An HEC Paris / Banque de France study published in 2026 documents for the first time the generational effect of AI on recruitment in France. Companies that adopt AI intensively hire between 17 and 25% fewer young people aged 21 to 29 on permanent contracts than the rest of their sector. Among software developers and computer engineers aged 21 to 29, permanent contract hiring was approximately 42% lower in March 2026 than in November 2022.
The essentials
- Among around 600 French AI-intensive companies, junior permanent contract recruitment is 17 to 25% lower than in the rest of the sector (HEC Paris / Banque de France, 2026).
- Across the field studied, permanent contract hiring of software developers and computer engineers aged 21 to 29 was approximately 42% below its November 2022 level in early 2026.
- Automation of entry-level tasks could reduce recruitment and training, but the study demonstrates neither this causal mechanism nor a shift toward senior recruitment.
- The breakdown of the generational pipeline raises a public policy question: who bears responsibility for training technical talent if companies disengage.
AI changes the recruitment calculation, not just work organization
For decades, major technology companies operated on a simple model: hire cheap juniors, train them over two to three years on execution tasks, then promote them as they gained expertise. This pipeline had clear economic logic. The junior cost little, learned fast, and became the next expert.
Generative AI changes that calculation. AI tools can assist certain tasks formerly entrusted to beginners, but the study does not establish any decline in their market value. The company would need fewer junior resources to handle a volume that a senior augmented by AI can absorb.
The study covers approximately 6,000 companies, of which nearly 600 declare intensive AI use—that is, those whose exposure to AI tools in business processes exceeds a significant threshold. The result is clear: these companies hire between 17 and 25% fewer 21- to 29-year-olds on permanent contracts than comparable companies with less AI exposure. The gap is not marginal. Across an annual recruitment cohort, it represents several tens of thousands of positions at the national level.
The figure for developers is even more striking. The decline of approximately 42% concerns permanent contract hiring of software developers and computer engineers aged 21 to 29 between November 2022 and early 2026. In late 2022, this is precisely when ChatGPT enters professional use at scale. The temporal coincidence does not prove causality, but it is hard to ignore.
Seniors required, juniors sidelined: the cold logic of HR departments
The rationality behind this shift is understandable, even if its consequences are uncomfortable. A senior developer augmented by AI produces more than a junior without AI support, and probably more than two juniors. The salary cost of a senior is admittedly higher, but the total cost—salary plus training plus supervision plus junior errors—often tips the balance.
Companies declaring intensive AI use hire relatively fewer 21- to 29-year-olds on permanent contracts; the study does not allow us to infer their HR strategy or increased senior recruitment. They import expertise rather than cultivate it internally. The problem is that this logic, rational at the level of a single company, becomes collective if it generalizes across an entire sector.
Actual AI adoption in European companies remains still limited, with fewer than 10% of SMEs truly integrating it into their processes. Among nearly 600 companies declaring intensive AI use, junior permanent contract recruitment trends less favorably; this association does not prove a causal effect of AI. The signal is early. It could announce a trend.
A technical labor market cannot function long on senior imports. Today’s seniors were yesterday’s juniors. If entry into these professions closes off, the pool of seniors will be depleted in ten years.
Generational discontinuity in technical professions
This mechanism has a name in labor literature: the breakdown of the generational pipeline. It describes the situation where entry routes into a profession narrow to the point of preventing the reproduction of skills within an economy. It was observed in certain manual trades after deindustrialization in the 1980s and 1990s. France still bears its marks in entire sectors of industry, where the absence of continuous apprenticeship created skill deserts that companies struggle to fill today.
The relative decline of intermediate professions on the wage scale signals a compression of career trajectories from the bottom in several sectors. In digital technology, the compression seems to operate differently: not through bottom-end wage pressure, but through closing the entry door. The result is the same: a generation struggling to find its first footing.
Data from the Dares on employment flows in the IT sector between 2019 and 2026 confirm the longer-term trend. After the 2020 decline, IT recruitment rebounded in 2021 and continued to progress in 2022; a subsequent slowdown must be dated and sourced separately. Companies declaring intensive AI use are associated with a more marked decline in young worker hiring, though the causal responsibility of AI cannot be established.
What makes the breakdown particularly concerning in technical professions is the nature of knowledge transmitted. Code, software architecture, managing complex systems: these skills are learned largely through practice and exposure to expert decisions. A junior who doesn’t find a job doesn’t just miss a salary. He misses the first two or three years of an implicit training that school cannot replace.
