Women only accounted for 26% of hires in jobs with AI skills—they could be missing out on a $100K

Women Only Accounted for 26% of Hires in AI-Skilled Jobs. They Could Be Missing Out on $100,000.

TL;DR

Women made up just 26% of hires in roles requiring AI skills, according to LinkedIn Economic Graph research. The median salary for those roles sits at or above $100,000. That is not a pipeline problem — it is a gatekeeping problem, built from structural bias in hiring systems, uneven access to reskilling, and policy frameworks that treat gender equity as a footnote. Whether the gap closes depends on decisions being made right now, in boardrooms and legislatures, not in LinkedIn Learning courses.

Key Takeaways

• LinkedIn Economic Graph reported that women accounted for just 26% of hires in jobs requiring AI skills globally, according to its Future of Work research, a figure that has remained stubbornly persistent across industry sectors.
• The World Economic Forum documented a significant gender gap in AI and machine learning specialist roles in its 2024 Global Gender Gap Report, with women holding fewer than one in three such positions worldwide.
• US Bureau of Labor Statistics data shows that computer and information research scientists — the occupational category most directly tied to AI development work — earn a median annual salary of $145,080, according to BLS Occupational Employment Statistics, placing AI-adjacent roles well above the $100,000 floor the hiring gap denies most women access to.
• The OECD documented that occupations at highest risk of automation are disproportionately held by women, creating a structural double bind: excluded from the high-wage AI jobs being created while overexposed to the lower-wage jobs AI is eliminating, according to its 2023 Employment Outlook.
• McKinsey Global Institute estimated that women will face disproportionately greater pressure than men to switch occupations by 2030 due to automation — concentrated in clerical, service, and care roles — according to its Future of Women at Work analysis.
• UNESCO found that structural discouragement — not lack of aptitude — keeps girls out of AI-adjacent STEM pathways, with teacher bias and curriculum design identified as primary mechanisms, undermining any argument that the talent shortage is organic.
• The AI Now Institute has argued that without explicit gender equity mandates in AI governance frameworks, voluntary corporate diversity commitments will consistently fail to close the hiring gap at scale, as documented in its annual state of AI assessments.

The 26% Is Not an Accident

Start with who runs the hiring systems.

Major technology companies have built AI-powered recruiting tools over the last decade. Those tools were trained on historical hiring data. That data reflected decades of male-dominated tech hiring. The result is a self-reinforcing loop: the AI screens for the past, and the past did not hire women into AI roles.

Amazon famously scrapped an internal AI recruiting tool in 2018 after discovering it systematically downgraded resumes that included the word "women's" — as in women's chess club, women's college. Reuters reported the story in October 2018. Amazon dissolved the team. The bias problem did not dissolve with it.

Other companies continued using similar tools. The mechanism is identical even when the word-matching is more subtle. Models trained on biased data produce biased outputs. Hiring managers interpret those outputs as neutral. Women do not get the callback.

So when LinkedIn's Economic Graph shows that women are only 26% of hires in AI-skilled roles, that is not a reflection of talent distribution. It is a reflection of system design. And the people who designed the systems benefit from not changing them.

That incentive structure matters. Pointing it out is not cynicism. It is the first step toward understanding why the number does not move.

What the $100,000 Gap Actually Represents

Here is the arithmetic.

The US Bureau of Labor Statistics puts median annual pay for computer and information research scientists at $145,080. The median for all US workers sits around $59,000. The gap is real, large, and growing as demand for AI skills outpaces supply.

The $100,000 threshold in this story is not the differential between women's current wages and AI wages. It is the floor. The jobs being created at scale — the ones LinkedIn tracks, the ones employers are hiring for at 74% male — start at $100,000 in most US markets. In major tech hubs, they start considerably higher.

Women who are excluded from those hires are not just missing a salary bracket. They are missing the compound effect: the salary, the title progression, the professional network, the access to the next tier of AI-adjacent roles. Labor economists call this occupational segregation. The downstream cost is not a single year's differential. It is a career trajectory that never bends upward.

AI Is Reshaping Professional Services, and Women Are Being Left on the Wrong Side

Finance and accounting are instructive. These sectors employ large numbers of women in mid-level analytical roles. They are also the sectors most actively restructuring around AI. When an accounting AI startup reaches a $1 billion valuation, the implicit signal to the market is that human accounting work is being repriced downward while AI-adjacent accounting work commands a premium. The question nobody is asking loudly enough: which workers in accounting firms are being positioned for the AI-adjacent roles, and which are being managed out?

