Carl Benedikt Frey is a kind and calm Swede who leads the Future of Work Program at the Martin School of the University of Oxford and is sought after for advice by the European Commission, the UN, and the US President's Advisory Group. He defines himself as a techno-optimist, although many of the things he says do not sound too optimistic.
Frey gained fame at a very young age. In 2013, before turning 30, this economist published an article with his colleague Michael Osborne that included a prediction that has been cited in over 6,000 academic papers and is a common phrase in speeches by any politician discussing innovation, whether at a business forum, a party conference, or the opening of a roundabout. In his forecast, Frey announced to the world that by the year 2033, automation will have taken over 47% of known jobs. He said this before ChatGPT arrived, before AI companies inflated the stock market, and before everyone searched the internet to see if their job is more at risk of extinction than the Iberian lynx.
During the week we celebrate May Day, Labor Day, it is convenient to study the many puzzling things happening in the workforce of the developed world. Just ask the young aspiring professionals, confident that technology was their greatest ally, with a degree in hand, ready to program, design, and serve clients. Overnight, they are discovering they are just the first link in the AI food chain in the office.
Official figures in the US show an increase in unemployment among recent graduates. Not only that: for the first time, a university graduate fresh out of college takes longer to find a job than a high school graduate. In Spain, the first collective dismissals directly related to the impact of AI are already being recorded, and according to the French digital services consulting firms Capgemini and Inetum, they particularly affect technical profiles, such as software developers, computer scientists, or cybersecurity analysts.
Age can be a key factor in being classified as a species threatened with ending up in the unemployment office of the future. But it is not the only one. According to the latest study by GovAI and the Brookings Institution, a Washington-based think tank, women will be the most affected by the changes brought about by AI. They represent around 86% of the workers most vulnerable to disruption. The reason is that they have much more difficulty accessing jobs with higher physical demands or those with a traditional male monopoly, such as plumbing, construction workers, or electricians, which are much more protected from AI. If this prediction is true, we would be facing a labor massacre focused on personnel in administrative and office positions with few options to redirect their careers. According to the study, web designers and secretaries are at higher risk, for example, than janitors.
The past has shown how difficult it is to predict changes in the job market decades in advance, given the unreliability of precedents, although current information suggests that the negative effects of automation will not be evenly distributed throughout society. Retrospectively, however, should give us some hope. Throughout the 20th century, women were also affected by technological changes and demonstrated extraordinary flexibility to reinvent themselves when they did not have access to higher education. For example, when hundreds of jobs in telephone switchboards held by telephone operators disappeared, new ones emerged, such as secretaries or waitresses. Perhaps something similar will happen again.
What cannot be denied is that the earthquake has begun, even though its intensity is unknown. Proof of this is the resurgence of a phenomenon that seemed extinct: the Luddites. At least that's what the survey by the Writer and Workplace Intelligence platform suggests, revealing that 44% of Generation Z workers in Europe and the US admit to sabotaging the implementation of AI in their company. Another Gallup survey shows the deteriorating reputation of this technology among young people: 30% feel "anger or rage" towards it, 11 points higher than a year ago.
This resistance sentiment is likely due to the extensive publicity from large companies attributing the increase in their profits and the decrease in their workforce to the advancement of AI. They report a boost in productivity and corporate efficiency, while simultaneously there are layoffs and rising unemployment. In other words, AI is creating an economic paradox.
This provokes a certain feeling of obsolescence in the spirit of the worker, which for the first time affects more qualified individuals with intellectual tasks. It is becoming common, as new apps and automation processes are announced, for office workers to feel the same as whalers in the whaling industry or horse breeders did when, at the beginning of the 20th century, they heard that someone named Edison wanted to bring electricity everywhere, dynamiting the whale oil trade, and that Henry Ford dreamed of people moving by car rather than by horse.
We asked Frey if the new revolution is moving too fast. "I admit that I expected many more changes considering how advanced the technology is," he acknowledges. "I still do not see a significant impact on productivity and employment. The gap between the enthusiasm AI generates and its real impact on the economy remains surprising."
-Well, the anxiety is real. In some countries, four out of ten workers believe their job will not exist in a decade.
