Which jobs are most threatened by AI? New research maps the risks

TL;DR: New research from GovAI and the Brookings Institution identifies which workers face the highest risk from AI automation and, for the first time, which can least adapt to displacement. About 6.1 million clerical and administrative workers score high on both measures. Women hold 86% of these most vulnerable positions.

The study goes beyond the standard approach of measuring AI “exposure,” which estimates how many job tasks overlap with AI capabilities. Researchers Sam Manning and Tomas Aguirre added a second dimension: how easily displaced workers could transition to other well-paying work, based on education, varied work experience, wealth, age, and local job availability.

The findings

Occupations like web design and secretarial work both score high for AI exposure, but they diverge sharply on adaptability. Web designers, typically younger with transferable technical skills and higher savings, are well-positioned to find alternative work. Secretaries and administrative staff are not.

The most vulnerable group consists largely of clerical workers with limited formal education, narrow work experience, and fewer financial resources to weather a career change. The gender dimension is stark: women make up roughly 86% of workers in this most-at-risk category.

Mark Muro, a senior fellow at Brookings, noted these workers “may be out of sight and out of mind” to policymakers. The research highlights a risk that AI-driven job losses will disproportionately affect people already with fewer options.

The prediction problem

The researchers themselves urge caution. Economists have consistently failed to predict how technology changes work. ATMs did not eliminate bank tellers. Earlier AI was supposed to replace radiologists. A prominent 2013 study predicted automation would destroy nearly half of all jobs, which has not happened.

What does appear to be different this time is that white-collar workers, not factory staff, are first in line. There is no measurable evidence yet that AI is reducing overall US employment, but the pattern of which jobs face pressure has shifted.

Looking forward

For UK businesses and policymakers, the findings carry direct relevance. Britain’s economy has a similar concentration of administrative and clerical roles, many held by women. If AI displacement follows the pattern this research suggests, the UK will need retraining programmes and transition support aimed at workers who currently have the fewest resources to adapt on their own.