Will Generative AI Hit Women Workers Harder? Brookings 2024 Data Says Yes

36% of women work in occupations where generative AI could reshape at least half of the daily tasks. For men, the figure is 25%. That 11-percentage-point gap is not a rounding error — it is a warning sign that Brookings pulled from ChatGPT-4 task exposure scores across 1,000+ U.S. occupations. Fact — [Brookings 2024]

If you are a woman reading this, you have probably already sensed it. The admin-heavy, documentation-heavy, back-office roles where women cluster are also the roles where large language models are quietly eating the workflow. The Brookings team did the math — and the numbers are worse than most headlines suggest.

Who Is Actually Exposed — And Why Gender Keeps Showing Up

Brookings used OpenAI's task-exposure framework, cross-referenced with O*NET task inventories and BLS occupational employment data, to score how much of each job's daily work can be meaningfully assisted or replaced by current generative AI. [Fact] They then layered in Pew demographic data to see who actually holds those jobs.

Here is the pattern that jumped out. Over 30% of all U.S. workers are in occupations where 50% or more of daily tasks are exposed to disruption. [Fact] Pull even further back and 85% of workers will see at least 10% of their tasks touched by the technology. [Fact] Almost nobody gets to sit this one out.

But the burden is not evenly spread. The five highest-exposure occupational families are:

• Computer and mathematical work (think software developers reviewing AI-generated code instead of writing from scratch)
• Business and financial operations (including financial analysts and accountants and auditors whose modeling and reconciliation tasks are partly automatable)
• Engineering
• Office and administrative support — roles like administrative assistants and bookkeeping, accounting, and auditing clerks
• Legal work, where paralegals and legal assistants are on the front line of contract review and research workflows

Three of those five — business and finance, office/admin support, and legal support — are majority-female occupations in the U.S. labor market. Office and administrative support alone employs roughly 19 million Americans, and the female share of that category is well over 70%. [Fact] That single fact is doing most of the work behind the 36% vs. 25% gap.

The Uncomfortable Layer Brookings Added: Bargaining Power

Exposure is only half the story. The other half is whether you have any leverage when the tasks inside your job start to change.

Brookings flagged a detail that rarely makes it into the splashy takes: union representation in finance is about 1%. [Fact] That is not a typo. When productivity software reshapes the work of a financial analyst or a claims processor, there is effectively no institutional counterweight negotiating over training, pay, or task redesign. Contrast that with education or healthcare — medium-exposure sectors — where unionization is meaningfully higher and where workers historically have had more voice in how new tools get deployed.

So the story is not "AI will replace women." It is narrower and more honest. Claim — [Brookings 2024]

The story is that the occupations most exposed to generative AI happen to employ a lot of women, and those same occupations happen to have among the weakest collective bargaining footprints in the U.S. economy. When the wave hits, the people standing in its path have the fewest formal tools to negotiate the terms.

What the Low-Exposure Column Is Trying to Tell You

Brookings' low-exposure list is interesting for what it contains and what it doesn't. Manual, blue-collar, and in-person service occupations — construction, food prep, personal care — score low on task exposure. [Estimate] That matches what most of us already suspect from watching the tools: current generative AI is strong at text, code, and structured data, and still awkward at embodied, physical, context-heavy work.

For the first time in a generation, a general-purpose technology is biting white-collar and office work harder than it is biting physical labor. That is a reversal of the automation story from the 2010s, when warehouse robotics and trucking (remember all the pieces about truck drivers?) dominated the headlines.

If your job is in the medium-exposure band — a customer service representative working alongside an LLM, a lawyer using AI for discovery, a nurse using AI charting — the Brookings data suggests a third lane. Tasks change. Jobs do not disappear wholesale. But the mix of what you do day to day does shift.

So What Do You Actually Do With This

A few things are worth saying plainly.

First, know your exposure score. If you are in one of the five high-exposure families, assume 30-50% of your current tasks will look meaningfully different within 3-5 years. [Estimate] That is not a forecast of unemployment. It is a forecast that the content of your workday changes, and the people who adapt their task portfolio fastest keep the most leverage.

Second, if you manage teams dominated by women — admin, finance ops, paralegal, customer support — this is a retention issue, not just a productivity story. The workers most affected by the task churn are the ones with the weakest formal bargaining position. Whatever training, redeployment, or wage policy you have today was probably designed before the exposure profile skewed this way.

Third, the Brookings data will keep updating. GPT-4 was the task-exposure proxy; newer frontier models push the exposure curve further into tasks that used to require judgment. [Claim] The gender gap in the 2024 data is a floor, not a ceiling.

Sources

• Muro, Mark, Maxim, Robert, Hathaway, Shriya Methkupally, Mark Muro. "Generative AI, the American worker, and the future of work." The Brookings Institution. October 10, 2024. Link
• Underlying data: OpenAI ChatGPT-4 task-exposure scores across 1,000+ occupations; O*NET task inventories; U.S. Bureau of Labor Statistics Occupational Employment and Wage Statistics; Pew Research Center demographic overlays.

Update History

• 2026-04-17: Initial publication based on Brookings 2024 report. Highlights the 36% vs 25% female/male exposure gap, the 19M office and admin support workforce, and the 1% finance-sector union density as the three defining data points.

---

AI-assisted analysis. This post was drafted by an AI research agent, reviewed for factual accuracy against the Brookings source, and published under editorial oversight at aichanging.work.