Alan Turing Institute Workforce Studies

In the AI Story

The research addressed a question that the AI industry had largely avoided: as the AI workforce was forming, what demographic patterns were being established, and what consequences would those patterns have for the technology being built? Previous technology revolutions had shown that early workforce composition tends to harden into long-term patterns — the early professionalization of computing, for instance, shifted programming from a feminized clerical task to a masculinized engineering discipline, with effects that persisted for generations. Wajcman's studies asked whether AI was being formed on similar lines.

The answer was unambiguous. AI work was disproportionately male across every measured dimension, and the disparities were not narrowing. Women were concentrated in lower-status AI roles — data labeling, model testing, project coordination — while higher-status architecture, strategy, and leadership roles remained disproportionately male. Women's educational qualifications in AI-relevant fields matched or exceeded men's, but their workforce outcomes consistently underperformed what their credentials would predict.

The venture capital findings were particularly stark. Between 2012 and 2022, eighty percent of UK AI venture capital went to all-male founding teams; all-female teams received 0.3 percent; female-founded startups that did secure funding received six times less capital per deal than comparable male-founded startups. These findings grounded Wajcman's concept of male-lens investing in empirical data that proved difficult for industry defenders to contest.

The confidence findings added a subtler layer. Even when controlling for qualifications and experience, women in AI roles reported lower confidence in their technical capabilities than their male peers — a pattern consistent with broader research on the gendered psychological effects of working in male-dominated environments. The finding suggested that the workforce disparities were reproducing themselves through mechanisms internal to individual women's self-assessment, making structural intervention more difficult.

The studies' methodological contribution was to insist on the intersection between workforce composition and technology output. Where previous research had treated workforce diversity as an equity concern separable from the technology itself, Wajcman's framework demanded that the two be analyzed together: the AI being built by a disproportionately male workforce encoded the temporal assumptions, problem framings, and workflow expectations of that workforce, and the resulting technologies would reshape the social conditions that had produced them. The analysis was circular by design, because the phenomena were circular in reality.

Origin

The research was conducted at the Alan Turing Institute, where Wajcman served as a Visiting Fellow. The institute, founded in 2015, is the UK's national institute for data science and AI, and its workforce research program was positioned to shape UK policy on AI development.

The reports built on Wajcman's earlier work at the London School of Economics and drew on collaborations with researchers across the UK university system. The findings informed policy recommendations that were partially incorporated into the UK government's approach to AI workforce development.

Key Ideas

The workforce is forming on gendered lines. Early-stage AI work is establishing demographic patterns that research on previous technology revolutions suggests will persist for decades.

Disparities span every measured dimension. Jobs, qualifications, seniority, funding, and confidence all show consistent gendered patterns that reinforce each other.

Funding is the sharpest disparity. The 80/0.3 distribution of UK AI venture capital is not explained by founder quality or opportunity differences but by systematic bias in evaluation.

Workforce and technology are inseparable. The demographic composition of AI development shapes the technology being built, which reshapes the social conditions that produced the composition.

Intervention must address structure, not attitudes. The disparities persist through institutional mechanisms — pattern-matching in hiring, intuitive evaluation in funding, cultural norms in workplace composition — that individual attitude adjustment cannot reach.