In this Onrec Q&A, Professor Nick Drydakis discusses new research into the value of AI-related qualifications for women from underrepresented backgrounds, and the fair recruitment practices needed alongside digital skills investment.
1. Your research found that AI qualifications improved interview prospects for women from underrepresented backgrounds, but significant inequalities remained. Were you surprised by the size of the gap, even when candidates demonstrated additional AI skills?
No, I was not particularly surprised by the persistence or magnitude of the inequalities, especially given the challenging economic conditions and cost-of-living pressures affecting the UK labour market. Existing research suggests that during periods of economic uncertainty and recession, discrimination against underrepresented groups can intensify, making it even more difficult for disadvantaged individuals to access employment opportunities. Indeed, in another recent study of mine, using longitudinal data collected in London, I found that worsening economic conditions, particularly rising inflation and interest rates during the cost-of-living crisis, were associated with increased experiences of discrimination among Black individuals. The research also revealed that certain groups, including women, older individuals and sexual minorities, experienced particularly pronounced disadvantages. These findings highlight how economic pressures can exacerbate existing social inequalities.
Against this background, it was perhaps expected that AI-related qualifications, despite their positive signalling value, would not be sufficient to eliminate recruitment inequalities. As my AI Capital theory suggests, AI-related skills can strengthen employability by signalling productivity, digital competence and adaptability. However, employers' recruitment decisions may still be influenced by stereotypes, discriminatory preferences and assumptions about particular demographic groups. What is particularly important is that even with additional AI qualifications, underrepresented women achieved an interview invitation rate of just 14%, compared with 25% for majority-group applicants without AI qualifications. This demonstrates that investing in AI Capital can improve employment prospects, but it cannot independently overcome deeply embedded labour-market inequalities. The findings reinforce the need to combine investment in digital skills with effective anti-discrimination measures, particularly during periods of economic uncertainty when vulnerable groups may face additional barriers.
2. The study found that interview invitation rates increased from 11% to 14% when applicants from underrepresented groups included AI-related qualifications. What does this tell us about how employers are currently valuing AI skills during recruitment?
The increase in interview invitation rates from 11% to 14% suggests that employers are increasingly recognising AI-related skills as valuable indicators of employability, technological adaptability and readiness for workplaces undergoing digital transformation. Although the improvement represents a relatively modest increase of three percentage points, it amounts to an approximately 27% relative improvement in interview prospects. This indicates that AI-related qualifications can provide applicants from underrepresented backgrounds with a meaningful competitive advantage during recruitment.
These findings are particularly relevant to Drydakis’s AI Capital theory, which conceptualises AI-related knowledge, skills and capabilities as valuable forms of capital that individuals can develop and mobilise to improve their labour-market opportunities. According to AI Capital theory, AI competencies have both a productive value, by potentially enhancing workers' efficiency and ability to perform tasks, and a signalling value, by communicating their technological capabilities to prospective employers. The findings provide empirical support for the signalling dimension of AI Capital theory, demonstrating that employers respond positively when applicants present evidence of AI-related training, even before they have had an opportunity to observe their actual workplace performance. Importantly, the results suggest that employers may interpret AI Capital as more than simply the ability to use particular AI tools. An AI-related qualification can also communicate broader attributes, including digital literacy, willingness to learn, adaptability and the capacity to work effectively alongside emerging technologies. As AI becomes embedded in administrative, analytical and operational activities across different occupations, these qualities are increasingly relevant to recruitment decisions. The fact that we observed improvements across non-graduate roles suggests that the labour-market value of AI Capital is not necessarily restricted to highly technical or specialist occupations.
However, the findings also reveal an important limitation in the ability of AI Capital to overcome established inequalities. Despite the improvement associated with AI qualifications, interview invitation rates for underrepresented women remained substantially below the 25% recorded for applicants from non-underrepresented groups without AI qualifications. This suggests that AI Capital can strengthen an applicant's position in the recruitment process without necessarily neutralising the disadvantages associated with demographic characteristics. From a theoretical perspective, employers may value AI Capital while simultaneously making recruitment decisions influenced by stereotypes, established expectations or discriminatory preferences. AI qualifications may help reduce uncertainty about an applicant's capabilities, but they do not necessarily eliminate the influence of these other factors. Moreover, the study cannot establish whether AI qualifications would produce a similar improvement for applicants from majority groups, or precisely which employer perceptions explain the observed increase.
