When the bias is built in: what AI's gender gap means for HR

Two questions came to mind

1. If the people building AI are overwhelmingly men, does the bias become embedded in the systems themselves?
2. And what does this mean for HR, a profession that is made up of predominately women?

How the bias ends up in the code

Representation in AI development affects two things: who advances within the field, and the outputs the technology produces once it's built.

A literature review published in Humanities and Social Sciences Communications found that algorithmic bias in AI-enabled recruitment produces discriminatory hiring outcomes based on gender, race and personality traits. The review traced this bias to limited or skewed training data and to the biases of the people designing the algorithms.

In practice, the bias compounds. A large-scale study using roughly 361,000 fictitious resumes to test five leading large language models found that AI hiring tools systematically favoured female candidates while disadvantaging Black male applicants, even when qualifications were identical. The researchers noted that most models are developed and trained on US data, so they reflect US social categories and labour market structures. That raises real questions for how well they translate to other markets, including Australia.

Biased AI also shapes human judgement. An experiment simulating hiring decisions found that participants followed AI recommendations 70 per cent of the time when candidates' qualifications were comparable, yet only 8 of 294 participants detected the gender bias built into those recommendations. Exposure to biased AI shifted participants' own independent judgement to align with that bias afterwards, unless they'd been shown an explanation of how the AI reached its decision.

There is a workable fix. A 2025 study on hiring algorithms found that awareness of debiasing significantly increased the willingness of qualified women to apply for competitive roles they'd otherwise have avoided, and that this happened without reducing the total pool of qualified applicants. Debiasing improves the outcome for employers as well as candidates. It requires someone in the room to notice the bias and act on it, which brings the problem back to representation.

HR is female-dominated, and that leaves it exposed

HR is one of the more heavily female-dominated professions, and the research suggests this leaves it more exposed to AI disruption.

The International Labour Organization's 2026 research brief found that female-dominated occupations are almost twice as likely to be exposed to generative AI as male-dominated ones, with around 29 per cent of female-dominated occupations exposed compared to 16 per cent of male-dominated ones. Among the highest-risk occupations, the gap widens further: 16 per cent of female-dominated occupations fall into the highest exposure category, against 3 per cent of male-dominated ones. The ILO attributes this to three factors: women being overrepresented in roles most susceptible to automation, underrepresented in AI and STEM roles, and exposed to AI systems that reflect embedded gender bias.

HR is named directly in this research. IBM's CEO, Arvind Krishna, has said the company is slowing hiring for roles that can be easily replaced by AI in back-office functions, including HR. A separate Australian analysis of Jobs and Skills Australia data found that human resources and payroll clerks are among the top 20 occupations most exposed to automation.

There's a structural shift under way alongside the exposure risk. Academic research on digital automation in HR argues that automation is likely to reshape the female HR workforce into what the authors call "screen-level HR": a female-dominated workforce operating in increasingly masculine-coded ways. The gender balance of who does the job may hold. What counts as valued, strategic work within it is what changes.

Adoption data bears this out. Sapient Insights research found women in HR are almost twice as likely as their male counterparts to adopt AI and automation into their roles. At the same time, women's presence in top HR leadership positions decreases as company size and visibility increase. Women are doing the bulk of the adapting, without a corresponding gain in seniority.

Where the two threads meet

These two threads describe a single feedback loop. AI is built predominantly by men, so it carries their blind spots into deployment. It's then rolled out fastest into functions like HR, which are staffed predominantly by women. Those women adopt it readily and do the work of integrating it into daily practice, while remaining less likely to be the ones deciding how it's designed or governed.

McNamara's article closed on Lobo-Pulo's question: what employers, institutions and workforce systems need to do differently. Women need to be in the rooms where these tools are built and tested, not only where they're deployed, to ensure both output quality and equity.