Work

Stop hiring juniors today, and in ten years you have no seniors left.

AI replaces beginner work. But careers don't start in the middle.

Companies hire juniors for three reasons. Cheap labour, work billed at day rates a senior wouldn’t charge, and a pipeline for their own seniors five, ten, fifteen years out. The first two reasons just collapsed thanks to AI. The third is long-term, and it doesn’t show up in next quarter’s report.

The work juniors used to start with, generative AI now does in seconds. The single company acts rationally when it hires fewer juniors. And that’s already happening: employment among career entrants is measurably falling, in the US as in Austria. Among older workers in the same jobs, it isn’t.

If every company acts this way, the senior generation goes missing in ten years. Judgment comes from ten years of cases you had to see through. From the contract you drafted wrong. From the patient you missed. From the code that crashed live at the customer’s site. AI delivers the answer, but not the experience of the question.

The hard cases remain, because those are exactly what no current language model solves reliably. Who handles them in fifteen years, once the path that leads there has been cut off? Whoever hires and trains juniors now is paying for a common good everyone will benefit from. That’s a textbook market failure.

The efficiency gains show up in every balance sheet. The career ladder we’re sawing off shows up in none. That’s the question nobody asks.

We need to talk about this

What does this topic stir up in you?

In depth

What the beginner work used to be

Pulling research together, drafting memos and standard contracts, cleaning up datasets, writing simple queries, coding building blocks to spec, prepping meetings. None of these tasks were ever the point. They were the practice. Review a hundred contracts and on the hundred-and-first you spot the clause that doesn't fit. Work through fifty bug reports and on the fifty-first you hear the bug in the user's tone of voice.

These are exactly the tasks generative AI now does almost for free. For the single company that's an efficiency gain. For training the next generation it's a hole. The tool that beginners used to sharpen their judgment is no longer in their hands.

What the data shows

The Stanford Digital Economy Lab study "Canaries in the Coal Mine" (November 2025, Brynjolfsson et al.) analyses the pay records of millions of US workers at the largest payroll provider. The finding: 22- to 25-year-olds in occupations whose routine work AI can take over (programming, research, customer service) saw a relative drop in employment of roughly 16 percent since the end of 2022. In the same occupations, but among older workers, employment held steady or grew. The effect shows up mainly in hiring, not in pay.

In Austria the number of apprentices fell by 3.4 percent in 2025 to around 103,000, according to the WKO. In industry, the number of first-year apprentices dropped by 14 percent. Demographics explain part of it, not the industry figure. In 2025 the US venture capital firm SignalFire analysed how large tech companies hire: the share of fresh university graduates sits at around seven percent. Three years ago it was more than twice that.

Who decides in ten years

The hard cases are the ones AI doesn't solve. The unusual contract, the patient with three pre-existing conditions, the system whose failure emerges from the interplay of four components. Maybe AI solves those too in two years. Even then, someone has to be able to judge whether the result is right, and that ability to judge comes from experience. These cases always landed on the senior's desk. But the senior becomes a senior by having been a junior. Take away the bottom rung and nobody reaches the top.

Structures that could organise this climb already exist: apprenticeships, trainee lawyers, university hospitals, graduate programmes. They cost money, the very money AI is now saving. You could train juniors differently too, more intensively, in less time. Except that costs exactly the money the savings were supposed to deliver, and nobody has yet shown how judgment forms without your own mistakes. As long as every single company rationally decides to stop funding juniors, the pipeline disappears collectively. No one is responsible for it. The gap becomes visible only once today's seniors leave and the cohorts are missing, the ones who should have started now: in the 2030s.

Sources