1 October is International Women in AI Day, a yearly reminder to celebrate the women shaping AI and technology, and to ask how we make room for more of them.
I was checking what models produced long before anyone was talking about generative AI. Around 20 years ago, I worked in an oceanographic research group where scientists built a system to predict ocean currents and temperatures. My job was to validate its outputs and make sure it was working as it should. There was nothing fancy or trendy about it; it was simply part of working with data.
Fast forward a couple of decades and generative AI fever arrived, followed by the constant stream of news, tools, agents, courses and opinions that seemed to appear every five minutes. I found the whole thing overwhelming, and I felt I had fallen behind. It wasn't until early this year that I properly opened that door.
Once I did, I started experimenting. I explored ideas, wrote code and built things with tools such as Lovable and Copilot, including PBIX A11y. I also learnt something I suspect many people discover fairly quickly when they start building with AI: it can produce something that looks completely convincing while being spectacularly wrong.
That old habit of checking the output came in rather handy.
Then I joined Lovable’s She Builds Season 3 hackathon. There were 150 women taking part, all building with AI. For the first time since getting back into this space, I had a real sense of how many women were experimenting, creating and figuring things out for themselves.
And then I looked at the numbers.
Women are already fairly close to parity as AI users. A February 2026 Pew Research Center survey found that 47% of US women and 50% of US men had used an AI chatbot.
The picture changes when you look at who is building the technology. Women made up around 30% of the global AI workforce in 2022, according to the International Labour Organization. On LinkedIn, women represented 29.4% of people listing AI engineering skills in 2025.
Go further towards research and leadership and the numbers become smaller again. Element AI estimated in 2018 that women accounted for around 12% of researchers publishing at leading AI conferences. In the US in 2025, women represented just 20% of new Head of AI hires.
These figures are measuring different parts of the AI ecosystem, so they aren’t directly comparable. Still, there is a pattern worth paying attention to. Women are using AI, while representation becomes thinner across many of the roles involved in building, researching and leading it.
The good news is that this wave of AI offers more ways in than before. You don't have to start as an AI engineer. You can come from data, design, accessibility, research, finance, healthcare or education, bring a deep understanding of a problem, and learn to use the technology to solve it.
This is important because who builds AI shapes what gets built. People bring their experiences and assumptions into their work, and different perspectives change which problems get noticed, which users are considered and which questions get asked during testing.
AI can also reproduce inequalities already in its data. In 2018, it was found out that Amazon had abandoned an experimental recruitment tool after it learnt from years of hiring data in a male-dominated industry and began favouring male candidates. The system had found patterns in historical data and reproduced them in its recommendations.
This serves as a useful reminder that data doesn’t arrive without a history attached to it. If the world represented in the data contains inequalities, an AI system can learn those patterns too. So, questions about where the data came from, whose experiences are represented, what is missing and who might be affected aren’t things to leave until the end of a project.
The discussion about women in AI often jumps straight to recruitment: how do we hire more women? This question matters, but by the time someone applies for a senior AI role, a lot has already been decided. Someone had to encourage them to build the skills, give them access to the technology, and trust them with a project before they had the perfect CV. McKinsey and LeanIn.Org's Women in the Workplace 2025 report found that only 21% of entry-level women say their managers encourage them to use AI tools, compared with 33% of entry-level men. Small differences in encouragement at the start can show up in the workforce years later.
Funding matters too. In the US, startups with at least one female founder raised a record $73.6 billion in venture funding in 2025, according to PitchBook, but more than $30 billion of it came from just two companies, Anthropic and Scale AI, while the number of deals for female-founded startups fell for the fourth year running. In the UK, The Alan Turing Institute found that between 2012 and 2022, all-female founding teams raised just 0.3% of the venture capital invested in AI, compared with 80% for all-male teams.
There isn’t one lever we can pull to change all of this.
It might mean giving someone time to experiment with AI in the job they already have rather than waiting until they have the perfect AI qualification. It might mean looking more carefully at who gets invited onto AI projects, who receives training and who gets the chance to lead. It might mean investors looking beyond the same networks and introductions. And it certainly means thinking about accessibility, privacy, safety and different users while AI products are being designed.
None of those things requires us to know exactly what the future of AI will look like.
Sometimes LinkedIn makes it feel as though AI has already happened. The big companies have their models, everyone has an AI strategy and there seems to be a new agent every time I open my phone. Yet the technology is still being built.
New roles are appearing. New tools are making previously specialist work more accessible. People with deep knowledge of a particular problem are discovering that they can build things themselves, sometimes with a much smaller technical team than would have been needed a few years ago.
I spent a while standing outside this particular door, convinced that everyone else had already gone through it. Then I realised I could just open it.
I wonder how many other people are doing the same thing right now, watching the speed of AI and assuming they have missed their chance.
For International Women in AI Day, that feels like a question worth asking. Who gets encouraged to have a go? Who gets the chance to build? Whose experience gets brought into the room? And what might we build differently if more people had the opportunity to take part?
The future of AI is still being built and there is still plenty of room to decide who gets to help build it.