Whenever a large change starts to take shape, we tend to reach for its most extreme version. Anything that has a firm edge (a date, a cause, a number of people affected) is the easiest to hold. The conversation about AI and work has settled into two positions that follow this pattern: either the jobs disappear for the 99%, or the technology becomes ungovernable and overthrows all humans. Both are simple enough to repeat over dinner, and both give you something definite to feel.

What I see happening is smaller than either of these stances, and it cuts deeper for being small. It takes the shape of a steady chiselling away at the value of ordinary, competent work, and it is moving fast. No one chip is large enough to count as an event, so it passes unnoticed, avoiding the argument we are busy having.

Dan Shipper is a useful example, partly because he is one of the most thoughtful voices on the reassuring side, and partly because his own figures show the carving clearly. Shipper runs Every, a media and software company that has built AI into almost everything it does. On Lenny's Podcast he said the job apocalypse is not what he sees from where he sits. His headcount doubled in a year, from 15 to 30. The explanation given is that every automated system needs a person to look after it: someone to give it context, catch its mistakes and keep it pointed at what the business needs. In his experience, more automation has meant more human work.

I believe him. The more useful question is who that work is for.

In the same conversation he described testing the latest model against his senior engineers on the rewrite of a failing codebase. The model scored 62 out of 100, and the engineers scored between 85 and 90. Two model generations earlier, the machine had scored around 30.

His team is chosen with great care and works with new models months before most of us see them. When Shipper talks about the humans every agent needs, those humans do not come from across the full bell curve; they are people who are already very good at what they do. What is happening inside Shipper's workflows tells us a good deal about the top of a profession and much less about its middle.

Consider the engineer whose own work would have scored somewhere in the sixties. Two years ago that person was comfortably useful, the reliable and employable colleague most teams are built from. In that time the model has climbed from 30 to 62, and it will keep climbing. Nobody has taken that engineer's job, yet each new release makes it a little harder to justify the human choice, as the room between what the model and an engineer can do gets narrower.

This is what the chipping away looks like in practice. As the gap closes between machine and skilled human, the organisation is carved towards its experts, though nobody sat down and chose that shape.

The steps a person used to take become offcuts, and they fall away along with the person who would have done them: a junior post that is not refilled, a contract that quietly lapses, a team that stays the same size while the work grows, or a job advert that asks for a slightly different person than it did last year. Each of these is a reasonable decision on its own, and none of them is newsworthy, because none of them has an edge.

Shipper is right that AI creates more work. It creates it for the people who can see where the machine is wrong, and it slowly lowers the standing of those who cannot. For anyone responsible for an organisation or a portfolio, that distinction matters more than any headline figure, because most organisations have been built from their middle.

So when you next hear the question framed as a job-apocalypse vs. tyrant machines, it is worth asking a smaller one: what is happening to the good-enough professionals most institutions depend on? The answer is around us, in a pile of small decisions that have fallen to the floor, and by the time it is large enough to be reported as news, most of it will already have happened.

Sources

The link has been copied!