Skip to content
After AI
02 7 profiles

What is exposed

Exposure is not displacement. It is a description of a task profile.

−19%

shortfall for ages 22–25 in the most exposed occupations, up from 15% a year earlier

3.3%

of global employment sits in the highest GenAI exposure band

52%

of advertised jobs are being deskilled by AI, per PwC

The most important correction in this whole literature: an AI applicability score measures overlap between what a model can do and what a job contains. It is not a headcount forecast. The Microsoft researchers who produced the most-cited version say so explicitly, and point at bank tellers, whose numbers held up after ATMs arrived.GeekWire 2025

What is observable is narrower and sharper. Payroll data through June 2026 shows employment for 22–25-year-olds in the most AI-exposed occupations sitting 19% below where it would be had it kept pace — 15% a year earlier. In levels, the two most exposed quintiles fell about 11% since November 2022 while the three least exposed grew about 10%. Experienced workers show no comparable gap. The adjustment came through hiring rather than pay, and it concentrated where AI substitutes for tasks rather than complementing them.Stanford 2026

The exposure itself is clustered. One in four workers globally is in an occupation with some generative-AI exposure, but only 3.3% sit in the highest exposure band, and clerical work dominates it.ILO 2025 Employers’ own projections put data entry, clerical and administrative roles at the fast end of decline to 2030.WEF 2025

The clearest new evidence on how this happens is a split by seniority. In the most AI-exposed occupations, the entry-level share of US postings fell from 29% to 10% between 2021 and 2026, while the senior share rose from 22% to 47% — and the tilt was about a third as large in the least-exposed occupations.Indeed Sep 2026 The same employers are hiring, just not at the bottom of the ladder.

Two honest counterweights. First, not all of this is AI: the OECD notes that the unemployment risk facing young university graduates has been rising since well before generative AI spread, which points at more than one cause.OECD 2026 Second, the overall market is not collapsing — US postings are up 0.7% year on year as of September 2026, the first positive annual reading in almost four years.Indeed Sep 2026 What is changing is who gets hired, not how many jobs exist in total.

The lab that produces the headline number now publishes it monthly, so the figure on this page is a snapshot that will move.Canaries dashboard

Observed, not forecast

Who is losing ground

Ages 22–25, exposed
−19%
Experienced, same jobs
stable
High-frequency US payroll data through June 2026. The 19% is the shortfall against the path young workers in less-exposed jobs stayed on; the adjustment came through hiring, not pay. The second row is a null result, not missing data.
Task-level exposure

Where exposure lands: by band, and by country income

2,861 tasks scored by three models plus expert review. The first two rows are exposure bands; the last two are the share of all employment exposed in high- and low-income countries — a different cut of the same index.
In levels, since ChatGPT

Two paths for the same age group

Employment change from November 2022 to June 2026 for 22–25-year-olds: the two most AI-exposed quintiles fell about 11% while the three least-exposed grew about 10%.
The same postings, tilted up the ladder

Entry-level share against senior share

20212026
Share of US postings in the most AI-exposed occupations that are entry level versus senior. In the least-exposed occupations the tilt was about a third as large.

Most affected by AI, measured from real usage

GeekWire 2025
  • 01 Interpreters & translators
  • 02 Historians
  • 03 Passenger attendants
  • 04 Sales representatives
  • 05 Writers & authors
  • 06 Customer service representatives
Ranked by the share of an occupation's activities that AI already performs in real conversations. Information work dominates the top of this list.
Employer projections to 2030

Roles expected to decline fastest

  • 01 Data entry clerks Clerical, specified, verifiable
  • 02 Bank tellers and related clerks Already reshaped twice by machines
  • 03 Material-recording and stock-keeping clerks Rule-based counting in a system
  • 04 Administrative assistants Scheduling and drafting
  • 05 Cashiers and ticket clerks Routine transactions with a queue
  • 06 Printing workers Downstream of a digitising medium
  • 07 Accountants and auditors Rule-based verification at volume
  • 08 Door-to-door sales and street vendors Channel replaced, not automated
Employers were asked which roles would decline in their own organisations — a direction, not a magnitude. No percentage is published per occupation, so none is shown here.

Profiles

Exposed, and in flux

← What survives Next: what is coming →