“AI is replacing everyone.” This slogan works well on LinkedIn, holds up for three hours on a TV talk show, then collapses as soon as you look at the market with any seriousness. The problem is more stark, more discreet, and also more difficult. Business continues. Teams keep producing. Roadmaps keep moving forward. AI applications are creeping into every corner of the industry. And at the same time, entry-level positions are becoming increasingly limited.

In , INSEE (the US National Institute of Statistics and Economic Studies) observes a decline in employment in IT and information services between the end of 2023 and the end of 2025, with a particularly sharp drop among 15-29 year olds, excluding apprentices. In the same sector, however, added value continues to grow. A similar pattern emerges in the United States, where Stanford University reports a decline in entry-level employment in professions most exposed to generative AI. The market is not collapsing; it is simply filtering its entry threshold more sharply.

The tech industry continues to advance, but it’s less adept at absorbing beginners.

The Us figures tell a rather stark story. Between the fourth quarter of 2023 and the fourth quarter of 2025, employment in IT and information services activities fell by 3.0% . 

This decline is almost entirely attributable to young people: those aged 15-29 (excluding apprentices) contribute to a decrease of 3.8 percentage points , while those aged 30-54 make a positive contribution. In the fourth quarter of 2025, employment of those under 30 in this sector fell by 7.4% year-on-year.

The crucial detail that changes the picture comes right after this. The sector isn’t experiencing a complete standstill. On the contrary, INSEE notes that the added value of IT and information services activities continues to rise and, by the end of 2025, will be well above its 2000 level. In other words, the machine is still producing. It’s even producing more. But it’s doing so with an increasingly competitive job market.

This is probably what makes the phenomenon so confusing. In the collective imagination, when employment falls, we look for a visible crisis: orders down, activity faltering, massive redundancy plans, budgets frozen at all levels. 

Here, the picture is less clear. Companies are delivering. Digital investment remains on track. Tech stacks are modernizing. Co-pilots, assistants, and agents are taking on some of the daily workload. And yet, landing that first paid ticket is becoming harder.

The American parallel reinforces this diagnosis. The Stanford Digital Economy Lab observes a relative 16% decline in employment for 22-25 year olds in jobs most exposed to generative AI since the widespread adoption of these tools. Here again, the core of the phenomenon does not lie in a uniform breakdown of the labor market. It lies in an asymmetrical adjustment, concentrated at the entry level of careers.

Why does the shock hit juniors first?

The market doesn’t initially remove junior positions out of abstract cruelty or a penchant for inverted youthfulness. It reacts to a logic of tasks.

However, the logic of tasks in 2026 has little to do with that of 2021.

A considerable portion of what a team used to entrust to a junior profile now falls within the scope of generative AI . Not all the work, no. Not a deep understanding of a system. Not the ability to make architectural trade-offs. Not the detailed analysis of a production incident at 3:12 a.m. 

But the first layer, the one that often served as paid learning, is being nibbled away: standard documentation, boilerplate, simple tests, utility scripts , first-level support, reformulation of information, first pass of analysis, small repetitive corrections.

These tasks were not merely menial . They also served as a gateway. They allowed entry, seeing how the machine works, reading real code, getting a feel for operations without breaking everything, understanding where errors lie, and grasping the logic of a database, a pipeline, an API, or a CI. In short, they provided training.

Based on recent literature, INSEE summarizes this shift well: AI is increasingly replacing early-career tasks and complementing experienced profiles, capable of framing, managing and validating what the machine proposes.

The adoption context is reinforcing this trend. In Us, 10% of companies with 10 or more employees report using at least one AI technology in 2024, compared to 6% a year earlier. In the information and communication sector, this figure climbs to 42%. 

And perhaps the most interesting aspect lies in the form this adoption takes: it very often occurs through ready-to-use, off-the-shelf software. AI, therefore, doesn’t only arrive through labs, sophisticated data teams, or large-scale transformation programs. It also arrives through standard tools, where day-to-day execution takes place.

Wages are not collapsing all at once. The market, however, is becoming more competitive.

The temptation is strong to tell the story in the most dramatic way possible: fewer junior positions, therefore an immediate drop in junior salaries. This simplistic approach doesn’t hold water.

The first adjustment doesn’t happen through the payroll. It happens through the front door.

Stanford puts it quite clearly: the effects observed at this stage are primarily on employment, rather than on compensation. This is an important distinction. A reported salary that holds up reasonably well doesn’t, on its own, tell us much about the health of the market. It can mask a narrower funnel, tougher processes, more rigorous selection, and higher expectations for the same level of compensation.

This is often how wage pressure begins. Not with a blunt announcement like “all junior salaries are being cut by 15%.” The market operates more subtly, and often more insidiously. Fewer job openings. More competition. More candidates for the same position. Slower career progression. Reduced room for negotiation. Job descriptions that resemble mid-level positions disguised as junior roles. And, in the same breath, a more substantial bonus for those already capable of working in a well-equipped environment without incurring hidden debt.

However, the market does not punish all early careers with the same intensity. The profiles that hold up better are often found in the less commoditized areas of junior work: cloud computing , security, data engineering, software quality, automation, product understanding, and verification skills. 

The blind spot: by closing the door, IT weakens its successors

A junior employee doesn’t join a tech team solely to perform simple, low-cost tasks. This purely accounting-based approach always ends up costing more than it promises. 

A junior employee also joins to learn the job in a real-world environment. Not in a tutorial. Not in a well-presented bootcamp. In the real world. The real deal. The messy logs. The poorly written tickets. The incomplete documentation.

When companies close this entry gate too quickly, they take on what could be called a transmission debt .

There’s a lack of people who understand the inner workings of the platform, its vulnerabilities, its operational habits, the compromises already made at a high price, the ways to review, correct, and anticipate. There’s a lack of intermediaries . And, in many organizations, the most toxic shortage isn’t the one you see right away. It’s the shortage in the middle. Not junior enough to build a talent pool. Not senior enough to absorb all the complexity .

The result: experts carry more weight, internal promotions become less frequent, managerial bandwidth is strained, and the real cost rises later, elsewhere, and more significantly.

Herein lies the paradox. AI streamlines a portion of the work while risking breaking the learning chain if the organization confuses local speed with structural health.

Junior value is shifting: less demonstration, more control

Ultimately, the question today isn’t simply “who knows how to use AI?”. The question becomes much more challenging: who knows how to work effectively in an environment saturated with assistance?

Value increases where control begins.

Debug. Review. Test. Define a small scope. Write documentation that doesn’t offend the next team. Explain a technical choice without reciting a template answer. Understand a business need well enough to see where the “smart” proposal is nonsense. That’s where the market weeds you out.

The real shift is here. The “impressive” junior is no longer the one who generates quickly. It’s the one who maintains a minimum level of technical accuracy when the entire ecosystem pushes towards a false sense of mastery.

So, at this stage, AI hasn’t eliminated young people from IT. It has done something more subtle, and perhaps more dangerous: it has compressed entry points, shifted value, hardened the balance of power, and begun to weaken the skills pipeline that sustains teams over time. This issue goes far beyond current recruitment. It concerns how the tech industry still intends to develop its future professionals.