-Yes. And I expected that anxiety. I anticipated it in my 2019 book The Technology Trap. We have seen it time and time again throughout history, and this is no exception. But I see a significant difference compared to the past: never before have the developers of a new technology been the ones announcing a "job apocalypse". This amplifies the public's fear to unprecedented levels.
Frey is not wrong. Over the past two years, industry giants have not stopped releasing announcements of wonders and also great calamities (for most, not for those proclaiming them, of course).
Dario Amodei, director of Anthropic, warned that AI could increase unemployment by 10% to 20% in just five years. Jim Farley, a senior executive at Ford, estimated that it will end half of white-collar workers, those professionals performing administrative and management tasks in an office. And Elon Musk -who is the baby at the christening, the groom at the wedding, and the deceased at the funeral in all futuristic matters- stated that work will soon be optional.
There are even questionable bets. Sam Altman, CEO of OpenAI, the company behind ChatGPT, admitted to having a chat with other industry executives where they bet on who will correctly predict the inevitable date when, with AI support, we will see a multi-billion dollar company with only one employee.
It is true that the newspaper library shows that, often, magnates tend to exaggerate in their public interventions or spread information to benefit their corporate interests, but in 2026 something previously unimaginable has happened: Silicon Valley has fallen silent. No more predictions and announcements of the benefits of AI on their balance sheets. This is further proof that selling AI is not as cool as it used to be. It no longer seems great for someone with a mortgage and children to lose a highly paid position that not long ago was considered as stable and resistant as steel. The work, as reported by Annie Lowrey in The Atlantic, has made AI unpopular.
"This change affects young people a lot," says economist José Carlos Díez. "We are talking about software engineers, economists, lawyers, and, among others, starting journalists. The key is for companies to take advantage of the substitution of routine tasks that AI cheapens to improve their productivity and competitiveness by using those saved work hours to increase their sales and margins."
Today, the concern reaches those aspiring to good jobs, not those who barely survive with the lowest paid jobs, who were the first to be affected by previous technological revolutions. Since AI went viral, its impact is no longer only discussed by digital prophets and employment scholars like Frey. Economists take it very seriously because, for the first time, they have figures to rigorously assess its effects.
Currently, the intellectual goal of experts studying it is clear: to find out as soon as possible whether the economy will be able to absorb the changes in the labor market or if, on the contrary, AI will progress so rapidly that it will cause high unemployment and social tensions.
One of the most interesting studies investigating how AI could lead to job displacement is of Spanish origin. Luis Garicano, a professor at the London School of Economics, along with other academics, suggests that labor markets value positions as a whole, not tasks. Garicano uses the example of radiologists, a healthcare profession that was considered a prime target of AI advances years ago but is highly valued in the current market. Why does this happen? As explained by the former Ciudadanos politician on his X account, the radiologist not only classifies images, a task that a machine could do, but also performs other highly valuable tasks: prioritizing cases, communicating with other doctors, training residents, signing diagnoses, and making complex medical decisions. In other words, the market buys the whole package of their services. For the economist, the question posed by AI is not whether it can perform a task within the package, but whether the task can be extracted.
In a normal situation, the labor market tends to make natural adjustments regarding professions. The retirement or dismissal of 4% of workers in a certain occupation over several decades can be absorbed without market distress, as history has shown us with countless professions that have naturally become extinct. This is evolution.
Anyone over 40 remembers that as a child, it was common to see knife sharpeners advertising on the street or finding a video rental store on the corner. Currently, these businesses that provided for many families are archaeological relics of the labor market, just as serenos, water carriers, or town criers were for our parents.
The question is whether this process will accelerate abruptly. That's why it's time to identify the most threatened species by AI.
Mar Manrique is 27 years old and represents the generation and gender that economists study by analyzing their behavior towards machines. She is a woman, young with a university education, and has entered the workforce in recent years. There is no doubt that she knows what it's like to deal with the difficulties of any newcomer in the job jungle. That's why she wrote Un trabajo soñado (Ed. Península), an interesting essay on the disillusionment of the generation that wanted to make a living from the internet.
"There is a false mirage that we, being digital natives, will adapt better to this great transformation, but this has nothing to do with the arrival of social networks, it is something much bigger," says Manrique, creator of the journalism newsletter Fleet Street. "AI is already another variable to add to our job insecurity, an addition to the mix that prevents us from building a stable career that allows us to climb socially and access housing."
