Harvard: AI is cutting junior hiring, and within three years the damage could spread across the whole industry

Generative AI is already reshaping hiring in IT: companies are cutting entry-level positions, while experienced developers lean ever more heavily on AI tools. If both trends persist, the market may face not a surplus but a shortage of people able to maintain and fix the code that has piled up.

What Harvard found

The Harvard study covered almost 62 million workers and 285 thousand companies in the US. Its main conclusion is simple: after adopting generative AI, companies start hiring juniors noticeably less often. The drop in hiring is estimated at 9–10% over six quarters, while demand for more senior roles barely changes.

The authors stress one important detail: this is not about mass layoffs but about a slowdown in entry into the profession. The bottom rung of the career ladder shrinks first. Other signals collected in the article reinforce the picture:

• employment of developers aged 22–25 fell 20% from its late-2022 peak to July 2025, according to the Stanford Digital Economy Study
• the number of entry-level job postings in the US fell 35% between January 2023 and June 2025, according to Revelio Labs
• in the UK, Revelio Labs puts the drop in such postings at 46%
• the share of juniors in IT hiring fell from 15% to 7% in three years, according to Stack Overflow

Taken separately, these figures could still be blamed on the market cooling after the overheated 2021–2022 period. Together, they point to a more troubling shift: companies are already behaving as if AI has taken over a significant share of the tasks once handed to newcomers. What matters most is that this is not a one-off spike but a series of matching signals across different samples and countries. If juniors don't enter the system now, there will be nowhere for mid-level engineers to come from in a few years.

Dependence on assistants

The second warning sign came from METR research. The organization tested 16 experienced open-source developers on real tasks in projects they knew well. Participants were optimistic: they expected AI to speed up their work by about 20%. In practice the opposite happened: when assistants were allowed, tasks took 19% longer on average. In other words, the tools felt useful, but in fact added time for prompting, waiting, checking and fixing code.

"It's like crossing the city on foot when you're used to taking an

Uber."

In the next phase, METR ran into a question not just of productivity but of habit. According to data cited in the article, 30–50% of developers refused to submit some tasks to the experiment if they had to be done without AI. The researchers explicitly pointed to a selection effect: the people who depend most on assistants are the ones dropping out of the sample.

This does not prove that AI is useless. But it shows how quickly it is becoming more than an accelerator, turning into a crutch that many now find hard to give up.

Where the risk lies

The problem is not limited to newcomers finding it harder to land a first job. The very logic of growing engineers inside teams is disappearing. Previously, a junior took on routine work, made mistakes, got code reviews and gradually became a mid-level engineer. Now that routine is often handed to a model, and the senior takes on three roles at once: their own main work, oversight of the AI and final review of the output.

This creates a talent doom cycle: companies save on junior positions today, and in 3–5 years discover a shortage of people at the next level. The cost of mistakes is rising too. The article cites estimates that AI code more often carries vulnerabilities, increases duplication and reduces the share of refactoring. Even if one argues over the exact percentages, the mechanism is clear: writing draft code has become easier, but untangling its consequences has not. So part of the speed gain turns into deferred tech debt.

If fewer juniors are hired and seniors spend more time reviewing machine output, then in a few years the industry may get not an army of hyper-productive developers but a bottleneck of overloaded specialists and bloated codebases.

What it means

The IT market is now testing a model in which AI covers the bottom layer of engineering work. In the short term it looks like savings, but in the long term it may break the talent pipeline. If companies do not bring back meaningful hiring and training of juniors, there will simply be no one to untangle the accumulated AI tech debt in a few years.