“How do we help them understand what they’re giving up?”

The question came towards the end of a conversation with the senior leadership team of an intellectual property firm last week.

We had been talking about AI and the difference between authoring work and approving work that something else has produced. Their attention had turned to junior team members.

They were seeing people hand work to AI that would previously have been part of learning their craft. The reasons were obvious. It was faster. It removed work that could be laborious. And if the output was good, why wouldn’t you?

Their concern was about something less visible.

Some of that work was also how people became good at what they did.

The junior team member could see the hours being saved. The experienced practitioner could see what had been learned in those hours.

Their question got me thinking… there is something slightly peculiar about expertise. By the time we become good at something, much of what once required conscious effort has become simply what we know how to do.

We can walk into a meeting and sense what is really going on. Read a piece of work and know the argument doesn’t quite hold. Know which question to ask, what matters and what can safely be left alone.

We weren’t always able to do that.

There was a time when we missed things. Paid attention to the wrong things. Followed rules because we didn’t yet have enough experience to know when they applied. Someone more experienced looked at our work and saw things we couldn’t see.

Then, gradually, we could.

Capability has a history, even when we can no longer see all of it.

That makes this an interesting moment for work.

AI allows us to separate the output of a task from much of the human activity that once produced it. The research, analysis, first draft or recommendation can arrive without someone having to work their way through its formation.

The output remains. But what else happened while we were producing it?

There is the work we can see, and then there is the work beneath the work: the repeated exposure that teaches us what to notice; the attempt that reveals what we don’t yet understand; the feedback that changes what we pay attention to next time.

We have understood this, in one form, for centuries. Apprenticeship was never simply about getting work done. The apprentice watched, attempted, was corrected and tried again. Over time, more of the work became theirs to do.

There was inefficiency built into the arrangement. The experienced practitioner could often have done the work faster.

That wasn’t entirely the point.

The work was also making the apprentice.

I don’t think that means we should preserve work simply because previous generations learned through it. Some of it may have been unnecessary. AI may give us the opportunity to build better pathways to expertise rather than reproducing old ones.

But we can only do that if we understand what those pathways were developing.

And that asks something of those of us who are already experienced. We are judging the value of work from the other side of having been formed by it.

The task that looks unnecessary to me now may look unnecessary precisely because I already possess what doing it helped me develop.

So perhaps there is another question to sit alongside Can AI do this?

What did learning to do this teach us to become capable of?

And if that capability still matters:

Where will the next person learn it now?