This viewpoint is supported by Carl Benedikt Frey, who sees how recent graduates, in addition to facing the work capacity of AI, have to deal with rising interest rates and the economic uncertainty of wars and tariffs.
"The convergence of all these factors is what has been most difficult to predict," says the Oxford academic. "Any of these forces is manageable when dealt with separately, but together they create a truly hostile hiring environment, especially when we talk about newcomers who do not have a track record to rely on."
The alarm signal detected for Mar Manrique's peers jumped last November through the work carried out by three Stanford academics, known by a name that sounds like a detective novel written by ChatGPT: The Report of the Coal Mine Canaries. In it, millions of payroll records of workers exposed to generative AI were analyzed. The conclusion was devastating: the canaries, a metaphorical reference to the birds that detect the presence of toxic gases before the miners, are workers between 22 and 25 years old, who have experienced a 13% drop in hirings since 2022.
This study was debated by some economists who believed that its conclusions were inflated by the ChatGPT effect, a tool that emerged in November 2022 causing an extraordinary frenzy in society. However, subsequent employment data analysis and the layoffs of these profiles announced in major tech companies, law firms, and consulting firms in the US are starting to support the canaries' theory.
This raises a fundamental question: How can someone who is not allowed to start stop being a beginner? It is so difficult to answer that, in the spirit of this report, we asked ChatGPT directly:
-It can't. And that is precisely the system's flaw. Do you want a longer answer for your article?
-Yes.
-The system has always needed "initial inefficiency." Historically, companies assume that a junior is not productive at the beginning, that making mistakes is part of the learning process, and that the investment in training pays off later.
This breaks the mold. Will computer programmers become the old video store clerks? Or have novice consultants become night watchmen? Does the human translator have the same future as a bell ringer?
In this environment of widespread confusion, it is worth asking what young people who have not yet entered the job market should do in the face of what is coming.
According to Carlos Victoria, an Economics professor at the Universidad Pontificia de Comillas, a teenager should consider three career paths to face the future: "The first is to pursue jobs that require skills that cannot be replaced by AI, such as caregiving; then there are jobs whose activities can be enhanced by technology itself, and finally, those that require transversal skills, including teamwork, public speaking, critical thinking, and the ability to discern information."
The problem seen by experts lies with the authorities, who seem oblivious to the challenge. The most industrialized countries do not seem willing to seriously develop contingency plans. In Spain, the Office of Prospective and Strategy, under the Presidency of the Government, drafted an extensive report called Hispania 2040 on the long-term impact of AI.
"Honestly, I think there is no strategic plan here," says a pessimistic José Carlos Díez, who believes that we are already late to adapt to the challenge. "The only plan I see is to regulate the ethics of AI and limit the use of data so that our companies cannot model AI and become less competitive."
If measures are not taken to soften the impact of AI, office workers with very routine and easily automatable tasks could quickly become the heirs of the workers who underwent industrial restructuring in Spain in the 1980s. Repeating a similar scenario to what was experienced then in Asturias, León, Galicia, the Basque Country, the Bay of Cádiz, or Sagunto would be very serious. Not only due to social conflict and public spending required to mitigate the unemployment hemorrhage but also due to the costs detected decades later. The main one being generational damage. The children of those workers who fell victim to competitiveness had access to very good education; however, the lack of an alternative business fabric to factories, shipyards, and mines in their localities forced them to emigrate in search of work.
In this regard, Mark Muro, a senior researcher at Brookings who has assessed the political relevance of the study conducted by GovAI on the labor vulnerability that could focus on women, believes that the great danger lies in these jobs being "out of sight and out of mind of policymakers and the population."
For Frey, the solution is a matter of will. "We created the welfare state in response to the Great Depression," he explains. "If AI has a significant impact on employment, what will be needed are not so much new ideas, but the political capacity to implement existing ones. We have the necessary political tools; the question is whether democracies can overcome the current polarization to put them into practice."
-ChatGPT, when will AI cause my job loss?
-I'll answer you directly because this is a very real concern: there is no specific date, but there is a process that is already beginning in journalism and that can affect you depending on your role.