Last episode, we said deep expertise survives automation. This episode tackles the question that answer creates: how does anyone become an expert now?

The boring entry-level work - routine contracts, standard drawings, first drafts - was never just cheap labour. It was the ladder. And AI is taking exactly that work. The latest payroll research shows employment for workers aged 22–25 in AI-exposed jobs has fallen 16% relative to other workers, while their older colleagues hold steady. Yet the same economist found the opposite result in a 5,000-agent call centre, where AI coaching made rookies 34% better, faster than any training programme ever had.

Same technology, opposite outcomes. Alex Rivers unpacks what decides which one you get — and what to do about it, whether you're starting your career or leading the team a graduate joins.

In this episode: why grunt work was actually the gym, three jobs where the training quietly vanished with the task, the flight-simulator lesson from aviation, four daily habits for early-career listeners, and three moves every leader can make to stop deleting their own talent pipeline.

Chapters (approximate)

00:00 — The listener question that stopped the show: how does anyone become an expert now? 01:45 — Canaries in the coal mine: the 16% decline hitting young workers first 03:15 — The twist: the call centre where AI made rookies 34% better 05:00 — What expertise is actually made of: reps and feedback 06:30 — Three jobs where the training vanished with the task 08:15 — Ladder remover or flight simulator: the fork every workplace faces 09:30 — Four habits if you're early in your career 11:45 — Three moves if you're the one leading 13:30 — This week's action 14:15 — How this episode was made (AI disclosure)

Key takeaways

The boring work was the training. When AI absorbs a routine task, it also absorbs the learning hidden inside it — the repetitions and feedback that build expert judgement. The same AI that removes the ladder in one workplace accelerates it in another; the difference is whether anyone designs for human learning around the tool. Early-career: do your own rep before seeing the AI's answer, use AI as a simulator not a substitute, chase feedback, and volunteer for the messy work machines can't touch. Leaders: protect a share of hand-done reps, have juniors predict before they see AI output, and redeploy veterans as mentors — because the tasks that used to transmit judgement are disappearing.