Machines wiped out most farm jobs, but factory and office jobs kept growing — why didn't automation just shrink employment overall?

This explores why farm employment collapsed when machines arrived while manufacturing and services kept hiring, and what that pattern tells us about automation today. The library doesn't contain a direct history of farm mechanization. It holds no notes on crop yields, rural migration or how food spending changed as people got richer. Treat what follows as the library's lens on the question rather than a full answer.

The closest material is Autor's argument that automating one link in a production chain raises the value of the links still done by people Does automating some tasks make other human tasks more valuable?. His well-known example is bank tellers: their numbers rose while ATMs spread, because cheaper branches meant more branches, and tellers moved toward relationship work. So why didn't farming follow the same path? The comparison suggests an answer: a technology creates jobs only when cheaper output leads people to want a lot more of it, or when the remaining human tasks become more valuable. People stop eating more food long before they stop wanting more banking, goods or services. When tractors made each farm worker far more productive, demand for food couldn't grow fast enough to absorb the extra output, so the work shrank. Spending moved to sectors where demand kept growing. This is a reading built on Autor's logic, not a finding stated in the library.

Read across several notes, the deeper point is that automation rarely just deletes jobs. It reshuffles which tasks people do. Acemoglu's model frames AI's effects as tasks moving between machines and workers rather than tasks disappearing Will AI increase inequality less than past automation?. Field studies of companies show the same thing at small scale. A shipbroking firm didn't eliminate manual classification work. It turned that work into checking and correcting the algorithm Does automation eliminate manual work or repurpose it?. MIT found that generative AI adoption moved workers from doing tasks to supervising them, without layoffs, a pattern it says aviation and manufacturing went through earlier Does generative AI adoption actually eliminate jobs or reshape them?.

The surprising thread is that the next 'growing sector' may not be a sector at all. It may be a scarce kind of human work. One note argues that as AI makes doing tasks cheap, human capacity to verify the results becomes the bottleneck on growth What becomes scarce when AI execution grows cheap?. A warning comes with that: if machines take over entry-level work, novices lose the chance to learn from experts. Output can rise while long-run growth slows Can automation raise output while slowing growth?. Early payroll data hints at this. AI shows no economy-wide job losses yet, but hiring of young workers in AI-exposed jobs lags sharply Is generative AI displacing workers at economy-wide scale?.

There is also a reason the farm-to-factory story may not repeat cleanly. Farm workers had somewhere to go. Anthropic's scenario model and Brynjolfsson's work both raise the risk that AI's gains go to whoever owns the technology rather than to new kinds of employment Does AI growth inevitably shift wealth away from workers? Does automating labor concentrate power more than augmenting it?. The open question the library leaves you with: when AI automates services, what plays the role that factories and offices played for people leaving the farm?