Women, Power, and the Systems Shaping Work

Picture a product sprint. Six people in a room — or on a call, because it’s always a call now. The agenda: finalizing the decision logic for an AI-powered performance review tool that will be deployed to 40,000 employees across a Fortune 500.

The model was trained on a decade of performance data. The team is debating edge cases. The conversation is fast, confident, technically fluent.

No one in the room asks whose performance history trained the model. No one asks whether the decade of data they just called “ground truth” was collected during a period when women in this company were routinely rated lower on “leadership presence” — a phrase so vague it functioned as a mirror for whoever was doing the rating. No one asks because no one in the room has reason to. They are not malicious. They are just — absent from an experience that would have made the question obvious.

This is not a story about bad actors. It is a story about structural blindness. And structural blindness, at scale, with the velocity of AI deployment behind it, is one of the most consequential forces shaping the future of work.

The Cultural Stakes No One Is Calling by Name

We have spent years debating AI’s technical challenges — hallucination rates, compute costs, latency, alignment. These are real. But the debate that will define the next decade is not happening in machine learning papers. It is happening in the decisions about what AI systems are for, whose workflows they optimize, whose language they speak, whose definitions of merit, risk, and competence they encode.

AI governance is not a compliance function. Model training is not a neutral act. Workflow design is not just engineering. These are acts of cultural authorship. Every dataset carries the fingerprints of the world that generated it. Every product decision reflects a theory about whose experience is default and whose is edge case. Every deployment choice encodes an assumption about who the worker is.

The dataset is not neutral. The model is not neutral. The sprint team that never questioned either — that is where the authorship actually happens.

When we talk about AI shaping work, we are really talking about a small set of people making large, durable decisions about the rules everyone else will live by. The question is not whether those decisions will carry cultural weight. They already do. The question is whether the people making them carry the range of experience — the friction, the skepticism, the pattern recognition — that comes from having navigated systems that were not designed with you in mind.

Women are 38% more likely than men to have ethical reservations about AI and 29% more likely to question its accuracy. This is not timidity. This is Truth. It is the kind of skepticism that catches the performance review model before it ships. It is the kind of friction that, right now, is being systematically kept out of the rooms where these decisions are made.

The Compounding Cost of Absence

Let me be direct about what the data shows, because abstraction is how urgency gets buried.

Only 22% of the global AI workforce is female — WEF 2025. Women hold just 26% of AI-related jobs globally. At the top — where architectural decisions get made and strategy gets set — only 10% of CEO and senior leadership roles in AI organizations are held by women. This is not a diversity metric. This is a description of who is authoring the systems that will determine how work is distributed, evaluated, and compensated for the next generation.

Now consider what those systems are being built on top of. 44% of AI systems studied showed measurable gender bias, according to research cited by the Berkeley Haas Center. Nearly half. These are not fringe academic findings — these are commercial systems in active deployment, making decisions about hiring, credit, healthcare, and performance.

And then consider who bears the cost when those systems go wrong. 57% of the jobs most at risk of automation are currently held by women — the World Economic Forum’s own projection. Women are almost twice as likely as men to predict that more women will be laid off due to AI. They are not being paranoid. They are reading the same systems that were designed without them.

57% of the jobs most at risk of automation are held by women. The people least represented in AI’s design are most exposed to AI’s disruption. That is not coincidence. That is architecture.

This is the compounding cost of absence: you are not just excluded from building the tools. You are excluded while the tools are being built to reshape your labor market. The gap between who designs AI and who is displaced by it is not a pipeline problem. It is a power problem — and it is accelerating.

The business case, to be clear, is just as stark. AI products from gender-diverse teams show 15% fewer bias-related errors, per McKinsey 2024 research. In a field where bias-related errors carry legal, reputational, and regulatory risk, that number should be driving board-level urgency.

The False Comfort of Pipelines

Here is the story the industry tells itself: we are working on it. We are funding STEM programs. We are building mentorship pipelines. Give it time, and the numbers will shift.

I want to challenge this story — not because the investment is wrong, but because it is being used as a sedative. Pipeline thinking assumes that the problem is supply: not enough qualified women moving into the field. Fix the supply, and the system self-corrects. This model is comforting because it requires nothing of the people currently holding power. It defers accountability to a future that will conveniently arrive after the most consequential decisions have already been made.

But power structures do not wait for pipelines. They consolidate while pipelines are being built.

The decisions being made right now — about how large language models are trained, about what data is treated as authoritative, about which industries get automated first, about how AI governance frameworks are written — these decisions have decade-long shelf lives. A woman who enters an AI policy role in 2031 after a pipeline program delivers her will inherit a regulatory architecture, a set of entrenched vendor relationships, and a corpus of encoded assumptions that were all authored during the years when women were told to wait their turn.

Pipelines are a mechanism for eventual inclusion. They are not a mechanism for present-tense influence. And in a technology cycle moving at the current velocity, eventual is functionally the same as never.

What Presence at the Table Actually Means

I want to be precise about this, because the word “inclusion” has been so hollowed out that it now functions as cover for exactly the dynamic it claims to address.

A seat at the table is not the same as authority at the table. Being invited into the room is not the same as having decision rights in the room. What presence at the table actually means, in the context of AI governance and deployment, is this: domain authority — where women are not advisors on gender considerations but architects of system design. Decision rights — where women hold formal veto and approval authority, not informal influence that can be overridden. Cultural accountability — where the organization is structurally answerable for the outcomes its AI systems produce on women and other marginalized groups, not just for its hiring optics.

Real presence means women are in the sprint, not reviewing the sprint’s output. Women are setting the training data criteria, not filing the post-deployment bias report. Women are writing the product strategy, not producing the case study about why it should have been written differently.

What the Decade Looks Like from Here

Two futures. Both extrapolations of current trajectories.

In one, the absence continues. AI systems designed by a workforce that is less than a quarter female continue to encode the asymmetries of the world that generated their training data. The jobs most at risk — disproportionately held by women — are the first to go, while the roles shaping the next wave of automation remain largely male. The gender wealth gap, already structural, becomes technologically reinforced — written into the model weights, the product architectures, the policy frameworks. In twenty years, we will write reports about this. They will be very well-cited.

In the other future, the organizations that grasp this earliest gain something their competitors cannot replicate quickly: AI systems that are more accurate, more trusted, more durable — because they were built by people who asked harder questions at the start. The companies that put women in rooms where actual power lives — not advisory councils, not ERG forums, not diversity slide decks — find that the friction those women generate is the friction that keeps systems from failing at scale.

The cultural decisions being made right now about AI are not footnotes. They are the text. They will determine not just what work looks like, but who gets to do it, who gets evaluated by it, who gets displaced by it, and who gets to define what counts as fair.

The question is whether women will be in that room — with authority, with accountability, with the unambiguous mandate to shape what gets built.

Not as a pipeline promise. Right now.

Miri Rodriguez is the CEO and Cofounder of Empressa AI, where she works at the intersection of AI strategy, brand narrative, and women’s economic empowerment.