Who will become senior if AI does the junior work?

Article link: https://open.substack.com/pub/johanos...

The article argues that AI is exceptionally good at the “junior tasks” that used to train people into becoming competent professionals: first drafts, summaries, basic research, formatting, simple coding, and triage. While this looks like pure productivity, it creates a deeper risk: if AI does the junior work, how do juniors develop the judgement, context, and craft needed to become seniors? The traditional apprenticeship model worked because work and learning were intertwined, and the repetitive tasks weren’t just labour — they were the training ground where people made safe mistakes, learned standards, and built mental models.

It then warns about three knock-on effects: shallow competence (people producing outputs without understanding), an illusion of productivity (verification becomes the real work, and it’s high-skill), and a “missing middle” problem where reduced entry-level hiring leads to a shortage of capable seniors a few years later. The solution isn’t to ban AI, but to redesign development intentionally: structured “manual reps” for core skills, stronger review rituals that teach verification and reasoning, simulation-style training for edge cases, and promotion criteria that reward judgement over output volume. The takeaway is that in an AI workplace, the scarce resource won’t be content — it will be human judgement.