Women Are 26% Of AI Hires. That's A Design Flaw, Not A Shortage

Eight years ago, I helped LinkedIn produce its first data on women in artificial intelligence and wrote the story that came out of it: There was a massive gender skills gap in a field about to explode. Women made up just 22% of AI professionals globally.

Very few people cared. The story got little engagement and almost no pickup.

Fast forward to today, and the research coming out of the world’s largest professional network is even more startling: That gap has resulted in women accounting for just 26% of U.S. AI hires last year, compared with 50% of hires into non-AI occupations. When you look at the C-suite globally, only 13% of AI leadership roles are held by women.

The time I spent working with engineers at a tech company taught me that when you discover a design flaw in code—when the code is doing exactly what it was built to do, but results in worse outcomes—the only fix is to go back to the decision that shaped it. Unlike a bug, it can’t be fixed with some tweaks to a line here or there.

We have eight years of evidence that the gender gap in AI is by design, something no one chose, but everyone shipped. That evidence produced conference panels I’ve sat on, pledges I’ve signed and reporting I’ve written.

What it hasn’t changed is any of the decision making that led us to this point in the first place.

Why Knowing Changed Nothing

Over a decade of reporting on gender bias in the workplace has proved to me that measuring something only gets you so far. Without having someone accountable for an outcome, nothing changes, whether it’s the pay gap or the broken rung.

For a brief period after 2020, it looked like companies had accepted that, with 68% of S&P 500 companies disclosing diversity hiring metrics in executive pay, according to The Conference Board. It was the closest we got to real accountability, even though most of those metrics rewarded effort rather than results.

Now that hope has all but disappeared. Just 35% of companies made the same disclosure in 2025 and among early 2026 proxy filers reviewed by Pearl Meyer, less than 10% carried a diversity goal in their incentive plan. Another 36% kept the goal, but renamed it “talent strategy” or “human capital,” blurring the problem entirely.

Companies have stopped linking consequences to the gender hiring gap at the same time that AI job postings have roughly doubled. The workforce that will build the technology of the decade is being hired right now, and almost no one’s compensation is tied to what that looks like.

The same technology is putting at risk the jobs women already hold. Brookings found in January that of the 6.1 million U.S. workers with the highest AI exposure and the lowest ability to adapt, more than eight in 10 are women.

We Keep Shipping The Problem

My 2018 story came with a warning even I stopped thinking about. Experts told me at the time that having mostly male teams build technology that would soon be run mostly by machines posed serious ethical challenges. Tech built almost solely from the perspective of one gender could ultimately disregard or discount the needs and values of another.

Eight years later, more than two dozen Meta employees are in federal court making a version of that argument. Their complaint—filed last month in Oakland over roughly 8,000 separations—alleges that selection metrics such as keystroke tracking and AI token-usage dashboards wrongly led to the termination of a disproportionate number of employees on medical or family leave. Meta said the allegations “lack merit and are not based on facts,” and that workforce decisions “were and are made by people, not AI.”

In other words, the well-researched “motherhood penalty” that women face in the workplace may be having an impact at an unprecedented scale. A hiring manager who has an unconscious bias against one of their direct reports because she just got back from maternity leave affects the career of one person. But a system with that bias encoded into its decision making can make the same call thousands of times, turning a manager’s bias into a company default.

And with women accounting for only 13% of the people in the room making those decisions, the chances of someone catching that bias before it ships come down to whether anyone in the room has lived it.

What Leaders Can Do

The Trump Administration’s stance on corporate DEI efforts has had a measurable chilling effect on what executives are willing to do to right this wrong. Still, there are measures that leaders who truly care about this can take today:

Measure the machine, not just the people.

Six in 10 companies have not assessed AI’s impact on fairness in hiring and promotions, according to data from Lean In and McKinsey.

Without properly auditing how the technology is both displacing female talent and stalling their careers, the numbers will continue to move in the wrong direction.

Clearly communicate how AI shows up in performance reviews.

Only one in 10 companies has clear policies guiding AI’s role in performance reviews and just 5% of managers are trained on ethical use. The Meta complaint is likely the first of many if more companies don’t put structures in place to prevent it.

Stop punishing women before they even start.

Lean In also has data that when women attempt to close the gender skills gap in AI themselves, they aren’t rewarded and sometimes are even punished. At the entry level, 21% of women are encouraged by a manager to use AI versus 33% of men. Early-career women are also more likely to fear being seen as cheating when they use AI at all at work.

Link fair AI hiring to executive pay.

Encouragingly enough, it’s clear that while most U.S. companies have stopped disclosing their diversity efforts publicly, many are still doing the work privately. That means there is still an opening where AI hiring can be linked to executive pay, an accountability metric a CEO can defend as talent strategy rather than a political action.

Eight years from now, someone will refer to the 2026 numbers and write a version of this article. My only hope is that they—unlike me—get to explain how it got better.

Click here to explore this article's original source for more.