The Efficient Company Is Not the One With the Fewest People

There is a boring question that keeps coming back: will AI take all the jobs?

The people who build with AI every day have mostly stopped asking it. They are watching the opposite happen.

Dan Shipper put the thesis on the table at Every. After automation, there will be more human work than ever.

That sounds like a contradiction. It is not.

The residue of expertise

AI gets good at whatever can be made explicit enough to train on. That is a real thing and it is valuable.

But here is the part that matters. As soon as something becomes explicit, the default output turns cheap. And when the default gets cheap, demand moves to what is different from the default. What is different is the human expert.

So the more you automate, the more expert judgment you need. Not less.

The reason is plain. An agent can do the stable, repeatable, well-framed layer of work. It cannot decide what to point itself at. It cannot judge whether the output is good. It cannot catch the places where it is wrong. It cannot turn a result into a real decision about a real customer, a real gym, a real supplier.

Someone has to be in charge of making sure it works. The further an agent gets from a human who owns the outcome, the worse it works.

The efficient company

That is where efficiency actually comes from. The efficient company is not the one with the fewest people. It is the one where agents do the repeatable layer and the humans spend their attention on the decisions that matter.

Agents absorb the busywork. Humans do the judgment. You end up with more done per human, not fewer humans doing less.

I have seen how this shows up in the work. AI can build the software, but it does not know how our staff works, what our customers ask, which mistakes are expensive, or where trust can break. I have to know the shape of the work first. Then AI helps me build the system around it.

The judgment stays mine.

What stays human

The jobs that survive need ambiguity and judgment and people. Managing. Communicating. Overseeing the agents. The creative work no model can plan for.

There is a common worry that AI makes people interchangeable. I think it is the opposite. As knowledge gets commoditized, what makes a person indispensable is not knowing more. It is judgment. Taste. The ability to read a situation and decide what matters right now.

That is the part I am actually optimistic about. AI does not have to compete with us on the things a training corpus can know. It frees us to do the parts that are alive. The part that is you.