The AI Employment Shock Among Women is Coming
Nourina Hertz: Emeritus Professor at the Policy Lab at University College London (UCL), where she leads research on artificial intelligence.
In 1986, “Ann” — a data entry clerk — watched an IBM personal computer on her desk. Within a year, her job vanished. Four decades later, “Nathalie,” a social media manager, observes the chatbot “ChatGPT” taking over the posts she once crafted. But her exit from the scene may come even faster than Ann’s.
In July, a new report published by researchers at Microsoft made headlines by outlining a list of forty professions most at risk of being replaced by artificial intelligence. The report included sales representatives, translators, editors, and other knowledge-based roles, signaling an imminent end for many administrative jobs.
However, the report’s authors and subsequent news coverage seem to have overlooked a critical point: the upcoming disruption will not be gender-neutral. According to the U.S. Bureau of Labor Statistics (BLS), women make up the majority — about 60% — of the listed occupations.
As AI begins to consume everyone’s lunch, it appears likely that this technology will target women’s jobs first and more rapidly. Just as the rise of computers in the 1980s displaced legions of secretarial and data entry positions predominantly held by women, this latest wave of automation is expected to disproportionately affect female workers. A recent study by the International Labour Organization projected that women’s jobs in high-income countries are about three times more likely to be automated compared to men’s.
The story of the computer revolution serves as a cautionary tale. Many women who lost their jobs during the 1980s never fully recovered; they either found lower-paying work (primarily in services and care sectors) after long periods of unemployment or exited the workforce altogether. When the Bureau of Labor Statistics tracked the outcomes of workers displaced during this period, the findings were stark: women were about twice as likely to leave the labor market afterward compared to men.
Given that women are already in an economically disadvantaged position relative to men — earning less, owning less, and retiring with smaller pensions — policymakers must prepare for the impact of AI, which will likely hit women’s jobs harder, and develop strategies to mitigate its effects.
In crafting their responses, they should also recognize that not all secretarial, data entry, and typing jobs performed poorly in the 1980s: women who adapted to technology and acquired new relevant skills were able to achieve better outcomes.
For now, let’s set aside whether the concept of “upskilling” is excessive in an era where AI is expected to surpass human intelligence, and instead assume that we will witness a transitional phase where those with AI-related skills outpace those without them. The AI Jobs Barometer for 2025 by PwC found that workers with AI skills receive a wage premium of 56%, a significant increase from the 25% premium the previous year.
This indicates that we must ensure women gain full proficiency with the new technology — or at least attain a comparable level of proficiency to their male counterparts — if we wish to prevent them from becoming the most adversely affected by AI.
While nearly equal numbers of men and women now use ChatGPT for personal tasks, a clear gender gap has emerged in the workplace. A recent survey of U.S. workers revealed that while 36% of men use generative AI daily at work, only 25% of women do. The survey also indicated that 47% of men feel confident using this technology at work, compared to around 39% of women.
This gap likely reflects the reality that women are generally more concerned than men about the increasing use of AI — a healthy concern that should be shared by all. However, another reason may be that companies invest more in upskilling their male employees in AI than in training women. In a global survey of 12,000 professionals conducted by Randstad this year, 41% of men reported that their employers provided access to AI, compared to about 35% of women, while 38% of men said they had opportunities to build AI skills, compared to about 33% of women.
Less technology use — coupled with fewer opportunities to use it — is a harmful mix for female employees, especially as firms increasingly cite “AI fluency” when deciding which employees to retain or promote.
Failing to address this issue could also expose employers to legal risks. In the UK, workplace policies that systematically harm women’s advancement opportunities — including providing fewer opportunities to upskill in AI — may constitute indirect gender discrimination under the Equality Act of 2010. This holds even if the company did not intend to discriminate. Under this law (and similar legislation in other countries), what matters is the outcome, not the intent.
Thus, business leaders should ask themselves fundamental questions. Who has access to AI tools? Who is invited to participate in experimental AI projects and initiatives? Who receives training on AI?
It appears that governments worldwide are utterly unprepared to tackle the potential job disruption ignited by AI, particularly as it affects women. As policymakers work on strategies to mitigate the risks of AI, it is imperative that gender concerns are prominent on the agenda — not just for ethical reasons.
At a time when political polarization is rising and traditional parties are losing ground, securing votes from women is crucially important. Ensuring that women do not bear the brunt of job displacement caused by AI, and addressing other gender disparities related to AI, is not only the right thing for governments to do but also highly pragmatic. After all, women like Nathalie are numerous.
