I’ve spent a good amount of my career on training courses. Which I am (mostly) grateful for. Some were excellent; others were probably excellent but I can’t really remember them. Some involved fancy handouts, many of which I have and do not use. I always liked the ones that gave you a folder to keep things in, and a superior lunch buffet.
The training I remember when I’m actually doing my work isn’t necessarily the training that taught me the most Stuff at the time. It’s the training that changed how I thought.
Lessons that permanently change how you think
Kristina Murrin CBE, who worked in both the Blair and Cameron No. 10 teams and later founded the National Leadership Centre for senior public service leaders, ran one of the most memorable training sessions I’ve ever attended. I was only there to help with logistics while much more senior people learned about innovation. I think there was some Edward de Bono chat in there somewhere. What I remember is a conversation about creativity and why new ideas rarely emerge from sitting at the same desk, talking to the same people and reading the same things.
The lesson wasn’t that we needed to work harder; quite the opposite. It was that if you want to solve wicked problems, you need different inputs: different experiences, different conversations. Sometimes you need to stop staring at the problem altogether and go for a walk.
Years later, that’s the bit that stuck. It was nearly 25 years ago and I can still picture Kristina in front of the whiteboard.
Lessons in how to keep learning
Another memorable course involved learning to juggle and then splitting wooden boards with our hands (in between some classic post-it heavy workshops). I know this sounds less like management development and more like a hostage situation, but stick with me.
It was part of a team event led by Michael Barber, then head of the Prime Minister’s Delivery Unit. It stayed with me not because I have subsequently needed to juggle actual juggling balls in any professional context (more’s the pity). What stuck was the realisation that most new skills seem impossible until you break them into smaller, totally achievable steps.
The board-breaking exercise taught a similar lesson: visualisation and focus matter more than strength. Focus on a point a few inches the other side of the board, listen to the instructions, picture yourself doing it, then do it.
Lessons that create new habits
I also remember being shown how to align boxes easily in PowerPoint. And am eternally grateful for learning some really good keyboard shortcuts.*
These lessons have no deeper meaning attached. They were just really, really useful and I immediately adopted and repeated them until they became unconscious habit.
What kind of lessons do we need now?
Organisations are currently spending a great deal of time worrying about how to build AI capability – rightly. But if we’re honest, we don’t really know what good AI training looks like yet, which is why we’re going to keep checking in on this every six months or so.
An awful lot of people have the same two or three years of generative AI experience, and they’re running courses written six months and fourteen model releases earlier. Everyone’s doing their best, but it doesn’t feel like we’re anywhere near the AI training pinnacle we all deserve.
I’m not even sure we’ve agreed on what people need to learn. Do they need to understand large language models? Do they need prompting skills? Do they need to know the risks and governance rules? Or do they need something much more fundamental: a different way of thinking about work?
The evidence is pretty murky.
Experts have spent decades worrying about something called “learning transfer”: whether people actually do anything differently after a course. The reason the field exists at all is that organisations keep ‘discovering’ that attending training and changing behaviour are not the same thing. We’re much better at delivering courses than measuring whether they changed outcomes. See this discussion from the CIPD on learning transfer if this has piqued your interest:
https://www.cipd.org/uk/knowledge/podcasts/transfer-of-learning/
It feels particularly relevant to AI, because when I look at the people who have become genuinely effective AI users, very few of them got there through formal training. Me included. Most of the people I know who use AI every day learned it because they had a problem, tried something, it worked, then they tried something else. And because they were curious, read a few articles and talked to people who were also on the AI road.
The learning happened inside the work, not at an external training centre.
The research tells us that’s not unusual. One recent study from Cornell University found that employees largely ignored formal AI onboarding materials and instead learned through experimentation, observation and conversations with colleagues. In other words, they behaved exactly like the first humans to pick up a sharpened rock, hit something with it, then think of better things to hit with it.
Learning that no one needs prompt training
Many AI training courses still spend a lot of time teaching prompting. It can be useful; like learning PowerPoint shortcuts is useful. But I suspect prompt lessons are the equivalent of teaching somebody where the pedals are in a car. Necessary, certainly, but not really the main event.
The bigger challenge is helping people change how they approach tasks that have stayed largely the same since the advent of email. That’s where the real productivity gains come from – not from writing slightly better prompts (while the AIs get rapidly better at working with bad prompts). It comes from deciding to ask AI for a first draft instead of staring at a blank page for a bit.
That’s much harder to teach in a formal sense, but much easier to remember once you’ve experienced it. It’s also why some people become AI power users while others never really get started: a change in habits. The popular explanation is confidence, age or technical ability, but I’m nearly 50 and have a degree in film studies, so we can probably rule out at least two of those assumptions.
The evidence points elsewhere. A study covering more than 36,000 European workers found that AI adoption was correlated with workplace culture, opportunities to experiment and the amount of autonomy people had over their work. In other words, people don’t adopt AI just because somebody showed it to them; they adopt it when they have permission to experiment with it. And that’s about seeing somebody they trust try it first.
So what does good AI learning look like?
If I were spending a training budget today, I’d check first if the rest of our training offer needed an AI-flavoured upgrade, and probably spend less of the budget on classrooms and more of it on short practical demonstrations linked to real work, a few AI champions who help colleagues solve actual problems, opportunities for teams to share what they’re learning, time and permission to experiment safely and exposure to people and organisations doing things differently.
The first few teach skills; the last two teach confidence. And confidence may be the thing we’re actually trying to build.
Then I’d want people to understand the concepts that stop them making avoidable mistakes: where their data goes; the difference between an AI model and a database; why AI sometimes sounds confident when it’s wrong; when to trust it and when not to; what good verification looks like; and how to recognise tasks where human judgement still matters. These are rapidly becoming the digital equivalent of understanding what “Reply All” does before you press it.
Maybe that’s the real test of good AI training. It’s not about whether people leave with twenty new prompts or a certificate saying they attended, but whether they come back to work and think differently. The training courses I still remember after twenty-five years didn’t fill my head with information; they changed the way I approached problems. If AI learning can do the same, we’ll probably look back on prompting in much the same way I now think about PowerPoint shortcuts: genuinely useful, but not remotely the most important thing I learned about how to communicate.
*Really good keyboard shortcuts. I have a personal top three – partly from years as a sub-editor trying not to give myself RSI by overusing the mouse. In reverse order they are: Shift + arrow key to select chunks of text; Ctrl + Shift + F3 to toggle highlighted text between lower case, all caps or capitalising the first word (Because Fully Capitalised Headings Are An Annoying Feature Of US English And AI-Generated Content so I have to get rid of them); then obviously the king of shortcut skills is the copy, cut and paste trio of Ctrl + c, Ctrl + x and Ctrl + v. But you knew that.
