If AI gets any better, that whole department can go

You hear it on the shop floor more and more: “If AI gets any better, that whole department can go.”

Sometimes as a joke. Sometimes because somebody has just worked out what is possible now. And sometimes in dead earnest.

I understand where it comes from. Look at what AI already does and extrapolating is a reasonable thing to do. A task used to take an hour, now it takes ten minutes. Work that needed specific expertise can suddenly be done by almost anyone. And things we said two years ago that AI would never manage are now a button in an application.

So yes. Work is going to disappear.

But a whole department? Something is off there.

A department is not a collection of actions

A department does not exist because we once decided that twelve people should perform certain actions for eight hours a day. A department exists, ideally, because something needs to happen.

Customers need helping. Invoices need to be right. Products need thinking up. Risks need spotting. Things need selling. People need hiring. Decisions need making.

That purpose does not disappear when AI makes the work faster.

Maybe we confuse the work too easily with the way we have organised it today. And frankly I find that more interesting than the question of how many jobs AI is going to replace.

Suppose AI can do eighty percent of a department’s current work. Why is our first thought that eighty percent of the people can go?

Why not that the same department can suddenly do five times as much? Or the things it never gets round to?

We built everything around scarcity

Look at the choices that were made in your organisation. Not the strategic ones. The practical ones, the choices nobody ever decided on, because they made themselves.

We check this on a sample basis. We call that customer once a year. We look at that report when something goes wrong. We will do that review next quarter. That idea stays on the shelf because nobody has time.

Not one of those choices was made because it was the best way. They were made because there are only so many people, so many hours, so much knowledge. We have spent decades designing organisations around a shortage, and that shortage has become so normal that we no longer see it as a choice.

And that is exactly what AI changes. Not the actions. The scarcity underneath them.

In my own trade it is clearest. We hold the keys to other people’s systems, and there is always more to check than there are hours. So we check what is most likely to be broken, and the rest we find out about when it goes wrong. Everyone in this industry knows that, and nobody thinks it is good. It is simply what fits.

The poorest version of this story

And of course this has consequences for how many people you end up needing. It would be naïve to pretend otherwise.

But that may be exactly where the risk sits.

That we get a technology which lets one person do the work of five, and our grandest fantasy is then to make four people redundant.

There is a practical problem inside that fantasy too. If you let four people go, your competitor does the same. You have both become equally efficient, you have both kept the same margin, and the only thing left to compete on is price. That is precisely the fight you do not want, and you used a technological breakthrough to walk into it.

The other route is harder to explain to a shareholder, but more interesting. The same four people, and suddenly time for the things that always stayed on the shelf. Calling every customer instead of once a year. Checking everything instead of a sample. Finally working out that idea nobody got round to.

That is not a soft, sympathetic thought. One route delivers the same service more cheaply. The other delivers a better one.

Redundancy is technically efficient.

But in terms of ambition, rather poor…