The sectors with the highest female employment — healthcare administration, education support, administrative services, retail — are also the sectors where AI adoption is moving fastest at the task-automation level. Scheduling, data entry, customer interaction, basic document analysis. These are not the high-wage AI jobs. These are the jobs AI is replacing. The high-wage AI jobs — the ones building, deploying, and governing these systems — are still 74% male.

That is not coincidence. It is occupational steering, operating at scale, without meaningful accountability.

AI Reskilling Tools: What's Available and Who Gets to Use Them

The standard policy response to labor displacement is reskilling. Let's examine what that actually looks like.

The table shows the problem clearly. None of these platforms was designed to address the barriers that keep women out of AI hiring. They address skill gaps. They do not address hiring bias. A woman who completes a six-month AI specialization and applies for a role that an AI recruiting tool will screen does not have a meaningfully better outcome if the screening model is still trained on biased historical data.

Reskilling is necessary. It is not sufficient. The gap between "we offer courses" and "we close the hiring gap" is the gap between a press release and a policy.

What This Changes for Journalists, Policymakers, and Labor Economists

For Policymakers: The EU AI Act Has an Equity Clause That Is Going Nowhere

The EU AI Act, which entered into force in August 2024, establishes risk categories for AI systems. High-risk systems include AI used in employment decisions — hiring, promotion, task allocation. The regulation requires conformity assessments, human oversight, and transparency documentation.

What it does not require is gender equity auditing. Employers must document their AI hiring tools. They do not have to prove those tools produce equitable outcomes across gender. That is a significant gap in a framework otherwise celebrated for its rigor.

Enforcement falls on national market surveillance authorities that are still building capacity. A meaningful enforcement action against a company for gender-biased AI hiring is at minimum three to five years away under the most optimistic implementation reading. The 26% figure will not improve on its own inside that timeline.

For Journalists: The Story Is in the Audits Nobody Is Running

The technology press covers AI hiring tools regularly. What it covers less often is the audit trail — or its absence. Most companies using AI in hiring do not publish third-party audits of those tools' demographic outcomes. A handful of US jurisdictions (New York City, Illinois, Maryland) have passed or proposed laws requiring bias audits of automated employment decision tools. Federal action has not materialized.

The questions worth asking on this beat: which companies are auditing their AI recruiting tools? What do the results show? Who performed the audit? Who paid for it? Who has access to the methodology? These answers are harder to obtain than the press releases. They are also the actual story.

For Labor Economists: The Measurement Problem Is Real

Standard labor market data tracks employment and wages by occupation and gender. It does not reliably track AI skill requirements within occupations. The 26% figure comes from LinkedIn — a platform with known selection bias toward white-collar, English-speaking, urban workers in formal employment. It is not a random sample of the global labor market.

That does not make the finding wrong. It means we need better infrastructure. National statistical offices — including the BLS and Eurostat — have not built AI skill intensity into standard occupational classifications. Without that data layer, policymakers are making structural decisions using information that is a year or more behind and skewed toward the most visible segment of the affected workforce.

When NOT to Trust the "Reskilling Will Fix It" Narrative

Don't accept reskilling as a gender equity strategy when there is no hiring reform attached. Every major technology company runs a women-in-AI program. Most teach skills. None audit the AI systems those women will apply to after graduation. Training someone for a market that will reject their application is not equity work. It is optically convenient.

Don't treat voluntary corporate DEI commitments as policy equivalents. Meta, Google, and Amazon all made public diversity commitments in 2020 and 2021. By 2023 and 2024, most had conducted significant layoffs that disproportionately affected diversity hires and DEI program staff. The commitments were not contractual. They were not regulated. They evaporated when the economic incentive shifted.

Don't assume that more women in AI courses means more women in AI jobs. If the pipeline were the bottleneck, the 26% figure would improve at the rate AI course enrollment is growing. It is not improving proportionally. The bottleneck is at hiring — in screening tools, in interview panel composition, in the salary negotiation dynamics that begin at the offer stage. None of that is addressed by a Coursera certificate.

Don't let "the data shows improvement" substitute for structural change. Any improvement in the 26% figure that comes from expanding the definition of AI-skilled jobs to include lower-wage roles — content moderation, data labeling, AI-assisted customer service — is not progress. It is reclassification. Watch what happens to the median wage as the denominator expands.

Where This Is Heading

AI hiring tools will get more powerful before they get more regulated. The EU AI Act is the most comprehensive framework in existence, and it does not require equitable outcomes — only transparency documentation. In the US, there is no federal law. State-level requirements are emerging but inconsistent. The tools that perpetuate the 26% figure are being sold, adopted, and refined without meaningful constraint on their demographic effects.

