When is it worth training the whole team in AI?

"Can I ask something that might be stupid? Isn't it enough that we give people the tool?"

The question came from a marketing director at a fairly large company. It isn't stupid. It's the only real question in this whole subject.

I sell training, so read what follows knowing that. But I've run enough sessions to know when they're wasted money, and it happens more often than anyone in my line of work likes to admit.

Here's how I'd think about it if it were my budget.

What you're actually paying for

Not knowledge about tools. Worth saying first, because it's what most people think they're buying.

Tool walkthroughs age in weeks. The interface changes, the model gets swapped, the button moves. Most of what I can show you about how a product looks today is out of date before the participants have used it on anything real.

What you're paying for is two other things.

The first is permission. At most companies nobody has said what's acceptable to put into an AI tool. Not no, just nothing. In that silence the careful people do the reasonable thing and abstain entirely, while the less careful paste in whatever they like and nobody knows about it. One session where someone from management stands in the room and says "this is how we do it here" settles it in an hour.

The second is a shared starting point. Right now you have people saving five hours a week and people who have never logged in, on the same floor, with the same tools available to them. That gap widens on its own, because the one who's already going gets faster every month while the other stands still. The only thing that closes it quickly is putting them in the same room once.

When you shouldn't book anything

This is the part my industry skips.

If you haven't bought the tools yet, wait. Training on something nobody has access to is gone within a month. The order is licences first, training second, ideally within a few weeks of the rollout.

If only a handful of people need it, don't pay for a full session. Let those three work at it for a quarter and then tell everyone else what they found. That's better and cheaper than anything I can deliver.

If management isn't willing to change how the work is actually done, skip the whole thing. I've run sessions where everyone was engaged, understood it, and then went back to a process nobody was allowed to touch. It was a pleasant day and worth nothing.

And if what you want is a receipt proving you're doing something about AI, don't buy it from me. There are cheaper ways to look busy.

When it does pay off

Three situations, and they share a shape. The tool exists, the willingness exists, and something is stuck between them.

You've rolled licences out to everyone and the usage stats are embarrassing. This is the most common one. The money is already spent. The only question left is whether it does anything.

People are already doing it, quietly and in five different ways. Now you're sitting on a risk and an unused asset at the same time. A session makes the practice visible and shared, which is cheaper than finding out afterwards what ended up where.

You're facing something concrete. A reorganisation, a new account, a stretch with fewer people doing the same work. Training that hangs off a real deadline sticks. Training because we probably should doesn't.

Do the arithmetic once

Take the number of licences, times the monthly cost, times twelve. That's your annual AI budget, and it's probably larger than you're carrying in your head.

Put half a day against that number. For a team of thirty, one session typically lands at a small fraction of what the licences cost over a year.

That isn't an argument that training is cheap. It's an argument that the large part of the budget is already decided, and the thing that determines whether it does any good at all costs a fraction of it.

How to tell whether it stuck

Don't ask participants whether they were happy on the way out. Everyone is happy on the way out. It tells you nothing.

Look six weeks later, at three things.

Has licence usage gone up, and has it stayed up? That's the only hard measure you have.

Have the questions got more specific? "How do I stop it sounding like a brochure" is a better sign than "which tool should we be using".

Is there something in writing about what may be entered, and do people know where it is?

If the answer is no on all three, it was a pleasant day. Don't book another session. Find out why instead.

If you want to see how I structure a session, and what participants leave with, it's on the services page. And if you're unsure whether you're even in the right situation for it, say so plainly when you get in touch. I'd rather tell you to wait than sell you a day you get nothing out of.