Why one-off AI training workshops don't stick, and what does

The workshop that changed nothing

You book an external expert, clear an afternoon, and the whole team comes out buzzing. They've seen some impressive demos and written a few good prompts, and everyone has a list of things to try. A month later, almost nobody is doing anything differently. The tools are still open in a tab, used for the odd email. The workshop wasn't bad. It just didn't change how anyone works.

This is one of the most common ways companies spend an AI training budget, and one of the most common ways they waste it. The trainer and the content are usually fine, but the workshop format itself works against them. A one-off session is built to inform people. That isn't the same as changing what they do on a Tuesday afternoon with a deadline looming.

What the research says

None of this is specific to AI. Learning and development has known for well over a century that people quickly forget what they only see once. Hermann Ebbinghaus's 'forgetting curve' describes how memory of new material drops sharply in the days after learning it, unless you revisit it and use it. A workshop is a single exposure by definition.

The second problem, and the harder one, is transfer: getting a skill out of the session and into the day job. Adults tend to learn a skill by doing the task itself, getting feedback and doing it again soon after. Watching a demonstration once doesn't do it. If the first proper attempt comes long after the session, with no feedback in between, the skill often stays in the training room.

AI arguably makes both problems worse, because the tools change month to month. A prompt technique shown in a workshop can be out of date before the team has built a habit around it. General fluency and good judgement about when to reach for AI hold up much better, and both are built by repetition. That repetition has to carry on after the trainer leaves, which is what an internal network of AI champions is for.

The same logic is why you should measure AI training by capability and usage rather than attendance. A full room on the day tells you very little about behaviour four weeks later.

What a programme that sticks looks like

If a single session is the problem, a longer single session won't fix it. You need a different shape:

• Spaced over weeks instead of squeezed into a day, so people practise between sessions and come back with real questions.
• Run on the team's own work, not canned exercises, so the skill gets used on the exact task it has to transfer to.
• Reworked between sessions around wherever people are getting stuck, rather than delivered from a fixed script whatever happens in the room.
• Tied to a few workflows each person owns, so you can answer 'did it stick?' honestly by looking at their work afterwards.

This takes more effort to design and deliver than a keynote, and that's exactly why it works. The reinforcement between sessions is the point. It's how we run our hands-on AI training as standard.

What this looked like in practice

We used this shape for a six-week AI programme we delivered with a training partner for a Fortune 500 tech company. Around twenty senior product managers and UX designers took part, live online. Every session was hands-on with their own product work, using real PRDs, user feedback and prototypes.

The part that mattered most for stickiness was the least glamorous. Between sessions we ran content syncs and reshaped the remaining weeks around what the cohort was struggling with. We split dense sessions in two and cut abstract material in favour of working sessions. The group ranged from AI beginners to experts, and everyone had to be able to follow. You can't do any of that in a one-off. There's no 'between sessions' to adjust in.

We won't put a headline number on it here, because the honest measures of a programme like this are capability and habit. Those show up in the work itself, in the weeks afterwards. The structure is the part worth copying. It was built to change behaviour, and that needs time and repetition designed in from the start.

How to brief a provider so you don't just buy a workshop

You might be commissioning the training on its own or as one part of a wider AI rollout. Either way, a few questions at the buying stage sort the programmes from the performances:

• Ask how the training is spaced and what happens between sessions. 'One full day' is a warning sign on its own.
• Ask whether sessions run on your team's own work or on generic examples. Generic examples rarely transfer.
• Ask how the content adapts mid-programme if a group struggles, and how the trainer would even know.
• Ask what they measure afterwards: attendance, or capability and usage. Attendance is not a result.

If a provider can answer those four clearly, you're most of the way to training that sticks. If the honest answer to all of them is 'it's a one-day workshop', at least you know what you're buying. To talk it through for your own team, book a discovery call.