Women Are Most Exposed To AI And Least Likely To Be Hired To Build It
Occupational segregation determines both who the technology displaces and who gets the roles it creates, and the two findings are rarely read together.
Sign up for my Substack daily AI newsletter here.
See my AI Training course portfolio for corporate Business Leaders here.
Follow me on LinkedIn: https://www.linkedin.com/in/johanosteyn/
Two sets of numbers have been published in recent months about artificial intelligence and women at work. One concerns who is losing ground. The other concerns who is being hired into the roles the technology creates. They are almost always reported separately, which obscures the fact that they describe a single structure viewed from opposite ends.
CONTEXT AND BACKGROUND
The International Labour Organisation published research in March finding that female-dominated occupations are almost twice as likely to be exposed to generative AI as male-dominated ones, at twenty-nine per cent against sixteen. At the highest automation risk, the gap widens sharply, with sixteen per cent of female-dominated occupations in the most exposed categories against three per cent of male-dominated ones. Women are more exposed than men in eighty-eight per cent of countries analysed, and made up only about thirty per cent of the global AI workforce in 2022, four percentage points higher than in 2016.
The hiring data arrived in August. Research from LinkedIn found that women accounted for twenty-six per cent of American hires into roles requiring AI skills, against fifty per cent of hires into everything else. Those roles advertise around one hundred and seventy-seven thousand dollars against roughly eighty thousand for non-AI positions, and postings have doubled since 2023. Across twenty-seven countries, women hold thirteen per cent of leadership roles at AI firms, against twenty per cent of the same roles in the wider technology industry.
INSIGHT AND ANALYSIS
Read together, these are not two problems. The ILO explanation for the exposure figure is occupational segregation. Women are concentrated in clerical, administrative and business support work, where tasks are routine and codifiable. The same segregation explains the hiring figure, because those roles were never the pipeline into technical positions.
The overlap is measurable. Brookings research combining exposure estimates with a measure of adaptive capacity, meaning savings, age, local labour market density and skill transferability, identified six point one million American workers facing both high exposure and low capacity to adjust. About eighty-six per cent of them are women, concentrated in clerical and administrative occupations.
The same study supplies the necessary qualifier. Of thirty-seven point one million workers in the top quartile of AI exposure, twenty-six point five million have above-median adaptive capacity. Most exposed workers are reasonably well placed to move. The concentration of vulnerability is narrow, and it is where the women are.
The practical consequence is that the standard policy response does not work as designed. Reskilling programmes assume displaced workers can move toward the roles being created. The hiring data says that market is admitting women at half the rate it admits them elsewhere, which means the destination is constrained independently of the training.
One caution about the evidence. The hiring research comes from LinkedIn, a Microsoft company analysing its own platform data about an industry its parent is heavily invested in. That does not make it wrong, and the ILO findings are independent and point the same way, but it is not a neutral source.
IMPLICATIONS
For South African organisations, the exposure profile differs from the American one. The ILO puts exposure at forty-one per cent of jobs in high-income countries against eleven per cent in low-income ones. That is not reassurance, because our business process outsourcing sector concentrates precisely the clerical and customer service work identified as most exposed, and it has been one of the more reliable routes into formal employment for women.
For boards, two questions are worth asking. When your organisation creates AI roles, does the hiring ratio for those positions match the ratio for the rest of the business? And of the people who complete your internal AI training, what proportion move into technical or product roles, and does that proportion match your overall gender split? The second is the more revealing, because it tests whether reskilling is producing transitions or only certificates.
CLOSING TAKEAWAY
There is no conspiracy here, which is what makes it difficult. The same structural fact produces both outcomes, and neither was decided by anyone.
Organisations that treat displacement and hiring as separate agenda items will address one and miss the other. The connection is the finding, and it will not appear in any report that examines only half of it.
Author Bio: Johan Steyn is a prominent AI thought leader, speaker, and author with a deep understanding of artificial intelligence’s impact on business and society. He is passionate about ethical AI development and its role in shaping a better future. Find out more about Johan’s work at https://www.aiforbusiness.net
Comments