European institutional models: gains and blind spots
Germany and Switzerland built their technical labor markets on a dual model: apprenticeship, that is, training shared between school and company, with a contract, reduced salary and guaranteed integration. This model produced lasting results: youth unemployment rates structurally lower than France’s, and transmission of know-how that does not rest solely on universities.
France has made real efforts to develop apprenticeship. The number of apprentices more than doubled between 2018 and 2023, rising from approximately 400,000 to over 980,000, according to Labor Ministry data. In digital technology, work-study contracts progressed strongly, notably in specialized schools.
These programs were designed for a labor market where companies had an incentive to train juniors, who brought value by their second or third year. If AI durably reduces the productivity of novices, AI-intensive companies could disengage from apprenticeship itself, not just from junior permanent contracts. This scenario extends the logic already documented by the HEC / Banque de France study.
Training software developers in 2035 without corporate commitment
The public policy question is straightforward. If the most technologically advanced companies reduce their absorption of juniors, two scenarios emerge.
In the first scenario, the sector self-regulates. Companies less advanced in AI, or located in sectors where AI penetrates more slowly, continue recruiting juniors and ensuring their implicit training. These juniors acquire the basics, then migrate to AI-intensive companies as they gain expertise. The pipeline remains intact, but shifted and lengthened. This scenario assumes that automation pressure remains differentiated in the economy, which is plausible in the short term.
In the second scenario, the logic generalizes. As AI becomes affordable and accessible, an increasing number of companies applies the same rationality: fewer juniors, more seniors augmented. The market for certain IT recruitments has slowed, but consulted sources do not demonstrate a structural contraction in junior IT employment. Schools graduate diplomas that no one hires in their field. Competition intensifies for senior profiles, driving up their salaries and widening the gap with beginners.
The generational pipeline breaks, and France enters the 2030s with a technical talent deficit that overseas skill imports can only imperfectly offset.
Between these two scenarios, public action has real levers. The first is a junior quota, or a financial incentive to its equivalent: conditioning certain tax benefits related to AI or R&D on maintaining a minimum percentage of junior recruitment. This mechanism exists partially in other contexts, apprenticeship obligations in public procurement, for example, and could be adapted. The challenge is not to conceive of it as a punitive constraint, but as compensation for the negative externalities that selective recruitment creates for the entire system.
The second lever is targeted continuing education. If AI devalues technical noviceship, we must accelerate the retraining of existing profiles—mid-level developers in less exposed sectors, engineers in transition, digital professionals in transition—toward the skills that AI-intensive companies seek. Germany did this at scale for industrial trades during post-reunification transitions. Switzerland is adapting it today to digital transitions in its financial sector. France has the tools, the personal training account, France Compétences funding, but they were designed for individual transitions, not coordinated sectoral reconversion.
The third lever concerns revising the curriculum in engineering schools and IT training programs. If execution tasks are automated, the value of a junior developer rests increasingly not on coding speed but on the ability to architect, question specifications and work at the interface between professions and technology. Training in these skills from the bachelor’s level would allow juniors to enter the market with added value that AI does not yet cover.
The signal to watch for decision-makers
The HEC Paris / Banque de France study documents the state of the labor market in 2026. It does not predict the following decade. But it provides an early signal that policymakers, professional associations and training institutions have every reason to take seriously now, before generational breakdown is complete.
Three indicators would measure whether the phenomenon amplifies or stabilizes in the coming years. The first is the evolution of the junior hiring rate in tech sectors broadly, not just in the 600 already-identified AI-intensive companies, but across all companies whose AI exposure is growing. The second is the apprenticeship contract dropout rate in digital technology: if companies also disengage from work-study programs, the signal is more serious. The third is job search duration for young IT graduates: consulted France Travail data do not show an established lengthening for this entire population, though some technicians face greater difficulties.
Automation of tasks by increasingly autonomous systems affects entire swaths of the productive economy. In technical digital professions, the effect arrives sooner and more concentrated. France built in a few years a genuine capacity for professional apprenticeship. The real test is whether this system can adapt to a labor market where the value of the beginner changes faster than training cycles.
Sources
- HEC Paris / Banque de France, Private French Firms Hire Fewer Young Workers in AI-Exposed Jobs (2026)
- Dares, Employment Flows in the IT Sector 2019-2026 (Research, Studies and Statistics Directorate, Labor Ministry)
- Labor Ministry, Apprenticeship Statistics in France 2018-2023
