Insights
What is human labour for in the age of AI? Discover how work contributes to purpose, identity, innovation, and social progress in a rapidly changing world.What is human labour for in the age of AI? Discover how work contributes to purpose, identity, innovation, and social progress in a rapidly changing world.
The standard argument about automation and employment has two moves. The first is historical: technological displacement of labour has occurred many times before — from the handloom weavers displaced by the power loom to the bank clerks displaced by the ATM — and in each case, new forms of employment emerged to absorb the displaced workers, often in industries that did not exist before.
The second move is forward-looking: the same process will occur again, because the displacement of routine tasks frees human labour for higher-value, more creative, more distinctly human activities. Together, these two moves produce a conclusion that is reassuring, historically grounded, and, in the view of a growing number of economists, increasingly inadequate to the nature of the current technological transition.
The inadequacy is not about the historical record, which is largely accurate. It is about what makes the current wave of automation structurally different from its predecessors. Previous technological transitions displaced specific categories of physical or cognitive routine — tasks that were repetitive, rule-based, and could be codified into mechanical or computational procedure.
What is distinctive about current AI systems is that they are beginning to displace not specific routines but entire categories of judgement — the ability to assess complex situations, generate plausible responses to novel inputs, and produce outputs that are indistinguishable, in their surface characteristics, from those produced by skilled human practitioners. This is not a difference in degree. It is a difference in kind.
The economist David Autor, who has studied the labour market effects of automation more rigorously than almost anyone, distinguished between tasks that are routine — whether manual or cognitive — and those that require flexibility, judgement, and the ability to respond to non-standardised situations.
His research showed that automation had, through the late twentieth and early twenty-first centuries, hollowed out the middle of the labour market — eliminating routine jobs in both manufacturing and services while leaving relatively intact the highest-skill cognitive work and the lowest-skill personal service work that remained difficult to automate. The result was a polarisation of the labour market that contributed significantly to the income inequality documented by Piketty and others.
The current generation of AI systems threatens to extend this hollowing into territory that Autor’s framework treated as automation-resistant. Legal research, medical diagnosis, financial analysis, software development, and creative production — domains that the previous wave of automation left largely untouched — are now generating substantial evidence of AI capability.
The question is not whether AI will replace all human practitioners in these fields, which seems unlikely in the near term. It is whether it will reduce the number of practitioners required, reduce the premium that skilled practitioners can command, and shift the distribution of value toward the owners of AI systems and away from the workers whose labour those systems partially replace.
This raises a question that the standard argument about automation tends to avoid: what is human labour for? If the historical argument is that technology displaces labour but creates new labour, the implicit premise is that human beings need work — not merely for income, but for the structure, meaning, social connection, and sense of contribution that work provides.
The political theorist Hannah Arendt distinguished between labour — the cyclical, biological necessity of maintaining life — work — the production of a durable world of objects and institutions — and action — the distinctly human capacity for initiative, for beginning something new in the public world.
A society in which AI performs an increasing proportion of what Arendt would call work and action does not merely face an economic adjustment problem. It faces a question about what forms of human engagement remain available, and whether the social and psychological structures that currently depend on work can be sustained in its absence.
These are not questions that economists alone are equipped to answer. They are questions about what human beings are for — which is, in the deepest sense, a question that no technological transition has ever made obsolete, even when it has made every previous answer inadequate.
Main Theme
The current wave of AI-driven automation is structurally different from previous technological transitions because it displaces entire categories of judgement rather than specific routines — and this difference forces a question that economic analysis alone cannot answer: what is human labour for?
Central Idea
Autor’s research showed that previous automation hollowed out the middle of the labour market. Current AI now threatens domains previously considered automation-resistant. The deeper challenge is not economic adjustment — it is the question of what forms of human engagement remain available when work, as Arendt understood it, is increasingly performed by machines.
Implied Idea
The standard reassurance that technology always creates new jobs is not wrong about the past but may be inadequate to the present — because it assumes the key variable is the number of jobs rather than the quality and nature of human engagement that work provides. A society where humans are fully employed in low-autonomy AI-adjacent roles is not obviously better than one grappling honestly with structural unemployment.
Conclusion of the Passage
The question of what happens when AI performs an increasing proportion of human work is not only an economic question. It is a question about what human beings are for — which is, in the deepest sense, a question that no technological transition has ever made obsolete, even when it has made every previous answer inadequate.
Summary of the Passage
The passage argues that current AI automation is structurally different from historical transitions because it displaces judgement rather than routine. Drawing on Autor’s labour market research and Arendt’s distinction between labour, work, and action, it argues that the standard historical reassurance is insufficient — because the deeper question is not whether new jobs will emerge but what forms of distinctly human engagement remain available when AI performs an increasing proportion of work and action.
Difficult Words with Contextual Meanings
- Labour market polarisation: the hollowing out of middle-skill, middle-income jobs by automation, leaving growth concentrated at the highest and lowest ends of the wage distribution
- Automation-resistant: tasks or roles previously believed to require human judgement and therefore unlikely to be displaced by technology; the passage argues current AI is eroding this category
- Rent: here used in the economic sense of returns captured from owning assets rather than from productive contribution; AI ownership concentrates rent away from displaced workers
- Action: Arendt’s term for the distinctly human capacity for initiative and beginning something new in the public world, as distinct from biological labour or productive work
