(Bloomberg) -- Over the past six months, as I’ve gone about my day as a journalist, teacher and parent, I’ve conducted a highly unscientific study. In conversations from the newsroom to the school gates, I’ve asked the same question repeatedly: “Do you use generative AI?”
The responses vary wildly.
At one end of the spectrum, a mother of three from New Jersey, who doesn’t work outside the home, told me that generative AI scares her “to death.” At the other, a father of one from Texas, who works in consulting, said that he uses it “all the time and for everything.” In between? Varying shades of enthusiasm. A young graduate student from California told me she’s never tried it but intends to. Her classmate from Chicago said he uses it “occasionally, mostly for writing emails.”
In my crude survey of about 30 or 40 people I regularly interact with, men were often quick to admit to using generative AI tools like ChatGPT, Copilot and Gemini frequently and shamelessly. The women I spoke to — with the notable exception of those who work in tech and media — were more reluctant. Many seemed disinterested.
Having reported and written for years on the causes and consequences of gender gaps, this dichotomy concerned me. Was I seeing evidence of the latest iteration of inequality between the sexes? And might it signal something consequential about how men and women work and are paid? Or was this merely a quirk of my particular social circle and small sample size? I started digging.
A Generative AI Gender Gap
Because of the relative novelty of generative AI models and technology, comprehensive studies on who uses them, how frequently and for what have so far been hard to come by. Nonetheless, one research paper from July corroborates my flimsy poll. Based on responses to a Federal Reserve Bank of New York consumer survey, economists at the Bank for International Settlements (BIS) indeed established the existence of “an economically and statistically significant ‘gen AI gender gap’.”
On average, they found that half of all men reported having used generative AI over the previous 12 months, while only 37% of women did. Among those who said they used it weekly, the gap was similarly large.
This is fresh evidence of a well-established technology gender divide: In its early days, the internet was dominated by men. More recently, research shows that women are significantly less likely than their male counterparts to use financial technology — or fintech — products. But the generative AI gap is noteworthy because of the extent to which this particular technology is already radically and rapidly transforming the labor market.
Last year, a McKinsey analysis predicted that 29.5% of the hours worked in the US economy at the time of their research will be able to be automated by 2030 — with 8 of those percentage points attributed specifically to generative AI. Goldman Sachs, meanwhile, has forecast that AI could automate up to the equivalent of 300 million full-time jobs globally over 10 years.
Research is also already showing that generative AI can substantially raise productivity across certain jobs, including grant-writing and marketing. One study found that consultants at Boston Consulting Group who used GPT-4 worked faster and more effectively compared to those who didn’t.
A Costly Fear
Why are women less likely to use generative AI? The BIS economists concluded that respondents’ self-reported knowledge about the technology accounted for almost two-thirds of the gap. The remainder could be explained by varying levels of trust in technology. “Women are generally more concerned about the negative consequences of sharing data,” they write, while men “see greater benefits for their job opportunities and lower risks from the use of gen AI.”
To be sure, the conclusions come with caveats. All of the 890 survey respondents were based in the US, and about 82% were white. Almost 60% had a bachelor’s degree or higher, and 43% were earning more than $100,000 annually. This naturally limits the extent to which we can extrapolate. But the findings do align with other studies showing that women are less likely to trust technology and that men are more likely to be confident in their tech savvy.
It’s therefore necessary for us to entertain an inconvenient prospect: If women are more risk-averse and fearful of technology, at least on average, and if an appetite and willingness to adopt new technology is a precondition of being able to thrive in a brave new labor market, generative AI could feasibly exacerbate the gender pay gap.
When I put this prospect to Sander van’t Noordende, the chief executive officer of Randstad, the world's largest employment agency, he gave it to me straight: “The way technology is today, you can’t afford to check out.” He urged me to think about it in terms of an equation: “Pay should be based on productivity. And AI — simply put — is a productivity booster.”
The Most Vulnerable Jobs
Today, the international gender pay gap stands at about 20%, with wide variations from country to country. Occupational segregation — or the tendency for a certain demographic group to dominate an industry or job type — is one of the most important drivers of the rift.
That’s particularly bad news in the context of generative AI, because the jobs it is most likely to disrupt — or even render redundant — are also the jobs disproportionately held by women.
Analysis done last year by the UNC Kenan-Flagler Business School found that eight out of 10 women, or about 59 million individuals in the US workforce, are in occupations that are “highly exposed to generative AI automation,” compared to about six out of 10 men. Top of the list of exposed jobs are office and administrative support roles. Jobs that require physical labor — and are still mostly done by men — are least likely to be affected by AI.
So, is the lesson here that women in those professions should pivot hard into another industry? Not necessarily. “As the saying goes,” the authors of the report accompanying the UNC analysis write, “You won’t be replaced by AI. You will be replaced by someone who knows AI.” If this seems like a familiar narrative, that’s because it is. Women, and notably women of color, also bore the brunt of Covid-related job losses, because they tended to be overrepresented in lower-skilled positions that were more vulnerable to cuts. But unlike the pandemic, generative AI is not going anywhere.
