Automation Doesn't Work for Everyone

AI could free us from tedious work. Under capitalism, it does the opposite.

Large corporations have spent months laying off workers, in what looks like an automation strategy driven by the incorporation of artificial intelligence into the production process. Some estimates point to more than one hundred thousand layoffs worldwide so far this year, a figure similar to the one recorded across the whole of 2025. In Spain, the Amazon collective redundancy plan — 1,200 people — and Capgemini’s — 748 people — are two clear examples, and there will likely be more in the coming months.

Since the dawn of capitalism, the fear of being replaced by machines has been one of the deepest and most persistent anxieties of the working class. Workers had been dispossessed of control over the means of production through violence and theft — as paradigmatically illustrated by the enclosures, those fencings-off of common lands that drove the peasantry into urban factories — so they always sensed that the reorganisation of labour around new machinery would not redound to their benefit. The Luddite movement, in the early nineteenth century, is perhaps the best-known expression of that popular resistance, precisely because it was a rebellion that sought to destroy the machines that were ruining traditional trades and casting thousands of workers into unemployment and destitution. It is sometimes forgotten that the State at the time responded with still greater violence, with laws that punished by hanging anyone who broke a loom, and with a clear defence of the machine owners’ private property over the workers’ livelihoods. The State was then, more than ever, an instrument fully committed to the proprietors of the means of production.

Those means of production were machines designed to cut costs. Although mainstream historiography has sweetened the inventive capacity of the mechanics who produced the first power looms or the steam engine during the Industrial Revolution, the truth is that the purpose pursued was, above all, to replace expensive workers — skilled adult men — with women and children, who were worse paid and less able to organise collectively. Moreover, and as Andreas Malm has documented in detail in Fossil Capital, those machines were at first very inefficient from a thermodynamic standpoint — the water wheel was often cheaper than the coal-fired engine — but they allowed far greater control over the workforce. Workers went from producing with a certain autonomy in their homes in the rural world, setting their own pace, to doing so, piled together and supervised, in the factories of the polluted and overpopulated periphery of industrial cities. From the outset, in short, capitalism’s technological innovations have been geared towards maximising profits at whatever social and ecological cost was necessary, not towards lightening human effort or democratising wealth. And this is the result of capitalism’s deep logic, not of technology in itself, which also explains why, more than two centuries on, the process feels so familiar to us.

Against this critical reading of machines, economists tend to insist that, although there is a painful adjustment process, after the introduction of each new labour-replacing technology new jobs are always created in other sectors — usually, it is said, better paid. Carl Benedikt Frey, the author of The Technology Trap, opens his book by recalling the replacement of lamplighters, who used to light oil lamps, by electric lighting. Marx, for his part, saw in this kind of change horrific consequences for the workers who lost their means of subsistence, but also evidence of the progressive character of the capitalist system as it moved towards more modern relations of production. David Ricardo thought the same at the beginning of the nineteenth century, and attributed any unemployment caused by machinery to a transient friction. However, in the third edition of his Principles — a famous change of position too often forgotten by contemporary thinkers — he introduced a modification that made his argument far more sceptical: he accepted that substitution could be permanent and harmful to workers. That the intellectual father of much of liberal economic theory ended up acknowledging this should invite us to take a much more cautious view of the promises of technological optimism.

It is true that, in general, mechanisation almost never entails the automation of entire jobs, but rather of specific tasks within each post. That usually forces workers to specialise in activities that the machine cannot yet perform, often requiring higher qualifications: supervising, managing, caring. This process also lies behind the failure of what Marx expected to be a total polarisation of society between the proletariat and the bourgeoisie. As it turned out, alongside the classic industrial worker, throughout the nineteenth century there grew a sizeable middle class of salaried employees specialised in activities involving more intellectual management than manual labour: technicians, administrators, professionals, supervisors. The so-called “problem of the middle class” — already debated by Bernstein and Kautsky at the end of the nineteenth century, and revisited in detail by Erik Olin Wright — is not new, but it has taken new forms with each technological wave. What is distinctive about the current one, as the Economic and Social Council’s report on the future of work emphasises, is that automation hollows out intermediate jobs in particular and polarises the labour market between highly qualified posts and highly precarious ones. In other words, it worsens an inequality dynamic that is already quite pronounced for other reasons.

