For two years now, we have been constantly hearing that AI will destroy jobs. Recruiters will disappear, junior staff will no longer have a place, and work as we know it will vanish before our eyes.
The problem is not that this discourse is anxiety-provoking, but that it is intellectually weak, based on very little, and above all, distracts attention from the real issues.
When we look at more serious studies, starting with those produced by Stanford, the picture is much less dramatic. Some jobs are partially exposed, there is a measurable effect on a very limited number of them, and the overall negative impact is in the tenths of a percent range. We are a long way from the predicted collapse, and also a long way from the disruption that some people like to predict ([FR]Stanford study: AI puts pressure on youth employment).
But this discrepancy between the facts and the review is no accident, quite the contrary.
In short:
- Alarmist talk about massive job losses due to AI is exaggerated and often based on confusion between tasks, occupations, and jobs in the macroeconomic sense.
- Serious studies (such as those by Stanford and the OECD) show that AI has a limited impact on employment, far from the doomsday scenarios often reported.
- In some businesses, layoffs presented as being linked to automation are in fact used to finance investments in AI, without immediate productivity gains.
- The main challenge lies in the transformation of career paths, as learning tasks are automated, which makes it more difficult for young people to enter the world of work.
- The productivity gains expected from AI are modest on a macroeconomic scale, and the real question is how they will be distributed: wages, working hours, or returns for investors.
Confusion between tasks, occupations, and jobs
Much of the discourse on AI is based on a constant confusion between three unrelated levels: tasks, occupations, and jobs in the macroeconomic sense. We begin by explaining, with figures to back it up, that some of the tasks involved in a job can be automated. So far, so good. Then we jump to the conclusion that the profession will disappear and, ultimately, that the job will vanish.
This reasoning is appealing because it is simple, but the problem is that it is also false. Occupations are not fixed lists of tasks, and economic history shows that they are constantly being reshaped. Photography destroyed certain occupations while creating others, often more numerous but different. AI is part of this continuity, not a radical exception.
This is exactly what the OECD’s recent publication on AI productivity (Miracle or Myth? Assessing the macroeconomic productivity gains from Artificial Intelligence) reminds us, between the lines.
The transition from micro to macro is not automatic, and forgetting this allows for anxiety-inducing headlines. Conversely, remembering this forces us to think a little.
Laying off employees because of AI or to finance AI?
First, we must be aware that a significant number of layoffs are the counterpart to the wave of “COVID hires”. A sudden increase in demand in certain sectors, particularly digital and consulting, led to massive recruitment, and now that we are back to a more “normal” level of activity, we are left with surplus employees.
There’s also another aspect of the debate that is rarely addressed clearly. In a number of businesses, layoffs are not the result of productivity gains linked to AI, but rather a condition for its financing.
AI is expensive: infrastructure, licenses, integrations, pilot projects, consulting firms, dedicated teams, etc. In many cases, it does not yet yield any measurable returns, not because it is useless, but because its integration into the work organization is slow, complex, and often poorly thought out (The great illusion of technological productivity gains (including AI)). Faced with this equation, some management teams make a choice. They reduce the payroll to free up investment margins, especially among AI solution publishers who invest heavily without checking in significant revenue, to such an extent that a business like Microsoft no longer publishes its AI activity figures in its accounts but buries them in the “Azure” line. Then they explain that these job cuts are the inevitable consequence of automation, but in the meantime, investors are reassured.
We are also beginning to see a new trend with businesses laying off staff on the grounds of AI, then rehiring them a few months later when the technological promise fails to materialize operationally (Employers regret AI layoffs and rush to rehire former talent, AI layoffs backfire: 55% of employers admit regret and The AI layoff trap: Why half will be quietly rehired).
AI is thus becoming less of a lever for transformation and more of a narrative alibi for short-term financial decisions.
The problem is not where we think it is
Saying that AI is not destroying jobs on a massive scale does not mean that everything is fine. There is a real unease, which is particularly visible among young people and in entry-level positions, but here again, the cause is often misidentified.
What is being undermined is not jobs as such, but learning mechanisms. The simple, repetitive tasks that allowed people to enter a profession, understand its codes, and progress step by step are precisely those that AI is taking over first. The risk is therefore not the disappearance of work, but the disruption of career paths (AI isn’t killing entry-level jobs. So, what is? and Automation could undermine the ability to renew skills).
This is an organizational issue rather than a technological one, and it directly engages management.
Who benefits from the gains?
When we take a step back, one constant emerges: the productivity gains expected from AI do exist, but they are modest on a macro scale. According to the OECD, they are in the order of 1% per year (see above) and, for the record, during the post-war economic boom, some countries consistently posted annual gains of 4 to 5% without destroying jobs. This is significant, but it is not a revolution comparable to the major industrial disruptions of the past.
And for Daron Acemoglu, who is, after all, a Nobel Prize winner in economics, “AI will lead to a ‘modest increase’ in GDP of between 1.1% and 1.6% over the next ten years, with an annual productivity gain of around 0.05%” (Daron Acemoglu: What do we know about the economics of AI?).
The central question then becomes one of redistribution. Will these gains go towards wages, working hours, and the financing of social systems, or will they be captured in the form of rents by those who finance and control the AI economy? In a context of geopolitical tensions, the end of multilateralism, and increased technological dependence, particularly in Europe, this choice is anything but neutral.
But talking only about jobs destroyed allows us to avoid this debate. It’s comfortable, but it’s also a way of not facing up to our collective choices.
Bottom Line
Ultimately, AI does not eliminate jobs, but rather reveals what we have accepted for years: poorly designed organizations, vague roles, jobs stripped of their substance, and, above all, financial decisions disguised as technological inevitabilities.
The question, then, is not whether AI will kill jobs, but what we do with this technology. How we integrate it, what we choose to finance, what we decide to preserve and, above all, what we are prepared to redistribute.
The rest—the prophecies, the bombastic announcements—are more about marketing than analysis (If you don’t buy my products and services, you’re all going to die.).
To answer your questions…
No, according to the studies cited, the overall impact of AI on employment remains very limited. The studies mainly refer to partial exposure of certain occupations and a measurable effect on a small fraction of tasks. On a macroeconomic scale, the negative impact is measured in tenths of a percent, far from the collapse scenarios often put forward. The alarmist discourse is based more on simplistic extrapolations than on solid data.
Because the debate often confuses tasks, occupations, and jobs. The fact that a task is automated does not mean that the occupation disappears. Occupations are constantly changing, as economic history has shown. Jumping directly from task automation to job collapse is intellectually appealing, but false.
In many cases, layoffs are primarily used to finance investments in AI. Infrastructure, licenses, and projects are expensive and do not yet generate measurable gains. AI is then used as a narrative justification for short-term financial decisions, sometimes followed by rehiring.
The problem is not the disappearance of jobs, but rather the disappearance of learning opportunities. Simple tasks that allowed junior employees to progress are often the first to be automated. This weakens career paths and poses an organizational and managerial problem rather than a technological one.
Yes. The productivity gains associated with AI are real but modest. The central question therefore becomes how they are distributed: wages, working hours, collective financing, or private pensions. Focusing on the fear of job losses mainly serves to avoid this essential debate.
Image credit: Image generated by artificial intelligence via ChatGPT (OpenAI)
Related articles:
How to reinvent work and professions in the age of AI
Will AI replace juniors? The false debate that’s only the tip of the iceberg
Like the climate, does AI deserve its “COP”?
AI and jobs: why I don’t believe in the “great replacement” of humans by machines
Work and AI: a transformation that is more organizational than technological
Who is handling your artificial intelligence projects? Probably not the right people.