Stubborn Conditioning
So, what’s to be done? It’s a hard question to answer, not least because of that risk aversion some women feel toward new technology — and toward sharing personal data — is likely rooted in stubborn social conditioning.
Research shows that women, in many social and professional circumstances, have historically faced graver consequences than men when they’ve failed. That has translated into a resistance to taking risks because women often (correctly) perceive themselves to be at a higher risk of negative consequences. In the context of using generative AI, failing might mean getting called out for trying to pass a piece of writing off as one’s own, when it was written by a machine — and then being labeled, correctly or not, as unethical or even just lazy.
“This shows up in many different ways,” Laura Globig, a cognitive neuroscientist and postdoctoral researcher at New York University, told me. “Studies show that women, for example, considered themselves to be at higher risk than men of getting sick during the Covid-19 pandemic.”
But when it comes to AI, there are ways to counter this dynamic.
“The more in control we humans feel we are, the more willing we are to take risks, and that applies to using AIs too,” Globig says. “So if we empower women with digital literacy, that could potentially be a really important way of closing the gen AI gender gap.”
One woman who is not risk-averse when it comes to technology is Shubhi Bhonsle-Rao, who has defied tired stereotypes around what a woman — and a woman of color, at that — should and shouldn’t do. An engineer by training, Bhonsle-Rao worked her way up the corporate ranks to serve as vice president and treasurer of Alphabet Inc. after a stint as the treasurer of retailer Tesco Plc in the UK. In 2020, she quit Big Tech to start her own company, Uplevyl. Its mission is bridging the gender gap by training women to become more confident in their digital skills — and it uses AI to do it.
Bhonsle-Rao acknowledges the gender risk disconnect when it comes to technology, and she agrees that — as the workplace evolves and as jobs are displaced by technology — roles predominantly held by women could be most vulnerable. But she’s also noticed things that are sparking hope. “A lot of women are reaching out to me because they want to improve their AI and digital acumen,” she says. “And that’s certainly encouraging.”
Having people like Bhonsle-Rao doing what she’s doing might also inspire optimism, as right now the industry producing generative AI is powered predominantly by men. One analysis, conducted in Germany of almost 1.6 million AI jobs worldwide, found that women comprise only 22% of AI talent — and even less at senior levels. That raises worries about gender bias creeping into AI design. If we want to broaden the users of AI, we need to ensure that the faces of those who are building it change, too.
One notable wrinkle here, notes Heather McCulloch, a senior fellow at The Aspen Institute, whose research focuses on women in the economy, is that learning something new takes time. “And time is not something that many women have in abundance,” she says. Women in many countries still do vastly more unpaid labor than men, McCulloch points out. Meanwhile, in some countries — including the US — women’s paid labor force participation is at a record high. Carving out time to learn something new might be a noble ambition but is not always a realistic prospect.
Acknowledging this, Globig explains that many women — because of work distribution patterns and occupational segregation — might not be as aware of just how much upside there is to using generative AI. Therefore, they might not be prioritizing it as much as men, especially considering how busy they are. “But the truth is, regardless of gender, gen AI can be a massive tool for economic mobility,” she adds. “So I just hope that women recognize that. The potential is too big to ignore.”
At the end of November, Deloitte published a report that found the share of US women using or experimenting with generative AI had tripled in 2024 from the year prior, compared to a 2.2 times increase for men, which helped to narrow the gender gap. The authors made a bold prediction: Despite the persistent generative AI gender divide, women are poised to eclipse men in its use — as early as next year in some countries.
They also found that women consistently trusted AI less than men did. But that trust gap disappeared when they looked exclusively at men and women working in tech. Women working in tech were also more likely than women in other fields to say their employer encouraged the use of generative AI and provided training.
Some of those skeptics I encountered at school drop-off and in the classroom no doubt still dismiss generative AI as a flash-in-the pan fad. To that I would respond: Don’t knock it until you’ve tried it.
Even while reporting this essay, I’ve gone from being a curious generative AI dabbler to a full-fledged convert. My last few prompts have included a request for holiday gift ideas and a suggested lesson structure for a class on storytelling. I’ve also asked ChatGPT to edit something I wrote in German — a language in which I’m fluent but don’t regularly write. The result was impressive. I’ve never asked generative AI to write emails for me. That still somehow feels fraudulent. I’ve not ruled out experimenting with it as a therapist, though. When I first heard that some people were doing that, I dismissed it as weird. But the lack of out-of-pocket fees is quite a selling point.
Finally, allow me to remind the skeptics what some predicted about the internet. It will never catch on, they said. What a way to be remembered.
Josie Cox is a freelance journalist and author, an associate instructor at Columbia University, and the author of Women Money Power: The Rise and Fall of Economic Equality.
©2024 Bloomberg L.P.