Despite all this complexity, in our societies we absolutely idolise technological progress to the point of refusing to problematise its role. Any criticism of the specific direction of innovation is quickly neutralised by labelling it “technophobic” or “Luddite”, as if the choice were between uncritically admiring every novelty and going back to live in caves. But anthropologists like Alf Hornborg have been warning for decades that the machine is a condensation of materials and energy flows extracted by the centre from the peripheries of the world economy. As is well known, behind every mobile phone there is Congolese coltan extracted under conditions of semi-slavery, lithium torn from the ecosystem of the Andean salt flats, Chilean copper, and silicon refined with enormous amounts of electricity. In short, machines do not fall from the sky and are built from materials and energy flows that come from somewhere specific, even if we don’t usually think about it. That turns every technological leap into an operation of unequal ecological exchange, in which the apparent well-being of a few rests on the depletion and contamination of many others, almost always far from our sight.

Perhaps the paradigmatic example of all this is so-called artificial intelligence. Although it is sometimes presented as an almost dematerialised process — a magical, ethereal “cloud” — the truth is that it entails an extraordinary consumption of water, minerals and energy: the large data centres that train generative models require huge amounts of resources, and the dispute over electricity supply has become one of the central concerns of the tech sector itself. Even so, and bearing in mind the need to manage these costs, the technology could be used for socially desirable ends: contributing to scientific research, helping to diagnose illnesses, translating minority languages, easing tedious or dangerous tasks, freeing up working time for care and rest. In practice, however, its use is widespread in the industrial production of digital nonsense — empty images, repetitive texts, advertising content designed to capture attention — or, worse still, in the hyper-exploitation of workers: algorithms that dictate every metre of a platform courier’s route, systems that time work down to the second, or apps that assign impossible shifts to people already earning subsistence wages. It is not so much a paradox as the incorporation of technology into capitalist logic: artificial intelligence could raise productivity in countless areas — eliminating a whole list of tasks that are tedious for the worker — and produce the same output in far less time, but that option is foreclosed under capitalism.

And it is foreclosed because the system’s blind objective is to produce more and more and more, regardless of the social and ecological costs and regardless of the availability of other social alternatives. Every increase in productivity translates into greater profits for owners and either intensified tasks or outright redundancies for those who carry out the work. Digital platforms — that latest frontier of wage labour — are a good portrait of the phenomenon: cutting-edge technology placed at the service of extreme precariousness, greater labour informality, permanent algorithmic surveillance, and an attempt to reimpose the old relations of servitude with a user-friendly interface. The more serious analysts acknowledge that the future of work will depend less on the technical power of machines than on the democratic governance — in labour, fiscal, and industrial terms — that we collectively decide to build. It seems obvious that without social dialogue, collective bargaining, and markedly progressive public policies, technological change will be just one more mechanism for redistributing wealth upwards.

And that, ultimately, is the real absurdity of contemporary automation. It is not that it replaces workers — in itself, having machines and algorithms take on some of the heaviest tasks could be excellent news — but that it is not directed towards any democratically planned social end. It responds solely to the insensitive stimuli of the market, to quarterly stock-market pressure, and to the individual calculation of profit. In that way we have become part of a mechanism that never stops, precisely because there is no common plan to justify the effort. We are at the mercy of a great unconscious force. To debate the automation of work is therefore not to debate whether machines are good or bad: it is to debate who decides what is produced, how it is produced, and what it is produced for. And that is also the question that the Luddite of two centuries ago was hurling with his hammer, and which the contemporary left should not give up on posing with every tool — technical, trade-union and political — within its reach: technology, like the economy, what for?