The Boom That Forgot Half the Workforce: Women Hold Just 26% of New AI Jobs

The numbers arrive with the industry’s boom cycle: AI job postings have doubled since 2023, and roles involving AI now pay, on average, more than twice the salary of comparable non-AI work. By LinkedIn’s count, this is the fastest-growing, highest-paying job category in the global economy. And yet the pipeline delivering people into it is remarkably lopsided — women filled just 26% of new AI hires in the most recent year measured, compared with roughly 50% of hires into non-AI occupations. In executive seats, the share collapses to 13%.

Those figures, published in a report by LinkedIn and revisited in depth by The Guardian on October 4, 2026, describe a two-sided squeeze. Women are underrepresented in the sector’s upside — the engineering, research, and product roles that carry the salary premium — and simultaneously overrepresented in the roles most exposed to AI-driven disruption, such as customer service and clerical work. One trend compounds the other: the same technology that generates the premium also erodes the occupational floor beneath it.

The Numbers Behind the Gap

The core dataset comes from LinkedIn’s global labor research. Its headline findings:

• 26% — women’s share of new AI-role hires in the past year, versus ~50% for non-AI roles
• 13% — women’s share of AI executive positions, versus ~19% executive representation in non-AI sectors
• 2x — the average salary premium for AI roles over non-AI roles
• 2x since 2023 — the growth in AI job postings
• $45,000 — the median pay gap between men and women across AI occupations, driven largely by role concentration rather than unequal pay for equal work

Sarah Steinberg, LinkedIn’s head of global public policy partnerships, put it bluntly: “AI is creating some of the fastest growing, highest paying and most consequential jobs in the global economy. Women are just strikingly underrepresented.”

The concentration problem is structural. Women who do work in AI are disproportionately found in lower-paying functions — data annotation, content review, support operations — rather than the modeling, infrastructure, and platform engineering roles where the premium concentrates. The $45,000 median gap is less a story about unequal paychecks for identical jobs and more about who gets sorted into which jobs at the hiring gate.

Why the Field’s Freshness Hasn’t Fixed It

A common assumption holds that because AI is new, it should be more meritocratic: no one has a decade of experience with a model shipped last year, so everyone starts level. Practitioners inside the industry say the opposite has happened. In a sector hiring at breakneck speed, firms have defaulted to the fastest filtering mechanisms available — personal networks, referrals, and reputational signals that long predate the technology.

“Companies are hiring at a breakneck speed, but they’re finding people through the same networks, the same referrals, the same filters that they’ve always used,” said Brenda Darden Wilkerson, president of AnitaB.org. “Generally, that’s not given women the same sort of exposure they should have.”

The pace itself is a filter. Jayeeta Putatunda, an AI engineering lead at investment firm Turing, described returning from four months of maternity leave to find “completely different frameworks and levels of models.” Catching up was possible only with supportive colleagues and a partner who split childcare equitably. “That is one way women get pushed out of AI,” she said. “They don’t lack the interest or ability to keep up. They may lack the infrastructure that makes keeping up possible.”

Founder experiences point in the same direction. Urvashi Batra, co-founder and CEO of Prioriwise, reports that investors take her less seriously than her male co-founder — so much so that the pair have learned pitches land more reliably when he delivers them. “A lot of people say: ‘Oh, AI is lowering barriers, it’s equalizing the playing field for women,’” she told The Guardian. “I actually think it’s the opposite.”

The DEI Retreat Arrives at the Worst Moment

The data lands amid a deliberate dismantling of the institutional machinery that once counteracted these dynamics. Since 2023, major US employers — Accenture, Deloitte, IBM, PayPal among them — have paid multimillion-dollar settlements under a Department of Justice campaign arguing that DEI programs violate anti-discrimination law. The programs that survived have been renamed, narrowed, or quietly shelved.

Felicia Newhouse, founder of AI Powered Women, says the political climate now follows advocates through the door: “It is getting harder to go into companies being called AI Powered Women. And our advocates inside those companies are also struggling with having women-focused initiatives.”

The result is a sector scaling faster than any in recent memory, with fewer institutional counterweights than at any point since the diversity-program era began.

What Is Actually at Stake

Advocates frame the stakes in generational terms. If AI roles carry a 2x pay premium and women fill a quarter of them, the arithmetic produces what several sources called the widest gender pay gap in generations — not through discriminatory pay for identical work, but through unequal distribution of an entire job class. Felicia Newhouse’s formulation is the sharpest: “We’re talking about who captures a major new source of economic mobility. The deepest risk is that a participation gap becomes a power gap.”

The power dimension is not abstract. The people building AI systems decide what problems get solved, whose use cases get prioritized, and how models behave at the edges. A workforce that is 74% male at the entry gate and 87% male in the executive suite is a workforce whose blind spots ship to production.

The Counterargument — and Why the Data Blunts It

Skeptics of intervention argue the gap is transitional: the AI workforce draws from computer science and ML graduate cohorts that were themselves male-skewed years ago, and as pipeline demographics shift, so will hiring. The pipeline story is real but slow, and the exposure data makes waiting expensive. Every year the gap persists, automation pressure grows on the female-dominated roles at the bottom while the premium compounds at the top for everyone else. Waiting for the pipeline assumes the two curves cross before the divergence locks in.

The second counterargument — that meritocratic sorting explains the numbers — struggles with the network-hiring evidence. When hiring runs on referrals and reputation rather than demonstrated skill, “meritocracy” mostly measures proximity to the existing workforce.

Where This Goes Next

LinkedIn’s data has tracked this gap across successive reports; the trend line, not any single year, will tell the story. The indicators to watch are whether women’s share of AI hires rises above the high-20s, whether the 13% executive figure moves as the sector’s first generation matures into leadership, and whether the annotation-and-support concentration thins as those functions themselves automate.

The uncomfortable synthesis is that AI has not invented workplace inequality — it has industrialized it. A technology praised for lowering barriers to capability is running on hiring machinery that reproduces the exclusions of the industries it is disrupting. The “permanent underclass” that Silicon Valley discusses as an AI-age hypothetical is, for a significant share of working women, already the working present.