Labor Market by AI Exposure
BLS's Job Openings and Labor Turnover Survey is the cleanest monthly read on US labor demand. We pull it and re-cut every series by AI-exposure tier — three cohorts of industries that face very different displacement pressures. The point isn't whether the labor market is good or bad in aggregate. It's whether the AI-exposed cohort is diverging from the rest. As of March 2026, Information is running a 2.4% layoff rate against a 1.2% national average — 2.0× the rest of the economy.
Layoff rate by industry — year over year
March 2025 (open) → March 2026 (filled), % of employmentBiggest year-over-year jump: Information moved from 1.3% to 2.4% — +1.1 pp in a year. Bars are colored by AI-exposure tier; the high-exposure cohort clusters at the top of the ranking, not the bottom.
Layoff rate over time, by AI-exposure cohort
Monthly, % of employment. Cohort = mean of industries in tierEach colored line is the simple mean of layoff rates across the industries in that AI-exposure tier. A widening gap between red and green is the labor-market evidence of asymmetric AI displacement.
Quits rate over time, by AI-exposure cohort
Monthly, % of employment. Workers leaving voluntarilyThe quits rate is the cleanest worker-confidence indicator in JOLTS. When it falls, workers are afraid to leave — and a tier-specific decline in quits is a leading signal that displacement pressure is being felt on the ground.
How the AI-exposure tiers are drawn
JOLTS reports at the supersector level (about a dozen industries). We group those into three tiers based on how much of the work is the kind of cognitive, language-and-symbol task an LLM can already do at a useful quality bar:
- High exposure — Information, Professional & Business Services, Financial Activities. White-collar, keyboard-bound, document- and code-heavy work.
- Medium exposure — Wholesale Trade, Retail Trade, Education & Health Services, Other Services. Mixed: some knowledge-work tasks, but lots of in-person and physical-product work.
- Low exposure — Mining & Logging, Construction, Manufacturing, Transportation & Warehousing, Leisure & Hospitality, Government. Physical work, machinery, and human-presence-bound roles. Robotics will eventually move some of this; LLMs won't.
The tier mapping is editorial — defensible but not surgical. Financial Activities, for example, is high-exposure because back-office finance work (analyst, accounting, audit prep) is heavily AI-assisted today; but the same supersector includes commercial bank tellers and insurance branches where exposure is lower. JOLTS doesn't go below supersector, so we take the lump.
Sources
- US Bureau of Labor Statistics — Job Openings and Labor Turnover Survey (JOLTS). Series fetched include layoffs/discharges, hires, quits, and openings rates per supersector industry, seasonally adjusted, monthly.
- JOLTS reports approximately five weeks after the reference month and routinely revises prior months. The latest figure on this page reflects whatever BLS has published at the time of the last refresh.
- AI-exposure tiers are a Takeover Tracker editorial classification. See our methodology for how we score AI exposure at the occupation level.