Search the Practicum

Type two letters · ↑↓ select · Enter open · Esc close

Type at least two letters.

What is happening to entry-level and mid-career work

Live Practicum page for Chapter 4 (called Ladder Watch in the book). Verified: 2026-08-18. Data on hiring young workers gets revised several times a year, and the studies disagree with each other. Check the date and the caveat on every row. Do not take our word for it. Check.

Chapter 4 holds two facts at once. AI raises the floor: a beginner working with a model does better than a beginner working alone. AI also narrows the door: the simple work people used to learn on is now cheap enough to skip. Both are supported by evidence, and neither cancels the other.

These are the sources you can use to check that picture yourself.

What the research shows

Source Date What it shows Caveat
Brynjolfsson, Li, Raymond, "Generative AI at Work" (NBER WP 31161 / QJE) 2023, journal version 2025 5,179 customer support agents resolved 14% more issues per hour on average. The largest gains went to novices and lower-skilled workers, because the model passed along what the best agents already did one company, one kind of work; it does not transfer to jobs without a clear right answer
Brynjolfsson, Chandar, Chen, "Canaries in the Coal Mine?" (Stanford Digital Economy Lab) November 2025 In the occupations most exposed to AI, employment for workers aged 22 to 25 declines relative to older workers in those same occupations the size of the effect depends on the model and the controls; the authors report a range, not a single number
Federal Reserve Bank of Dallas, young workers in AI-exposed occupations January 2026 The decline among young workers comes mostly from a weaker move from outside the workforce into a job, not from mass layoffs. Aggregate unemployment is barely affected use this one as a brake on panic: AI is not the only cause here
Hartley, Jolevski, Melo, Moore, "The Labor Market Effects of Generative AI" (SSRN) 2026 GenAI adoption among U.S. workers is climbing fast, but the authors find no statistically significant drop in job openings or total jobs in exposed occupations working paper; the adoption numbers shift between PDF versions
Brookings with Opportunity@Work, "How AI may reshape career pathways to better jobs" 2026 STARs are workers who built real skill through jobs, military service, apprenticeships, or community college rather than a four-year degree. AI erodes the gateway roles they used to climb through a report with estimates; check the exact counts against the tables in the original
Eurostat, euro area unemployment June 2026, released Jul 30, 2026 Youth unemployment in the euro area was 14.8%, down from 14.9% in May, with 2.351 million people under 25 out of work. Overall euro area unemployment was 6.3% an official statistical series, and the strongest available brake on collapse narratives. It is an aggregate: a real squeeze inside one occupation can hide inside a stable average
Matt Beane, The Skill Code (Harper Business) 2024 Skill grows on three things: challenge, complexity, and a live connection to someone who already knows. Take the simple work away from the beginner standing next to an expert, and you break how experts get made a book and field research, not labor market statistics

How to read this

A narrower door is not a collapsing market

Stanford finds young workers losing ground in specific occupations. Hartley does not yet see it in the economy-wide numbers. The Dallas Fed explains part of it as young people struggling to get in rather than getting pushed out. All three can be true at once, and together they say "the entrance got tighter," not "the work is gone."

The picture outside the US

Almost every study above uses US data, which is a limit worth naming. The best non-US check available is the official European series, and it does not show a collapsing entrance: euro area youth unemployment was essentially flat through 2026 and slightly lower in June than in May.

There is also a pattern worth knowing by name. Where junior postings survive, employers have been loading them with judgment, stakeholder work, and review duties that used to belong a level up. Researchers call it seniorization. It matters because it changes what "entry level" means: the door is not only narrower, the first room behind it is different. That is the same movement Chapter 5 describes, arriving earlier in a career than anyone planned.

Helping the beginner and removing the beginner are different events

Brynjolfsson shows AI lifting weaker workers. But if the same model makes that job unnecessary as a position, there is nobody left to lift. That is the trap in this chapter: the tool that teaches is also taking away the place where people learned.

Notice who is counting

Predictions from AI company executives about "half of entry-level jobs" are the position of an interested party, not a measurement. The table above leaves them out on purpose. It holds research and official statistics only.

What to do with this

Sources

Version: 2026-08-18.

What is happening to entry-level and mid-career work