Automation and Jobs
The Jobs Robots Keep Creating While Stealing Others
Every major wave of automation that was supposed to eliminate work has, so far, produced more jobs than it destroyed — and that fact is both reassuring and completely beside the point.
The Idea
The standard debate about automation and jobs is stuck in a binary: either robots take everything, or history repeats itself and new work emerges. Both camps are partially right, and that's what makes the current moment genuinely strange. The historical record is real — the mechanisation of agriculture didn't produce mass unemployment; it freed labour for factories; factories gave way to services; and so on. Economists call this the 'lump of labour fallacy': the assumption that there's a fixed amount of work to go around, which automation steadily eats away at. In reality, work tends to reshape itself. But here's the underappreciated tension: the speed and breadth of the current wave may matter as much as its direction. Previous automation was largely domain-specific — a loom replaced a weaver; a spreadsheet replaced a bookkeeper. What's different now is that AI systems are encroaching on cognitive tasks that span multiple domains simultaneously: writing, analysis, coding, diagnosis, legal reasoning. The transition costs — retraining, relocation, lost years of career momentum — are real and unevenly distributed. The jobs that emerge from AI-driven disruption may well outnumber those lost. But they won't go to the same people, in the same places, at the same skill levels. That gap between the aggregate story and the individual experience is where the real policy and human challenge lives.
In the World
Consider what happened to bank tellers after ATMs arrived. Between 1980 and 2010, the number of ATMs in the United States roughly quadrupled. Most economists predicted a bloodbath for teller jobs. Instead, the number of tellers increased. The reason, as economist James Bessen documented, was almost counterintuitive: ATMs reduced the cost of running a bank branch, so banks opened more branches. More branches meant more tellers — but tellers doing different work, focused on relationship banking and financial advice rather than counting out notes. On the surface, this looks like a clean vindication of the 'automation creates jobs' thesis. But look closer. The tellers who survived weren't the same tellers who'd been displaced. Many older workers in smaller markets didn't transition. The new teller roles required different skills, different temperaments, and were concentrated in urban financial centres. Now fast-forward to today. Klarna, the Swedish payments company, announced in 2024 that its AI assistant was doing the work of 700 customer service agents. Simultaneously, the company reduced its overall headcount significantly. New roles are appearing — in AI oversight, prompt engineering, workflow design — but they require substantially different training, and they are far fewer in number, at least so far. The teller story took decades to play out. This one is moving faster.
Why It Matters
Understanding this distinction — between aggregate outcomes and individual ones — changes how you think about your own working life, not just policy debates. If you're mid-career, the relevant question isn't 'will my profession survive automation?' It's more granular: which tasks within my role are most exposed, and which require the kind of contextual judgement, relational trust, or embodied skill that remains genuinely hard to automate? Most jobs are bundles of tasks, not monolithic entities. Radiology as a profession isn't dying; the specific task of reading a scan from scratch is under pressure, while communication, treatment planning, and navigating clinical uncertainty are not. Thinking in tasks rather than job titles gives you a sharper map. It also suggests where to invest attention: in the parts of your work that require you to be a person, not a processor. And it reframes the broader social question — not 'how many jobs?' but 'for whom, where, and on what timeline?' Those are harder questions, and they don't have tidy answers. But sitting with them is more useful than either the optimism or the panic.
A Question to Ponder
If the tasks in your current role that could be automated disappeared tomorrow, what would be left — and would that remainder feel like a job worth having?
Get a new one of these every morning.
Start learning with Thinkable