Not because someone told you to. Because this one’s different. | Women’s Day edition | 8 min read
Nobody handed me a seat at the AI table.
I came to it sideways, through years of marketing work, through curiosity, through a vague but persistent feeling that something enormous was happening and I didn’t want it to happen without me.
I didn’t have a computer science background. I didn’t have a technical co-founder or a startup or a plan. I just started building. Experimenting. Paying attention.
And the more I built, the more one question kept surfacing: where are the other women?
Not because this is a women’s issue. Because this is a power issue. And power, historically, goes to the people who show up early.
We’ve seen this before. It didn’t go well.
Every major technological shift in modern history has created enormous wealth and opportunity. The internet boom. The mobile revolution. The rise of platforms. Crypto.
And every single time, women were underrepresented at the table where the decisions were made, and overrepresented among the people those decisions affected.
AI is no different.
Right now, women hold between 22% and 30% of global AI jobs. In data science, that drops to 12%. In cloud computing, 15%. In the UK, women account for just 14% of STEM roles.
At the leadership level, the people actually deciding how these systems work, what they prioritize, whose needs they serve, women hold just 10% of CEO and top tech positions across AI organizations.
That’s not just an equity problem. That’s a design problem. When the people building a technology don’t reflect the full range of people who will use it, the technology reflects their blind spots. Hiring tools that discriminate.
Medical diagnostics that underperform on women’s health data. Voice assistants that default to female-coded subservience. These aren’t accidents. They’re the output of rooms that lacked the right people.
Curious about the numbers?
Checkout the interactive version to take a closer look: Women & AI: The Numbers
I built a small interactive tool to go alongside this piece. 12 statistics, all connected, click any bubble to see the full story behind the number and follow the threads to related data points. Some of these stats will surprise you. The one in the middle surprised me the most.
The risk nobody is talking about loudly enough
The International Labour Organization analyzed 436 specific occupations globally to understand real AI exposure. And here’s what they found:
29% of female-dominated roles, clerical support, translation, administrative work, face high exposure to AI displacement. For male-dominated roles, that number is 16%. Almost half. And zooming out further: 57% of all jobs currently at risk of automation are held by women.
This isn’t because AI is targeting women. It’s because generative AI is entering a labor market that has been gender-segregated for decades.
Female-dominated professions heavily rely on text processing, scheduling, and routine communication, the exact capabilities that large language models are engineered to replicate at scale.
Meanwhile, male-dominated fields index higher in physical trades and hands-on work. A language model cannot fix a broken pipe.
The shockwave is hitting office-based economies first. In high-income countries, 41% of total employment is exposed to this disruption. If you are mapping your career for the next decade, this is not a distant concern.
The algorithm judging you right now
Modern fintech platforms have moved beyond traditional credit scoring, your repayment history, your debt-to-income ratio, to what researchers call learning algorithms.
These systems analyze alternative data: your mobile phone usage patterns, your social media connections, even which games you play in augmented reality. All to determine your financial credibility.
The problem is that this alternative data doesn’t exist in a vacuum. It reflects the same structural inequalities that traditional scoring was supposed to move past. Women in precarious employment, or those who took career breaks for caregiving, get flagged as high risk, not because they’re bad borrowers, but because their life patterns don’t match the model’s idea of financial reliability.
They get approved. But with APRs ranging from 300% to 700%. It’s not credit access. It’s a debt trap engineered by machine learning. And it’s completely invisible to the end user.
We have a tendency to trust machines as impartial. The computer says so, so it must be objective. But these models were built by humans, trained on human data, and they inherit every bias baked into that history. The gap is just hidden in the code now.
The confidence gap is real. And it’s costing us.
A Paris School of Economics study gave university students access to ChatGPT and watched how they used it:
Male students were 34% more likely to get the correct answer from the tool. Not because they were smarter, when you looked at the top-performing female students, they matched their male counterparts exactly. The gap wasn’t ability. It was persistence and confidence. When ChatGPT didn’t give the right answer, 71% of male students tried again. Only 55% of female students did.
When asked how confident they felt about their prompts, over 40% of male students said very or extremely confident. Only 18% of female students said the same.
I recognize that feeling. The hesitation before asking the dumb question. The assumption that someone else probably knows better. The sense that the tool was built for a different kind of person.
But here’s what the data also shows: the gap closes completely at the top. The most capable women perform just as well as the most capable men. This isn’t a ceiling. It’s a confidence tax. And AI is just the latest place it shows up.
But here’s what’s actually happening
If the story so far sounds heavy, this is where it turns.
In the past year, the proportion of US women adopting generative AI has tripled. That outpaces men, who grew at 2.2 times. The gender gap in daily AI use is closing, fast.
In Europe, 77% of female founders are actively using AI to scale their businesses, contributing to 5.76 billion euros in funding in 2024 alone, a third of which went into deep tech. Not basic consumer apps. Advanced AI, quantum computing, synthetic biology. Women building the fundamental infrastructure of the future.
In India, women-led AI startups grew 300% between 2020 and 2024, from around 520 to an estimated 2,100 companies. They’re building in health tech, ed tech, e-commerce, solving deeply entrenched problems that the mainstream tech industry has historically ignored.
And at a recent 72-hour vibe coding build-a-thon, women with no computer science background turned raw ideas into functioning software. Not prototypes. Not mockups. Working products.
Because when you remove the gatekeeping of traditional coding, what’s left is just: can you think clearly about what you want to build?
Turns out, a lot of women can.
What getting involved actually looks like
This is the part I want every woman who thinks “AI isn’t for me” to read slowly.
You do not need a computer science degree. You do not need to know how to code. What AI actually needs, what it is genuinely bad at and desperately requires humans for, is empathy, judgment, communication, creativity, and the ability to understand how real people actually live. The skills that have been historically undervalued in tech.
The skills that a lot of women have been quietly developing for their entire careers.
If we want AI systems that are less biased, that serve the full range of human experience, we don’t just need more computer science PhDs. We need anthropologists, psychologists, designers, storytellers, and people who understand what it actually means to live in a body in the world.
The Council of Europe’s new AI treaty, the first legally binding international framework on AI, is built on exactly this argument.
Getting involved looks like using AI to move faster in the job you already have. It looks like building something small just to see if you can. It looks like automating the mental load, the family logistics, the scheduling, the endless coordination, so there’s more space for the work that actually matters.
It looks like learning to prompt well, which is just another way of saying: learning to ask clearly for what you need.
None of that requires permission.
This one’s different
The internet required infrastructure. Mobile required hardware. Crypto required a particular kind of risk tolerance and tribal belonging. AI requires curiosity and willingness to try.
The barrier has never been lower. The upside has never been higher. And the window, where getting in early actually means something, is open right now.
The data is clear that diverse teams build better AI, less biased, more creative, more attuned to the full range of human experience. Which means women being at the table isn’t just good for women. It’s good for the technology. It’s good for everyone who’s going to live with whatever gets built.
And here’s the question I keep sitting with, the one that came up at the end of a research deep dive I did recently: if we eventually build AI that is perfectly equitable and endlessly patient, will we start preferring the empathy of our machines over the messy reality of each other?
The better these systems get at understanding us, the more it might change how we connect with each other as humans.
That question doesn’t have an answer yet. But the people who will shape it are the ones building right now.
I’m one woman who came to this sideways, who started building before she felt ready, who found the table and pulled up a chair without waiting for an invitation.
There’s room. There’s always been room.
The question is whether we’re going to take it.
Happy International Women’s Day.
Top comments (0)