FEATURE: Artificial Intelligence Is Eroding Jobs; France Is Feeling the Impact

Maxime Marquette

2026-07-22 15:27:10

A Slowdown in Youth Employment

The Treasury report highlights a particularly concerning trend: young people’s entry into the workforce appears to be slowing since the rise of generative AI. This finding aligns with observations from INSEE, which documented a decline in employment among 22- to 25-year-olds in the occupations most vulnerable to automation—approximately 16 percent—according to a report published in March 2026.

In practical terms, this means that recent graduates seeking their first job in IT or programming are facing a job market that is hiring less than before—not because economic activity is slowing down, but because entry-level tasks are increasingly being automated.

The IT Sector: A Real-World Testing Ground

Between the fourth quarter of 2023 and the fourth quarter of 2025, employment in the information and communication sector declined by 3.0% in France, according to data from INSEE. This decline comes even as the sector’s value added continues to grow—a paradox that clearly illustrates the dynamics at play: more is being produced with fewer workers.

This pattern is not unique to France: the United States is experiencing a comparable trend, where digital investment is driving growth without creating jobs at the same rate, according to the same INSEE report.

Seeing an entire generation struggle to find their first job because of technology they didn’t choose strikes me as the kind of silent injustice that we’ll still be talking about ten years from now, with the hindsight and regrets that always accompany poorly anticipated transitions.

Between 5% and 60% of jobs at risk: an admission of widespread uncertainty

A range that speaks volumes about collective ignorance

The Treasury cites available estimates that place the share of jobs potentially at risk in the short or medium term from generative artificial intelligence between 5% and 60%. Such a wide range is not a technical detail: it is an admission that no one today truly understands the full extent of the upheaval underway.

Specifically for France, a study by Arquié et al. (2026), cited by the Treasury, suggests that 3.8% of job tasks are currently at risk of automation by generative AI—a figure that could rise to 16.3% within two to five years.

Five Million Jobs in the Crosshairs

This same study, conducted jointly by Coface and the Observatoire des Emplois Menacés, estimated that approximately five million jobs out of a workforce of thirty million people could be affected by this growing exposure to generative AI. Notably, unlike previous waves of automation, it is high-wage earners who appear to be most at risk, with an exposure rate of 22.1% for the top 10% of earners.

This reversal upends conventional wisdom: for decades, skilled workers were promised that their degrees would protect them from automation. This is clearly no longer the case, and this reality deserves to be stated plainly to young people who are investing in lengthy educations today.

There is something rather ironic—almost cruel—about seeing the highest-paying and most skilled jobs suddenly find themselves in the crosshairs of a technology that, just five years ago, was presented as a threat limited to repetitive, low-skilled tasks.

PwC's Global Barometer Confirms a Two-Tier Labor Market

Soaring Productivity for a Minority of Companies

PwC’s Global AI Jobs Barometer 2026, published in June, analyzes more than one billion job postings across six continents to reach a conclusion similar to that of the French Treasury: AI is creating a two-tiered labor market, where judgment and leadership are becoming even more highly valued skills, according to PwC.

According to the same barometer, the 20% of companies most exposed to AI recorded a 163% increase in labor productivity compared to a 2018 baseline—a considerable gap compared to the rest of the global economy.

Job creation that masks targeted job losses

Other data, particularly that compiled by specialized platforms such as Snowflake, Anthropic, and LinkedIn, show that 77% of organizations worldwide report net job creation linked to AI, often accompanied by targeted job losses in specific technical and administrative roles. France is following this global trend, with a gradual redeployment of talent rather than a sudden, drastic net loss of jobs.

This nuance matters, but it should not be used to downplay the very real pain of those whose jobs are actually disappearing—even if another position, elsewhere, is being created at the same time in a department or company they may never join.

To say that AI creates more jobs overall than it destroys is statistically true, but it offers absolutely no consolation to the person who is actually losing their job this week, and I refuse to treat this distinction as a negligible detail.

France Faces a Choice: Invest or Fall Behind

International Competitiveness as a Key Argument

The Treasury is clear on one point: in the context of international competition, investing in artificial intelligence is becoming essential to sustaining competitiveness—and thus employment—which will be increasingly tied to the adoption of this technology. In other words, France cannot afford to back down simply because the transition is daunting.

This approach is part of the national “France 2030” strategy, which sets ambitious goals: 16% growth over ten years driven by AI and a 20% increase in productivity, according to data reported in several recent economic analyses.

Massive Hiring Needs Despite Everything

Paradoxically, a survey on the workforce needs of French companies published on April 21, 2026, identifies 2.2 million planned hires this year, including 84,000 in the technology sector alone—where 49% of companies are nevertheless struggling to fill their positions. The healthcare sector, meanwhile, anticipates 322,000 new hires and has been largely spared from the direct threat of automation.

This contrast between labor shortages in certain occupations and fears of replacement in others illustrates the true complexity of this transition—a far cry from simplistic narratives of robots simply replacing humans.

