Are AI going to steal our jobs? Is it stealing them already? Are we headed toward an “employment apocalypse”? The big problem, as with all great transformations, is the transition. The first Industrial Revolution took seventy years to permeate and benefit the entire British society —in fact, Marx and Engels wrote about that transition, and in the new transition sits my anticipation novel Be Water (2025)—. And this time? It is no longer the third, the digital revolution, but the new one, marked by artificial intelligence (AI), quantum computing, biotechnology or renewable energies with nuclear fusion on the horizon.
“There are jobs that we see have disappeared with automation, and even more with the new generative AI; this is evident in finance, the legal world, or among programmers”
The academic debate is in full swing, and at first glance it only resembles the one from a decade ago (which I treated in The Unstoppable March of Robots, 2016) when it focused almost exclusively on automation. There are jobs we see have disappeared with automation and, even more, with the new generative AI: in the financial sector, for example (customer service and analysts who learned their craft this way); in the legal world, with the work that interns and trainees do less and that is now performed by some programs; among programmers (AI programs program better and better); among interpreters and translators (knowing languages will carry fewer advantages); or among taxi or other vehicle drivers. And many others, which in general are middle-class jobs, an essential social stratum that has been hollowing out for years. We are living a situation of de-skilling, with consequences that we already see in politics.
Some large consultancies, such as PwC or Accenture, or even the New York Federal Reserve, now attribute to the impact of AI the decline in hiring, selective layoffs and a new reorganization of work. Although it acknowledges that there are still not enough data, a recent study from Yale University concludes that, currently, measures of exposure, automation and increases in AI and related technologies show no signs of being related to changes in employment or unemployment. Others think the real challenge is to equip people with the expertise to use these new technologies.
Amazon, which has been investing a lot in this field for years, has long classified its operations into six types of automation: movement, handling, sorting, storage, identification and packaging. It bets, it invests, to have first-rate capacity in each of them, according to one of its executives. Currently it is deploying a new generation of AI-powered robots, which could replace ten million jobs, and generate economic growth (how to share it?). According to Morgan Stanley, automation is about to trigger the greatest capital shift since the advent of the Internet. We are facing a new qualitative and quantitative leap in robotics.
“In Spain, Amazon has proposed a redundancy procedure that could affect up to 1,200 workers, in part due to the adoption of AI”
The multinational founded by Jeff Bezos plans either not to hire, or to lay off 600,000 employees in its distribution warehouses in the U.S. with intelligent robots. In Spain, Amazon has proposed a redundancy plan (ERE) that could affect up to 1,200 workers in its corporate offices, part of a global adjustment that will affect about 14,000 employees worldwide, partly due to AI adoption. About a thousand Amazon employees (engineers, product managers and others), though anonymously, have signed a manifesto warning that the purported “AI development at full speed and cost justification” approach by the company could cause “devastating damage to democracy, to our jobs and to the planet“.
Naturally, new jobs will be created, or at least new tasks or occupations, to be performed by humans. Some see a shift of work rather than a net destruction of jobs. Others are less optimistic. Destruction or creative destruction of innovation, as Schumpeter noted and Philippe Aghion has studied (the latest Nobel Prize in Economics)? Too often the person who loses their job to AI is not prepared for the new roles demanded, among other things, the know-how to use AI fully, which leads to unemployment or to settling for poorer-paid jobs (de-skilling). This is the great problem of the technological transition.
The new wave of automation and labor de-skilling
Executives of large companies tend to be more radical in their predictions than academic studies. Thus, for Jim Farley, CEO of Ford, “artificial intelligence will literally replace half of the white-collar workers in the United States.” Of course, the automotive industry adds to the impact of AI the fact that electric cars have half as many parts, and therefore need half the assemblers, who are highly paid white-collar workers. According to a survey of UK business leaders, 26% of large firms expect to reduce their workforce in the coming year due to the impact of AI, and junior positions will be the most affected.
“Reports published by the Chinese government point to a significant decline in demand for traditional labor”
Even the Chinese are in this line. They are the economy that robotizes the most. In 2024 (latest data), more than 295,000 new robots were installed in Chinese factories, four times more than in the European Union, seven times more than in Japan and almost nine times more than in the United States. The latest Frame 2.0 for AI Safety Governance published by China states that “the value of work as a production input is diminished, causing a significant drop in demand for traditional labor”. (To remember the high youth unemployment, despite the aging population).
How AI is changing the way we work
It’s not just about jobs — whose rates, compared to tasks, will lose meaning as a fundamental element of analysis and policy — but about the transformation of how we work. “The change is changing”, says a McKinsey report. From the European Commission, a Joint Research Centre team notes some paradoxical effects of the digitization of work. For instance: although digital technologies reduce the labor needed for routine tasks, they simultaneously increase the routinization and bureaucratization of professional functions that were not routine to begin with. It points to one of three drivers of change, more important than robots, which it calls the “platformization of work,” an expansion of data-driven systems for coordination, digital management and supervision and algorithmic management across various economic activities.
Then there is the geography of impact. According to a Brookings study, based on occupation-specific exposure data, metropolitan areas with high levels of education, high wages and white-collar jobs, which were previously considered at relatively low risk of automation, appear to be the places most exposed to generative AI. This wave of automation is very different from previous ones: “The possible patterns of impact of generative AI seem quite different from those of earlier forms of automation”, the report notes.
Perhaps a broader view is needed. According to Azeem Azhar and Chantal Smith, we are in the midst of restructuring the economy: from an economy that uses computation to one that is based on computation. What they call the “infinite-compute economy” (infinite-compute economy) has already begun. And it is pulling us along.
