• As AI transforms workplaces, the technology has an outsized impact on women’s jobs, according to new data from the United Nations’ International Labour Organization and Poland’s National Research Institute. To help future-proof their careers, women can use AI to augment their jobs, but are less likely to engage with the technology than their male counterparts, according to Harvard Business School professor Rembrand Koning.

As workers grapple with anxiety around artificial intelligence replacing them, women in the workplace may have extra reason to fear. Jobs traditionally held by women are much more exposed to AI than those traditionally held by men, according to new data from the United Nations’ International Labour Organization (ILO) and Poland’s National Research Institute (NASK).

In higher income countries, jobs with the highest risk of AI automation make up about 9.6% of women’s jobs, compared to 3.5% of jobs among men, the report released Tuesday found. More broadly, 25% of global jobs are potentially exposed to generative AI, a percentage that increases to 34% among higher income countries.

The report notes clerical and administrative jobs have the highest exposure to AI, which could be one reason why AI poses an outsized risk to women workers. Between 93% and 97% of secretary and administrative assistant positions in the U.S. were held by women between 2000 and 2019, according to the U.S. Census Bureau. Comparatively, women made up between 40% and 44% of the workforce in the same 20-year period. Secretaries and administrators are the fifth most common professions for women in the U.S., according to the Department of Labor.

Notably, the study does not mention caretaker jobs such as health aides that require emotional labor and are more likely to be held by women; they are considered more AI-proof.

Automation versus augmentation

While AI has shown potential to gobble up jobs like software engineers and computer programmers, the technology may also threaten entry-level positions across white-collar industries beyond administrative roles. A Bloomberg report in April found AI could replace more than half the tasks performed by market research analysts and two-thirds of tasks done by sales representatives. The technology could perform only 9% and 21% of the respective tasks of those positions’ managers.

The ILO-NASK report isn’t meant to say that AI will eliminate clerical or entry-level jobs. Rather, these jobs still require human involvement in some capacity, and identifying jobs that AI can partially complete can help prepare the workforce in those industries for technological changes.

“This index helps identify where GenAI is likely to have the biggest impact, so countries can better prepare and protect workers,” Marek Troszyński, senior expert at NASK, said in the report.

Rembrand Koning, associate professor of business administration at Harvard Business School, believes one key to women future-proofing workplace roles that may be more exposed to AI is to follow the framework of viewing AI as a tool, not a threat.

“This goes back to the distinction between automation versus augmentation when we think about AI,” Koning told Fortune. “We can think of this as a threat, which is that it’s going to automate away a lot of these clerical jobs that might be held more by women. On the other hand, we can think of AI as automating a lot of this work, of allowing [workers] to take on tasks that might be higher paying, or that there might be more competition.”

Gender differences in AI use

While Koning sees a path forward for workers to use AI to their benefit, he also sees a gender barrier: Women are using AI tools at an average 25% lower rate than men, his research found.

There’s not one clear reason for this disparity, Koning said, but one explanation outlined in a working paper co-authored by Koning is women are more concerned about the ethics of AI. Some fear they will be judged as cheating for using the technology or that leaning on AI tools will cause male colleagues to question their intelligence.

“Men seem to be much more confident—shall I say, overconfident—that, if they use AI, they’ll still get all the benefits,” Koning said.

The onus of changing who feels comfortable accessing AI falls not on the women workers, but on leaders in the workplace, Koning said. In many workplaces, workers, usually men, experiment with AI tools in the shadows. Even if an office doesn’t have a license for or partnership with an AI company, its management should still set clear expectations and resources on how to use the technology, Koning suggested.

“If we want to make sure it’s inclusive, it includes all workers, it’s the job of a leader to bring everybody in,” he said.