The wage premium for AI skills will widen. Every AI labor market analysis from the past three years points in the same direction: demand for AI-skilled workers is outpacing supply. Employers respond to scarcity with higher wages. If women remain at 26% of hires into that category, the structural wage gap between male-dominated AI work and female-dominated non-AI work will compound over the next decade. The $100,000 floor will become a $150,000 floor, then higher.

Mandatory equity audits for AI hiring tools will come — on a timeline that matters less than it should. New York City's Local Law 144, which requires bias audits of automated employment decision tools, went into effect in 2023. Enforcement has been uneven and the audit methodology is not standardized. But it establishes a direction. The OECD's AI Principles and the EU's regulatory infrastructure both give advocates tools to push for outcome-based requirements. Whether that happens before the current generation of workers is priced out is genuinely uncertain.

The countries that build AI governance with gender equity built in will have a structural labor market advantage. This is not a sentiment. It is economics. A workforce that excludes a majority of its potential AI talent from AI roles is a workforce that is slower to build, deploy, and adapt these systems. The 26% figure is not only a social justice problem. It is an efficiency problem. Policymakers who understand the latter will build faster coalitions.

The data infrastructure problem will define what we can actually know. Without AI skill intensity in standard occupational classifications, every intervention is partially blind. The ILO, BLS, and Eurostat have the technical capacity to build this infrastructure. Whether they prioritize it depends on political will and budget decisions made in the next two to three years. What gets measured gets managed.

FAQ

Is the 26% figure global or US-specific? LinkedIn's Economic Graph covers global hiring activity on its platform, which skews toward white-collar, English-speaking, urban workers in formal-sector employment. The 26% is a global average for LinkedIn-tracked AI-skilled hiring. What it reliably captures is the segment of the AI labor market that matters most for wages and policy: professional-grade, formal-sector AI roles. The actual figure in lower-income countries, where LinkedIn penetration is lower, may differ — but those workers also have less access to the high-wage AI roles in question.

Doesn't the pipeline problem explain the gap? There aren't enough women with AI skills yet. This is the most common deflection, and the evidence does not support it as a complete explanation. Women are enrolling in AI and machine learning courses at rates that should be closing the gap faster than the hiring data reflects. The gap between skill acquisition and hiring outcomes points to something happening at the screening and selection stage — bias in AI recruiting tools, homophily in interview panels, salary negotiation norms. The pipeline argument is also circular: if the reason there are not more women in AI is that the field has not historically welcomed them, "wait for more qualified women to enter the pipeline" is not a solution. It is a delay strategy.

What can individual women do with this information? Negotiate using market data. The BLS and LinkedIn publish salary ranges for AI-skilled roles by metro area and sector. Use them in offers and counteroffers. In jurisdictions where bias audits of automated hiring tools are legally required (New York City, parts of Illinois), ask employers whether their screening tools have been audited — it is a legal question, not an intrusive one. Document and report screening outcomes that seem inconsistent with stated qualifications. These are individual-level actions. They do not close systemic gaps. But they are not useless, and they generate the documented cases that support regulatory enforcement.

Are there sectors where women are better represented in AI roles? Healthcare AI is frequently cited as a relative bright spot, partly because the underlying workforce in health systems skews female and AI adoption in clinical settings has moved through channels where women already hold professional authority. But "better represented" in healthcare AI does not mean equitable. Leadership in health AI development still skews male. Product and engineering decisions are still predominantly made by men. The sector is not an exception to the broader pattern. It is a slower version of it.

What would a serious policy response look like? Mandatory third-party equity audits for AI hiring tools used in regulated industries, with published results and standardized methodology. Outcome-based diversity requirements tied to public procurement — companies that want government contracts must demonstrate that their hiring tools produce equitable outcomes, not merely that they have documented them. Publicly funded reskilling programs that include parallel hiring reform requirements for participating employers. And better national data infrastructure: occupational classifications that capture AI skill intensity, maintained at a cadence that reflects the speed at which the market is actually moving.

Isn't this a problem the market will solve once companies recognize they're leaving talent on the table? The talent argument is theoretically correct and empirically weak. Companies have been hearing the "you're missing out on women's talent" argument for forty years in technology. The representation figures in technology leadership have not improved at anything approaching the rate that argument would predict. Market incentives that have not closed the gender gap in four decades of sustained advocacy are unlikely to close the AI skills gap in four years of quarterly earnings pressure.

Will the EU AI Act fix this? Not in its current form. The Act requires transparency and documentation for high-risk AI in employment contexts. It does not mandate equitable outcomes. National enforcement authorities are still building capacity, and a meaningful enforcement action against a company for gender-biased AI hiring is realistically five years away at current trajectory. The Act creates infrastructure that advocates and regulators can use. It is not, on its own, a solution. The solution requires political will to use the infrastructure it provides.