Sam Wong.黃力桐Let’s talk training

TRAINING DECISIONS

Why your AI champions give up within a few months

Published:

In 2024, I was the AI champion.

At RENPHO, I had been executive assistant to the CEO. No IT background. I pitched my boss, got the chance to lead AI in the company, and ended up as Leader of the AI Division.

Part of that job was sitting down one-on-one with more than 300 colleagues across departments, to understand how they felt about AI and where they got stuck. Not everyone was curious. One colleague told me to my face: this so-called high-tech stuff, I don't know it, I'm not interested, don't ask me to use it.

I remember that sentence because it was honest. And because the colleagues who did want to learn needed something from me every week: time, answers, a reason to keep going.

That is the job a company hands to its AI champions. Most of them get the title without the time, the method or the measures that should come with it.

The hand goes up in month one

In every programme I run, a few people lean in. At Garden, the Hong Kong food manufacturer, seven of the 19 colleagues in the first cohort volunteered to champion AI in their own departments.

Seven out of 19 is a good start. It is also the easy part.

Raising a hand costs nothing in the week after a good workshop. Everyone is still excited. When I ran a follow-up clinic for the surgeons at the College of Surgeons of Hong Kong, I described the problem like this (translated from Cantonese):

"In the few days, or the week or two, after a workshop, people should be very motivated to try things and learn a bit more. But after a week, two weeks, a month, two months, what then?" — Sam Wong, CSHK Q&A Clinic, 5 October 2026 (01:21–01:36)

For a champion, the answer often comes within a few months. The excitement has worn off, the day job is back at full speed, and the extra role has no owner except the person doing it.

Why they stop

They are doing two jobs

Dan Cumberland put it plainly in October 2026: champions are "running two jobs simultaneously: their real job, and the job of making everyone else better at AI."

The second job rarely comes with time in the calendar. In July 2026, the Conference Board found that only 48% of workers agree their organisation gives them enough time during work hours to build AI skills. If the learners don't get time, the person helping them usually doesn't either.

When the quarter gets busy, the AI help stops first. Nobody decides that. It just happens.

Knowing AI is not the same as teaching it

Most champions are picked because they are good with AI. That makes them useful to sit next to. It does not make them able to run a session for 20 colleagues, build an exercise around someone else's work, or recover when the tool fails in front of the room.

In Adaptig's trainer programme, where I teach, the most common habit of a new trainer is walking through every slide. Champions do the same, except nobody has taught them anything else. So they answer the same question for the tenth time, one colleague at a time, and it wears them out.

Nobody measures what changed

If nobody counts what a champion changed, nobody notices when they stop. BCG's 2026 AI at Work survey found that 66% of frontline employees who use AI regularly get limited or no guidance on what to do with the time it saves. The time leaks away, and so does the case for the champion's role.

What I would put in place before month one

None of this needs a big budget. It needs decisions made before the first session, not after the first champion goes quiet.

Write down the time. Agree with each champion's manager how many hours a week the role gets. Cumberland suggests at least two to three hours a week, separate from existing work. Your number may be different. What matters is that it is written down.

Citi shows what structure looks like at scale. According to Business Insider, about 25 to 30 champions lead around 4,000 volunteer "AI accelerators" across the bank, and its CEO reported adoption of the firm's own AI tools above 70%. That is a design, not a mood.

Teach them to teach. Give each champion one session to prepare and run for their own colleagues, with feedback afterwards. A short session on a task their department already does is enough. After one real session, they stop being "the person who knows AI" and become the person who runs the AI session.

Settle the rules first. Champions get stuck on "am I allowed to?" more often than on prompts. At Garden, the IT directive allowed only Microsoft Copilot on the web and Perplexity, and the most confidential tier of data was barred from AI tools altogether. Every session opened with a three-tier data card. A champion who can answer the rules question in one sentence saves everyone a week. I wrote more about staff AI data rules in Hong Kong.

Give questions a home. At HKCT, a whole-college talk turned into a monthly AI learning community, with more than 200 members in its WhatsApp question group. Later monthly workshops were run by other trainers on the same programme. That is the shape: a place to ask in the moment, and a regular session nobody has to reinvent.

Count behaviour at day 90. Sessions run, colleagues reached, one task per champion timed before and after, questions asked, and whether the next department asks to join. If those numbers move, the role is working. If they don't, you find out in month three, not month nine. I explained why I stopped trusting satisfaction scores in why AI training doesn't stick in large organisations.

What happened after I left

I left that company in 2024 to teach AI on my own. A few months later, former colleagues were still messaging me. Some told me the AI projects we had started were still helping their work. A few had carried them further and trained up more colleagues themselves.

That is what a champion is for. Not to be the AI person forever, but to start something that keeps growing after they step back.

Seven people raised their hands at Garden. The hand is the easy part. What you build around it decides whether they give up within a few months.

Want your AI champions still going after a few months? I train a small group of your own staff as in-house AI trainers, on your approved tools and your own work. Tell me how many staff you have and which AI tools you're allowed to use. WhatsApp me · How the in-house AI trainer programme works

FAQ

Why do AI champions burn out?

Usually not because they lose interest. The role sits on top of a full-time job, nobody teaches them how to teach, and nobody measures what they change, so their work stays invisible.

How much time should an AI champion get?

Agree it in writing with their manager before the programme starts. One practitioner recommends at least two to three hours a week, separate from existing responsibilities.

What is the difference between an AI champion and an in-house AI trainer?

An AI champion is usually an enthusiastic user whom colleagues ask for help. An in-house AI trainer has also learned to teach: they run sessions for their department, prepare exercises on the team's own work and keep a question channel going.

Sources

Sam Wong (黃力桐) is a co-founder and Director of Academy at Adaptig. He designs AI training for teams and trains the trainers who deliver it.