The entry rung is thinning: what the 2025 evidence says about AI and careers
Exposure by task and career stage, the apprenticeship problem, and what a person does about it
The question: If AI takes the work juniors used to learn on, how do people become experts — and what should you do about your own ladder?
Lesson 14 showed exposure is wide and replacement narrow. The 2025 payroll evidence adds a sharper finding: the effects concentrate on early-career workers in exposed occupations. That changes the advice to individuals and the duty on firms.
What the lesson covers
The evidence, carefully. Brynjolfsson, Chandar and Chen ("Canaries in the Coal Mine", August 2025), on payroll data for millions of US workers: employment for early-career workers (22–25) in the most AI-exposed occupations fell relative to less-exposed ones, while experienced workers in the same occupations held or grew. Exposure is still not replacement — but the entry rung is thinning where the competent-output layer commoditised (Lesson 38).
Why juniors first. The work that trains a novice — the first draft, the routine analysis, the standard reply — is exactly the competent cognitive output that got cheap. Firms stop hiring for it; the people who would have learned by doing it do not get to. The Anthropic Economic Index (2025) shows usage concentrated in exactly those augmentable, mid-skill tasks. WEF's Future of Jobs 2025 sees roles created and displaced in parallel through 2030 — churn, not collapse, with the burden landing on entrants.
The apprenticeship problem is the firm's version: if the junior work is automated, where do the next seniors come from? The answer has to be designed — juniors who verify the machine's output against the expert's, who learn to frame before they generate, who get accountability in small doses early (Lessons 38 and 42). A firm that only banks the junior saving will buy its experts later from whoever did not.
For the individual, the advice follows the premium (Lesson 38): move toward accountability, relationship, framing, verification and proprietary knowledge; use the tools hard but keep doing the thinking steps whose loss you would regret (Lesson 34); and treat Weiss's "jagged frontier" exercise as a personal map — find where the tools fail in your work and stand there.
Be honest about uncertainty: the 2025 studies are early, US-centred and about the first years of diffusion. They are strong enough to act on — redesign the apprenticeship, aim your own development at the rising layers — and not strong enough for prophecy in either direction.
Key points
- Early-career workers in AI-exposed occupations are losing ground while experienced workers hold (2025 payroll evidence).
- Juniors go first because the work that trains them is the layer that commoditised.
- The apprenticeship must be redesigned: verify, frame, small accountability early.
- Individuals: aim development at accountability, relationship, framing, verification, proprietary knowledge; keep the thinking steps you would regret losing.
- The evidence is strong enough to act on and not strong enough for prophecy.
Framework — Map exposure by task → Locate your rung → Move toward the premium → Protect the apprenticeship
For a person: exposure map, career-stage honesty, development aimed at accountability/framing/verification. For a firm: redesign how juniors become seniors.
The lab
Map your own role's exposure and write the apprenticeship problem for your team.
Open this lesson, its lab and its quiz
Sources and further reading
- Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of AI — Brynjolfsson, Chandar & Chen, 2025
- Future of Jobs Report 2025 — World Economic Forum
- Anthropic Economic Index — Anthropic, 2025