The findings reinforce a central implication of Drydakis's AI Capital theory: investing in AI-related competencies can generate tangible labour-market benefits and potentially support greater economic inclusion. However, the returns to AI Capital are realised within labour markets that continue to exhibit structural inequalities. AI Capital can help underrepresented women compete more effectively, but it cannot, on its own, guarantee equality of opportunity. This is why investment in AI training must be accompanied by fair, transparent and consistent recruitment practices if technological progress is to translate into genuinely inclusive employment opportunities.
3. With employers increasingly looking for candidates who can use AI, do you think there's a risk that AI skills could become another barrier to employment for people who haven't had the same opportunities to access training?
Yes, there is a significant risk that AI skills could become another source of labour-market inequality, particularly if access to AI training remains uneven. This concern relates directly to the emergence of AI digital divides, whereby differences in access to AI technologies, training opportunities and digital resources may translate into unequal employment prospects and earnings.
My earlier research introducing AI Capital theory provides empirical evidence of this risk. In a correspondence experiment conducted in England, I found that university graduates possessing AI Capital, received more interview invitations than otherwise comparable graduates without AI Capital. They were also invited to interview for better-paid positions, with particularly strong advantages observed among large employers. These findings suggest that individuals without AI Capital may face reduced employment opportunities and greater concentration in lower-paid jobs, potentially reinforcing existing economic inequalities.
My latest study extends these concerns to women from underrepresented backgrounds. Although AI qualifications improved their interview prospects, substantial inequalities persisted. The findings suggest that AI Capital is becoming an increasingly valuable labour-market asset, but unequal access to acquiring it could deepen existing disadvantages. Individuals who cannot afford training, lack access to AI technologies or have fewer opportunities to develop digital competencies may find themselves increasingly excluded from attractive employment opportunities. This highlights the importance of addressing AI digital divides through accessible, affordable and inclusive training programmes, particularly for disadvantaged and underrepresented groups. Employers must also ensure that AI qualifications are assessed proportionately to actual job requirements, without creating unnecessary barriers to recruitment. The challenge is to ensure that AI Capital becomes a pathway towards greater social mobility and economic inclusion, while preventing existing labour-market inequalities from being reinforced or deepened.
4. Your research also identified evidence of 'wage sorting', where applicants from underrepresented groups were more likely to receive interview invitations for lower-paid positions. What might explain this, and what should employers be doing to address it?
The evidence of wage sorting is particularly concerning because it suggests that labour-market inequalities are not limited to whether individuals secure employment opportunities, but also extend to the quality and remuneration of the jobs they can access. This may help explain why income inequalities persist between underrepresented and majority groups, even when individuals possess comparable qualifications and skills. Several mechanisms could explain this pattern. Taste-based discrimination suggests that discriminatory preferences may restrict underrepresented applicants' access to better-paid positions, while statistical discrimination theory indicates that employers may rely on stereotypes or group-based assumptions when evaluating candidates' productivity and suitability for higher-paying roles. Consequently, equally qualified individuals may face different opportunities depending on their demographic characteristics. Over time, unequal access to better-paid employment can reinforce earnings disparities, limit career progression and contribute to wider economic inequalities.
These findings also connect directly to my AI Capital theory. My earlier research demonstrated that applicants possessing AI Capital received more interview invitations for higher-paying jobs than otherwise comparable candidates without these skills. The latest study similarly suggests that AI qualifications can improve access to better-paid vacancies for underrepresented women. However, the persistence of wage sorting indicates that AI Capital can improve access to higher-quality employment without fully overcoming discriminatory barriers. Employers should therefore examine not only who receives interview invitations, but also which candidates are considered for better-paid roles. This requires transparent recruitment and salary-setting procedures, consistent evaluation of qualifications, structured assessments and regular monitoring of recruitment outcomes across demographic groups and salary levels. Addressing wage sorting is therefore essential to tackling income inequality and ensuring that underrepresented individuals have equitable access to employment, better-paid jobs and long-term career opportunities.
5. Many employers are moving towards skills-based hiring, placing less emphasis on traditional qualifications and previous job titles. Based on your findings, do you think this approach could help reduce recruitment inequalities, or could existing biases simply carry over into how skills are assessed?
The study was not designed to assess skills-based hiring directly, so we cannot draw definitive conclusions about whether this approach reduces recruitment inequalities. However, an encouraging finding is that employers appear to value AI-related skills, including those acquired through vocational training providers outside traditional university education. This suggests that alternative training pathways can provide meaningful labour-market advantages, particularly as demand for AI competencies continues to grow. From the perspective of my AI Capital theory, these findings suggest that AI-related knowledge, skills and capabilities can serve as valuable signals of employability, productivity and technological adaptability, regardless of the educational pathway through which they are acquired. This is particularly important for individuals who may have limited access to university education but can develop AI Capital through shorter, more accessible training programmes.