AI is often presented to us as a uniform threat, but the reality of the French labor market more closely resembles an archipelago of contradictory situations, with glaring shortages on one side and fears of job losses on the other—sometimes even within the same building.

The European Central Bank is also sounding the alarm

Five Million Jobs Projected by 2030

The European Central Bank acknowledges the positive effects of artificial intelligence on the French economy while warning of short-term negative impacts: up to five million jobs could be lost in France by 2030 as a result of AI-driven automation, a figure that aligns with projections from Coface and the Observatory of Threatened Jobs.

The ECB also highlights a concerning structural risk: job losses could temporarily outpace job creation, with a ripple effect on tax revenues, since the jobs most at risk are often the most skilled and highest-paying.

Unemployment Is Already Rising, Though Not Systematically Linked to AI

Unemployment in France rose by 1.7% year-over-year in 2026, though this increase cannot be directly and entirely attributed to artificial intelligence, according to data reported by France Travail. This nuance is important: it serves as a reminder that AI is not the sole variable in a much broader economic equation, which is also shaped by international competition and multiple sectoral shifts.

But this nuance must not serve as an excuse for inaction: even a partial contribution by AI to rising unemployment warrants a structured public response, before the phenomenon worsens further.

I am as wary of arguments that blame everything on AI as I am of those that exonerate it entirely: the truth likely lies somewhere in between, and it is precisely this uncomfortable gray area that policymakers must have the courage to address publicly.

Education: The Only Credible Response to Uncertainty

What Economic History Teaches Us About Technological Revolutions

The Treasury points to a lesson from previous technological revolutions: in the past, the jobs created by new technology have ultimately offset those that disappeared, while the very nature of professions underwent profound changes. There is no guarantee that this pattern will repeat itself exactly with generative artificial intelligence, but history at least offers a cautious frame of reference.

This time, however, the pace of change appears to be significantly faster than that of previous industrial revolutions, leaving workers and institutions less time to adapt gradually—a factor that many economists consider crucial for the future.

Public Support Deemed Essential by Experts

The Treasury’s report is explicit: support through public policy—particularly in the area of training—is deemed necessary to assist in the retraining of workers whose jobs may be at risk and to promote the adoption of AI across the entire French economy. This is not just wishful thinking; it is a condition laid out in black and white by the government’s own economists.

It remains to be seen whether this support will actually materialize on the ground—in vocational training budgets and retraining programs—or whether it will remain nothing more than a good intention recorded in yet another report, quickly forgotten after its publication.

I’ve seen too many government reports make common-sense recommendations that end up shelved due to a lack of funding: I sincerely hope I’m wrong this time, because what’s at stake directly affects the lives of millions of French workers.

Manual and healthcare jobs: a relative refuge, but not a permanent one

Sectors That Have So Far Been Spared by the Wave

Some sectors are weathering the surge in automation driven by generative AI better than others. Healthcare—with 322,000 new jobs projected for 2026—along with the hospitality and construction industries, remain largely unaffected, according to data from several recent studies on the French labor market. Manual trades, which require a physical presence and constant adaptation to unpredictable situations, remain, for now, beyond the reach of large language models.

This relative protection should not, however, be confused with permanent immunity: rapid advances in robotics combined with AI could, in the medium term, extend automation to tasks currently considered out of reach.

Human Expertise as the Last Line of Defense

Experts consulted in several studies agree on one point: jobs requiring judgment, empathy, and the ability to adapt to unpredictable situations remain, for now, the best protected against automation, regardless of the pay associated with these positions. This is relatively good news for nurses, educators, and skilled tradespeople.

But this relative protection alone is not enough to solve the overall employment equation in France, since not all workers threatened by automation will be able to simply retrain overnight to move into these protected sectors.

Thinking that manual and caregiving jobs are spared for now may seem reassuring on the surface, but it doesn’t solve anything for the 40-year-old engineer or accountant who, in theory, would have to retrain for a job they’ve never done and for which they have no training.

Conclusion: A transition that will not tolerate a wait-and-see approach

A report that raises more questions than it provides answers

The Treasury’s report on artificial intelligence and employment offers no definitive predictions, and that is precisely what makes it credible. With between 5% and 60% of jobs at risk, and job creation and job losses that may or may not offset each other depending on the scenarios considered, the only certainty that emerges is that a profound transformation of the French labor market is already underway.

What young graduates struggling to land their first job—or experienced employees seeing their positions redefined or eliminated—are experiencing today will not wait for statisticians to definitively settle the debate between the various economic studies.

Political Time Versus Technological Time

A worrying disconnect is emerging between the pace of government reports and the pace of technological adoption in businesses. France still has a window of opportunity to invest heavily in training and support, but this window is closing as generative artificial intelligence continues to advance in the capabilities it offers to businesses.

I’ll conclude this piece with a simple conviction: waiting for perfect scientific consensus before taking action would be the worst possible decision, because in the meantime, real people are already losing real jobs—today, not in some hypothetical future.

Signed, Maxime Marquette, columnist