After enough cohorts, patterns you weren't looking for start asserting themselves, and some of them are unwelcome. Here's the one that took me longest to accept, because it flatters my curriculum least. When I check back with organisations months after a training engagement, the ones where AI use has genuinely changed daily behaviour almost always have one or two people who became the informal "go ask them" person after I left. The organisations where nothing changed usually don't. Same material, same delivery, same me at the front, different outcomes, and the differentiating variable was never on my slides.
The mechanism, once you finally see it, is almost insultingly obvious. Real adoption doesn't stall on the big concepts covered in training. It stalls on the small, specific, slightly embarrassing questions that surface in week three. This prompt worked in class but not on my actual report, why? Is it safe to paste this particular document in? How were we supposed to check this figure again? Each question is tiny, but each one is a fork in the road: get an answer within a day and the new habit survives; hit silence and the person quietly reverts to the old way of working, permanently and without announcement. An external trainer cannot be there for those forks. A colleague two desks away can. Whether that colleague exists is, as far as I can tell, the strongest predictor of whether anything I taught survives the month.
I didn't design for any of this early on, because I was carrying an assumption I never examined: that the content itself was doing the work. That a well-built curriculum, well delivered, transfers on its own merits. Every trainer wants to believe this, it puts the outcome within our control and credits our craft. It's also wrong, or at least badly incomplete, and the way I protected the belief was worse than the belief itself. When engagements succeeded, I credited my material. When they failed, I blamed "organisational culture," which is the trainer's all-purpose absolution. It took several puzzling post-engagement check-ins before I admitted both outcomes hinged on a variable I wasn't even tracking.
So now I track it, and design for it, which mostly means watching differently during sessions. The champion candidates give themselves away, and the tell isn't technical fluency. It's who other participants already turn to when they're stuck. Who answers a neighbour's question before I can get there. Who visibly enjoys the explaining. Once I've spotted them, I invest disproportionately, harder questions their way, extra depth during breaks, sometimes a direct conversation about the role they could play after I've gone. Yes, that's deliberately unequal treatment, and I've made my peace with it, because trying to make every participant equally capable is a worse use of the same hours. One durable reference point beats twenty uniformly trained people with nowhere to take their week-three questions.
Exhibit · Spotted during the session, not after
Tells of a future training champion
- 01
The one others turn to
Who other participants already turn to when they're stuck.
- 02
The one who answers first
Who answers a neighbour's question before the trainer can get there.
- 03
The one who enjoys explaining
Who visibly enjoys walking someone else through it.
If you're commissioning training for a team, the practical move costs nothing: choose your champions before the programme rather than hoping they emerge, and give them explicit permission to spend work time being asked things. That permission matters more than any curriculum decision you'll make. Including, it pains me slightly to say, mine.
If you take one thing: training sticks where a colleague is answering the week-three questions. Find that person before the programme starts. The curriculum, anyone's curriculum, won't save you on its own.
