A Nobel Prize winner in Economics does not believe that AI will cause massive layoffs: “Replacing workers is a losing proposition”

“Replacing workers with AI is a losing proposition”he stated in an interview published by MIT Technology Reviewwhere he especially questioned the discourse of some large technology companies about the future of work.

Even before receiving the Nobel Prize alongside Simon Johnson and James A. Robinson, Acemoglu had already gone against the prevailing enthusiasm in the industry. In 2024 he published the work “The Simple Macroeconomics of AI”, in which he calculated that the impact of artificial intelligence on US productivity would be “modesto” over the next decade.

According to his estimate, the improvement would not exceed 0.66% in ten years and could even be below 0.53% if it is considered that many future tasks will be more difficult to automate than the first applications of generative AI.

That vision clashed with the projections of technology CEOs who have been anticipating a radical transformation of office work for years. However, two years later, Acemoglu assures that data still does not support a “job apocalypse”.

In fact, it mentions that different studies so far they have not found significant effects on employment rates nor a general increase in layoffs associated with AI. Even a report from the United States Federal Reserve published in March 2026 concluded that there was “no evidence of a reduction” in job openings in industries with greater adoption of artificial intelligence.

The economist clarifies, in any case, that this does not eliminate the risk of specific impacts on certain professions or groups of workers most exposed to automation.

One of the main current focuses of debate are the so-called “AI agents”: systems capable of executing tasks with greater autonomy than a conventional chatbot. Several technology companies present them as tools capable of replacing entire workers.

Acemoglu does not share that idea. For him, a job is made up of multiple heterogeneous tasks that people coordinate naturally and flexibly, something that AI still cannot easily replicate.

As an example, he mentions the case of an X-ray technician. His job is not limited to taking medical images: it also involves collecting medical records, organizing files, managing databases, adapting procedures and coordinating with other professionals.

“A worker can naturally switch between formats, databases and work styles,” he explained. The doubt, as stated, is How many different tools, protocols or systems would an AI need to perform the same combination of tasks without errors?.

For the Nobel Prize winner, the real challenge of AI agents lies in the ability to “orchestrate” different activities with the fluidity that humans have. And he maintains that, as long as that does not happen, many positions will remain far from total automation.

Another point that he observes with concern is the growing landing of economists in artificial intelligence companies. OpenAI hired in 2024 Ronnie Chatterjee as chief economist and then announced joint work with Jason Furman to investigate the impact of AI on employment and productivity.

Also Anthropic brought together a group of economists to study the phenomenon, while Google DeepMind incorporated the economist Alex Imas as director of economics at AGI.

Acemoglu acknowledges that it makes sense for companies to want to understand the economic impact of their technologies, especially in a context of growing social skepticism regarding AI and employment. But he warns of another risk: that the most influential research on the future of work will remain in the hands of companies with direct interests in promoting favorable conclusions.

“What I hope doesn’t happen is that they are interested in economists just to push their views or feed the hype (exaggerated expectation),” he noted.

The economist also questions another central aspect of the current boom: the lack of truly simple and massive applications. To explain, he compares the current state of AI with programs like Word or PowerPoint, which could be installed and used quickly to solve specific tasks.

“Anyone could install those programs and get them to do what they wanted,” he explained. Instead, he maintains that he still no AI-based tools appeared with the same level of usability and practical adoption for the majority of workers.

Although millions of people already interact with chatbots, Acemoglu believes that this does not automatically mean a sustained leap in productivity or an immediate transformation of the labor market.

His look at the future, anyway, not completely optimistic. In another recent interview, the Nobel Prize winner warned that the economic model promoted around artificial intelligence can increase inequality and deteriorate the quality of employment if it is mainly oriented towards mass automation.

“This development is bad for equality, bad for the working class, bad for democracy and bad for social cohesion,” he stated.

According to him, The problem is not only technological, but also political and economic: how companies, governments and AI laboratories build the story around artificial intelligence and decide to apply it to human work.

Meanwhile, Acemoglu insists on a central idea: despite the climate of alarm surrounding AI, it still exists “an enormous uncertainty” about its true economic and labor impact.