A new report in France is delivering a blunt warning: the next wave of artificial intelligence, so-called “agentic” AI that can make decisions on its own, could threaten roughly 5 million jobs.
The twist is who’s most exposed. It’s not restaurant servers or construction crews. The study says many of the biggest targets are well-paid, white-collar roles clustered in major cities, exactly the kinds of jobs that have long been considered safer from automation.
What “agentic AI” means, and why it’s different
The analysis, published in March 2026 by credit insurer Coface, estimates that agentic AI could put 16.3% of jobs in France’s public and private sectors at risk, about 5 million positions, according to reporting byLe Monde. The research was led by Axelle Arquié, founder of an observatory tracking threatened and emerging jobs, along with Coface economist Aurélien Duthoit.
Unlike earlier automation that handled narrow, repetitive tasks, agentic AI is designed to act more like a digital worker: it can plan, execute steps, learn from feedback, and operate with minimal supervision. That’s what makes it a direct competitor for many office jobs built around analysis, writing, coordination, and decision-making.
White-collar jobs take the hit, manual work looks safer
The report’s headline finding is the profile of the “likely losers.” White-collar roles, often higher-paid and concentrated in big metro areas, rank among the most vulnerable. The study argues that many of these jobs are packed with cognitive tasks that AI systems can increasingly replicate at speed and scale.
Industry isn’t immune either. In wealthy countries, Duthoit notes, manufacturing includes a lot of brainwork, engineering roles and research-and-development-heavy work, not just assembly lines. Germany, with a larger industrial base than France, actually fares slightly worse on this measure.
By contrast, hands-on trades, especially construction, along with hospitality and food service appear relatively protected for now, largely because they require physical presence, dexterity, and real-world problem-solving that AI still can’t do without expensive robotics.
Young workers are already feeling it, research suggests
The pressure may already be showing up in early-career hiring. A Stanford research team led by economist Erik Brynjolfsson found that employment rates fell by more than 10% since 2022 for 22- to 25-year-olds working in AI-exposed roles, including software development and customer service, according to French business outletChallenges.
In software development alone, the decline was nearly 20%, the report said, an ominous signal for new grads aiming for the kind of entry-level tech jobs that have served as a reliable on-ramp to the middle class.
Meanwhile, Gartner projects that by the end of 2026, 40% of enterprise applications will include AI agents, up from 5% in mid-2025, suggesting the technology is moving from pilot projects into the core of how companies run.
Wages split into two economies: automation pressure vs. talent premiums
The labor market picture isn’t uniformly bleak. While automation threatens some roles, workers who can build, deploy, secure, or audit AI systems are becoming more valuable, and more expensive.
Consulting firm WTW expects median pay increases of 3.1% this year across all sectors in France, roughly flat compared with 2025. But pay is projected to run hotter in AI, cybersecurity, and the broader tech-media-telecom sector, at about 3.2%, on par with energy, chemicals, and pharmaceuticals, according toLa Tribune.
WTW’s Khalil Ait Mouloud said the biggest raises should go to “in-demand” roles where employers can’t find enough qualified candidates, especially in AI and cybersecurity, where demand is rising faster than the talent pipeline.
A reshuffling, not a wipeout, at least at first
The report paints a two-speed transition. On one side are cognitive jobs that once seemed protected by education and credentials, now facing competition from AI agents that can work nonstop for a fraction of the cost. On the other are scarce specialists who can design and govern these systems, whose bargaining power is growing.
And the shift won’t erase existing needs overnight. France’s Apec, an organization that tracks white-collar employment, counted 17,800 job postings for accountants in 2025, evidence that even as AI spreads, companies still need people in core business functions. The change may be less about jobs disappearing instantly and more about which tasks, and which workers, get valued in the new economy.
