AI is supposed to come for women's jobs first. The usage says the reverse.
The common claim is that AI comes for women's work first: admin, care, service. Real Claude usage runs the other way. Male-dominated occupations lean more toward automation-style use than female-dominated ones do. What AI doesamplify is the sorting that was already there. The roles it's scored to shrink are 69% female; the ones it's scored to grow are 68% male. So the occupational gender gap doesn't close. It hardens.
Augmentation runs 49.3% in female-dominated occupations, 43.8% in male-dominated ones.
Cross real Claude usage with the gender make-up of each occupation and the automation tilt points away from where the headlines aim it. In male-dominated occupations (156 of them, 10.4M workers), usage leans more toward doing the task. Augmentation share sits at just 43.8%. In female-dominated ones (10.1M workers) it leans more toward assisting the person, at 49.3%.
Measured: AEI augmentation share, Claude.ai usage, Apr+May 2026, CC-BY. Claude users only; “automation-leaning usage” is how AI is used, not a count of jobs lost; 356 occupations with 5k+ workers, employment-weighted. This complicates the guess-based zone story below, so both stay in view. Method →
Same AI exposure. Opposite gender composition.
1.0M women in roles that are both high-exposure AND female-dominated.
| Role | Women employed | Outlook |
|---|---|---|
| Admin assistant | 365.0K | -3% |
| Office manager | 95.0K | -3% |
| Bilingual secretary | 67.0K | -3% |
| Medical secretary | 43.0K | -3% |
| Health records clerk | 51.0K | -3% |
| Payroll administrator | 118.0K | +3% |
| Bookkeeper | 118.0K | +3% |
These seven roles alone employ 857,000 UK women. They concentrate in admin / clerical pipelines where AI now performs most of the routine task load. Transition paths do exist (trades apprenticeships, technical qualifications, regulated professions), but none are well-signposted for this cohort yet.
2.2M women across the UK in AI-exposed + inelastic roles.
The two rankings tell different stories. London and the South East have the biggest absolute counts, because they're the biggest regional economies. By density, Northern Ireland (23.4%) and the East (20.5%) carry the heaviest proportional load. In those regions the mid-skill admin and clerical economy is a larger share of total female employment.
The gap the AI split amplifies.
The gap peaks at 50-59 at £15.1K (35.3% of the male median). Those are the prime earning years, when male careers tend to sit in Jevons-protected sectors (engineering, finance, tech) while female careers concentrate in the Anti-Jevons zone (admin, clerical, care). The AI split doesn't create the gap. It cements it.
The gender split here comes from pre-existing UK sector sorting, not from anything AI is doing to women specifically. What's worth noticing is how cleanly the split maps onto AI exposure, and it does that because admin and care were already where AI overlaps.
Regulated professions (nursing, allied health, accountancy) preserve more female employment than general knowledge work. Trades apprenticeships remain ~2% female, and that number would have to move a lot to change the 2030 composition. It's a signposting problem more than a training problem.
The /regions detail pages break this down geographically. Northern Ireland and the East of England carry the heaviest proportional female-at-risk load. /education shows which education tracks lead into the AI-exposed cohort versus the Baumol-protected one.