Jobs aren’t disappearing: tasks are

Everyone asks which jobs will disappear with artificial intelligence.

That’s understandable, but I think it’s slightly off target. At first, what vanishes isn’t jobs as labels on an org chart: it’s tasks, the smallest units of work that make up a day.

And when enough tasks disappear, even if the job title stays the same, the number of people needed to get the same outcome tends to fall.

What I mean by a task

A task isn’t a formal duty in HR language. It’s something concrete: copying data, answering with a script, building a standard slide, doing first-pass screening, producing a text variant, running a repetitive check, filling a form, triaging a queue.

Many modern roles, especially in front of a screen, are chains of tasks stitched together. Some require judgment, context, accountability. Others don’t, or barely.

That’s where AI is strongest today: where there are patterns, rules, examples, fast iteration, and where the cost of mistakes is manageable.

Intellectual work isn’t over, quite the opposite. But a huge slice of everyday knowledge work in companies isn’t high-level at all: it’s mechanical, even when you do it with Excel and four browser tabs.

Losing a trade is a different story

If the blacksmith trade disappears tomorrow, you see it: the shop is gone.

If 30% of an office’s tasks vanish, what often happens is this: the same people remain, but the team shrinks over time, or the hiring that growth would have required never arrives, or the load shifts onto whoever is already there.

The labour market doesn’t always translate that into an explicit layoff headline tied to AI. It translates into fewer hires, more pressure, higher expected output, leaner organizations.

That’s why asking which jobs will go can create a false comfort: it makes you imagine a sharp event, like a sector collapsing. Often it’s erosion: slow, distributed, hard to photograph in news titles.

Who feels the pressure first

Saying that people who work with their hands are safer is largely true in the short run, because many physical jobs live in contexts that are hard to standardize: variability, safety, materials, surprises, logistics, environment.

But don’t turn that into a comforting myth. Automation, sensors, robotics, planning software show up there too. The cycles are just different, and the task isn’t always digital.

The core point remains: the wave hits hardest where output is digital, repeatable, measurable, and where AI lowers the marginal cost of producing another piece of the same kind.

The real risk isn’t losing your job tomorrow morning

The most insidious risk, in my view, isn’t sudden unemployment.

It’s becoming replaceable without realizing it, because the value you bring is tied to work someone else, or something, can replicate with standard tools.

In that world your value doesn’t go to zero overnight: it becomes lower, more negotiable, more compressible. You end up competing on price, speed, availability, on who does it for less.

And it isn’t only an employee story. It applies to freelancers, professionals, micro-teams: if your edge is doing X that has become a commodity, commoditization arrives faster than you think.

Staying close to real complexity

So the useful question becomes: how much of your work lives where complexity is real?

Not complexity from a slide deck. I mean ambiguity, accountability, trade-offs, people, shifting contexts, expensive mistakes if you get it wrong, decisions under uncertainty.

The closer your contribution is to that zone, the harder you are to replace with a template.

AI is useful there too. But there it becomes an amplifier for people who already know how to hold the handle: connect pieces, ask the right questions, sign, own consequences.

If your role is mostly executing known steps, even if you execute them well, the defensive margin is thinner.

This isn’t an anti-tech pamphlet

AI isn’t a flood to hide from. It’s a tool, and in many cases it’s already a massive competitive advantage for people who use it well.

The point isn’t to demonize it. The point is not to confuse short-term comfort with medium-term safety.

Because transformation doesn’t ask permission: it accelerates where it pays, where there is demand, where margins bite.

Closing

Talking about AI eliminating work, in a philosophical sense, is imprecise. It transforms work, shifts the centre of gravity, and in many contexts raises the bar for what we mean by a useful contribution.

If you take away one idea: don’t look only at your role title. Look at your task list, and ask how many items on it are still hard to clone.

That list, far more than the title, tells you where you’re heading.