan for Digital and Technologies names “AI champions” as a way to spread AI skills through government teams. The idea is that champions build capability in their colleagues, not just in themselves.

I like the idea. I’ve also been the champion, twice, and I’ve seen the model succeed and fail. The two experiences were close to identical on the surface, and that’s what makes them useful.

What champion schemes usually get wrong

A champion scheme usually starts with one enthusiastic person. They get a title, perhaps a Slack channel, and an instruction to go and make everyone else good at the thing.

That person can run a brilliant workshop, and people will say nice things afterwards. But a workshop is an event, and capability is a habit. If nothing in the organisation changes after the event, the enthusiasm drains away within a few weeks and everyone drifts back to how they worked before.

The trouble is that this failure looks like success. Attendance is high, the feedback is warm, and a slide in a quarterly review says “trained 200 staff”. None of that tells you whether anyone is working differently.

The school: great feedback, no traction

For a while I was the IT coordinator at a school, with a brief to champion online learning and digital literacy among the teaching staff. I used a range of packages to teach, to develop skills and to record progress, and I could show clear benefits for each one. The staff training I delivered was well received amongst those who attended, and the feedback was excellent.

In two years, almost nothing changed.

There were no extra computers for students to use, so a teacher who wanted to try a new approach had nowhere to try it. There was also no way to put digital tooling into anyone’s learning plan, which meant that using it was always an extra, on top of the job as it was already defined. Leadership partly turned up to the workshop, but did not compel others to train. A teacher with a full timetable is going to do most of the things they’re accountable for, and nobody was demanding competence.

I did the part that was mine to do, and the system around me did nothing with it.

The camp: the same gap, a different outcome

Later, as a communications technician in the military, I noticed a similar gap. People leaving camp were being issued with satellite phones, and few of them could use one. The issue was a tick-box exercise: sign for the kit, take the kit, go. In practice, without a person who could use the phone, those phones were useless boxes.

So I organised a programme, then rolled it out to train everyone in the camp. Satellite phone use went into the core “welcome to theatre” training. Within a short time, issuing the phones had stopped being a box-ticking exercise; they had become equipment people could rely on when it mattered.

The problem was the same as in the school: a tool was available, people weren’t confident with it, and a formality was standing in for competence. What differed was everything around me. Leadership bought in and demanded better. The equipment was already there. And the training became part of what everyone was expected to do, rather than an optional extra for the keen.

My ability and enthusiasm were roughly the same in both places. The conditions weren’t.

Three principles that transfer

Looking at the two side by side, three things separated real capability from the appearance of it.

• Deliberate practice. People get good at a tool by using it, repeatedly, on real tasks, and with feedback. Watching someone else use it doesn’t do that. The satellite phone training involved picking the phone up and making the call.
• A clear minimum standard. Everyone knew what “can use this” meant, and it was something you could check. A standard turns “I went on the course” into “I can do the thing”.
• Someone noticing when it isn’t working. There needs to be a line of accountability so that when practice slips or the standard isn’t met, someone sees it and acts. At the school nobody was in that position. In the camp it was part of the job.

None of these depends on the champion’s personality. They depend on structure.

What this means for a software team

Take a software team or a consultancy that wants its people to get real value from AI tooling. The version that fails is easy to picture: one person who is excited about it, a lunchtime demo, a shared channel full of tips, and everyone else carrying on as before.

The version that works looks more like the camp. The team agrees what “competent” looks like, for example, being able to use an assistant to draft tests, summarise a pull request or investigate an unfamiliar codebase, and can show it. The practice happens on real work, in work time. And somebody, probably the line manager rather than the champion, notices when a person has quietly stopped using it and asks why.

The champion’s job in that setup is to be a good teacher, which is a skill in its own right. The structure is somebody else’s job.

What I’d want to see in a government scheme

If I were designing or joining an AI champion scheme in government, this is what I’d look for.

• Time. Time set aside for people to learn how AI can speed up their own role. If it only happens in spare moments, it won’t happen.
• Leadership buy-in. Leaders who ask for it, and who expect people to hand the automatable parts of their roles over to AI. A champion can’t create that demand from the middle of the organisation, and in my experience it’s the single biggest difference between the school and the camp.
• Resource. AI credits made available to staff, with the choice of provider left to them. People learn fastest when they can experiment with the tool that suits them, rather than waiting for one approved option.
• A shared kata. In martial arts a kata is a short, repeatable practice sequence. The same idea works for AI: small, shared exercises that show people what the tools can and can’t do, and that they can repeat as the tools change. This is a learning journey, and kata give a team something concrete to share along the way.
• Boundaries. Clear guidance on what should still be done by a human. People say “accountability” a lot, but it means little unless people understand the consequences and follow up on poor use. A boundary that nobody checks is only a suggestion.

The test

When I think about both of my experiences, the question I’d put to any champion scheme is simple. If the champion left tomorrow, would anything keep going?

At the school, the answer was no. In the camp, it was yes, because the practice, the standard and the accountability had been built into how things were done.

Champions can start something good. Whether it spreads depends on what the organisation builds around them.

This post was first published on Medium on 16 October 2026.