Women are more exposed – but use AI less
Saco’s report Mitt i skiftet shows that 65 per cent of women in the Swedish labour market work in occupations with high exposure to AI. The corresponding figure for men is just under 53 per cent. The same pattern can be seen internationally. ILO data shows that 29 per cent of female-dominated occupations globally are exposed to generative AI, compared with 16 per cent of male-dominated occupations.
However, exposure does not automatically mean that a job is at risk of disappearing. Saco’s analysis also shows that a larger proportion of women’s exposure to AI is found in occupations where the technology is expected to complement people rather than replace them. There are therefore significant differences between occupational groups. Healthcare and social care, for example, face a very different situation from administrative roles such as accounting assistants, office assistants and secretaries. This makes the issue more complex than any single percentage can convey.
When we look instead at the use of AI, the picture becomes clearer. Harvard Business School has analysed 76 sources covering more than 100 countries and over 300,000 people. The findings show a global adoption rate of 47.8 per cent among men, compared with 39.3 per cent among women. The gap has narrowed since generative AI entered the mainstream, but it remains.
Why does the AI gender gap exist?
The researchers behind the Harvard study identify five types of friction that affect adoption: knowledge, perceived usefulness, institutional support, social legitimacy and trust. Several of these barriers can be reduced relatively quickly. Practical training, access to the right tools and clear support from employers can make a significant difference.
The question of social legitimacy is particularly interesting. Studies show that women are more likely to worry that using AI will be perceived as cheating or as a sign of lower competence. Moreover, that concern appears to have some basis in reality. Research suggests that perceptions of a person’s competence can differ depending on whether it is a woman or a man who says they used AI to complete a task.
This also changes the question we need to ask. Rather than simply asking why women experiment with AI less, organisations need to consider whether different employees have the same conditions and confidence to experiment in the first place. Curiosity and experimentation do not emerge in a vacuum; they are shaped by workplace culture, expectations and how mistakes are received.
Experimenting with AI is not the same as changing how you work
Discussions about AI often conflate three different stages: experimenting, adopting and transforming. Trying an AI tool once does not mean it has become part of everyday work. Regular use, in turn, does not automatically mean that the technology has changed how we solve problems, make decisions or organise our work. These are three distinct stages, each with its own barriers.
This presents an important challenge for organisations. The people who have experimented most with AI are often also those who speak most confidently about the technology, setting the benchmark for what it means to be ‘at the forefront’. For those who have not yet progressed as far, the distance can quickly feel greater than it actually is. One person described it to me like this:
‘It feels as though everyone has got so much further than I have that there’s hardly any point in me trying. And the questions I have are the ones everyone else was asking six months ago.’
That is precisely the kind of barrier organisations need to address.
AI strategy also needs to address culture
A successful AI strategy cannot focus solely on technology, licences and training. It also needs to create an environment where people feel able to experiment, ask basic questions and share things that did not work.
This is particularly important because the greatest value is often created when AI is applied directly to people’s everyday work. The employees doing the work are often best placed to identify which tasks can be improved, automated or approached in entirely new ways. If some groups face greater barriers to participating in that process, the organisation risks not only creating a skills gap but also missing out on some of the business value its AI investments could generate.