Nevertheless, the findings also highlight an important concern: valuing skills does not necessarily mean assessing them without bias. Despite possessing additional AI qualifications, women from underrepresented backgrounds continued to experience substantial disadvantages in interview invitations. This is consistent with the possibility that discriminatory preferences and group-based stereotypes influence recruitment decisions even when applicants demonstrate relevant competencies. Skills-based hiring therefore has the potential to broaden employment opportunities, but its effectiveness in reducing inequalities will depend on how employers implement it. Transparent skills assessments, consistent evaluation criteria and fair recruitment procedures are essential. The key challenge extends beyond recognising AI Capital to ensuring that it is valued fairly and consistently across applicants from different backgrounds.
6. AI is increasingly being used by employers to screen applications and shortlist candidates. Could this technology help make recruitment fairer, or is there a danger that it could reinforce some of the inequalities highlighted in your research?
AI has the potential to make recruitment fairer by introducing more consistent and objective assessments of applicants' qualifications and skills. However, there is also a significant risk that AI-driven recruitment systems could reproduce or even amplify existing labour-market inequalities if they are not carefully designed and monitored. A central concern is algorithmic bias, particularly when AI recruitment systems are trained on historical hiring data that reflect past discriminatory practices. If certain demographic groups have historically been disadvantaged in recruitment, algorithms trained on these patterns may learn to reproduce those inequalities. Therefore, employers must ensure that AI screening systems do not simply replicate historical biases and that the criteria used to evaluate candidates are genuinely fair, relevant to job requirements and regularly tested for discriminatory outcomes.
Transparency is equally important. If employers use AI to screen applications or shortlist candidates, they should clearly disclose which algorithmic criteria influence recruitment decisions, how applicants are assessed and whether historical recruitment data have been used to develop or train these systems. Trade unions should play an active role in scrutinising the transparency, fairness and accountability of AI recruitment tools, including examining screening filters, evaluating potential discriminatory effects and ensuring that workers and job applicants are protected from unfair automated decisions. The research demonstrates that women from underrepresented backgrounds continue to face substantial recruitment inequalities even when they possess additional AI qualifications. Although the study did not investigate algorithmic recruitment directly, its findings highlight why introducing AI into hiring processes must be accompanied by appropriate safeguards. AI should be used to reduce recruitment discrimination and prevent the automation of existing inequalities. Achieving this requires transparent algorithms, carefully assessed training data, independent auditing, meaningful human oversight and active involvement from trade unions to ensure that technological innovation promotes genuinely equal employment opportunities.
7. If you were advising an HR director or recruitment agency looking to make their hiring processes fairer, what practical changes would you recommend they introduce?
My main recommendation would be to ensure that fairness is embedded in everyday recruitment decisions. The research shows that even when women from underrepresented backgrounds possess valuable AI-related qualifications, substantial inequalities persist. This suggests that investing in skills alone is insufficient; employers must also examine how they assess, shortlist and reward applicants. First, I would recommend clear, transparent and standardised recruitment criteria, with structured assessments that focus on skills and qualifications genuinely relevant to each position. Recruiters should receive regular training on unconscious bias and discrimination, while organisations should systematically monitor interview invitation rates across demographic groups. Importantly, they should also examine whether underrepresented applicants are disproportionately shortlisted for lower-paid positions, as the findings on wage sorting suggest. Addressing these patterns is essential for reducing longer-term income inequalities.
Second, where employers use AI-powered recruitment tools, algorithmic screening criteria must be transparent, fair and regularly audited. Particular attention should be paid to the historical data used to train these systems, ensuring that past discriminatory recruitment patterns are not reproduced through automated decisions. Trade unions should have an active role in scrutinising these technologies, including their screening filters, fairness and accountability. Third, drawing on my AI Capital theory, I would encourage employers to recognise AI-related competencies acquired through vocational training and alternative educational pathways, alongside traditional university qualifications. Firms should also support accessible AI training, particularly for underrepresented groups, to help reduce emerging AI digital divides. Finally, equality, diversity and inclusion policies must translate into measurable improvements in recruitment outcomes, going beyond statements of organisational intent. Employers should regularly evaluate whether their practices provide comparable opportunities to equally qualified candidates and take corrective action where disparities emerge. Fair recruitment is both a matter of social justice and an important contributor to organisational efficiency. When employers overlook qualified applicants because of discriminatory barriers, they risk losing valuable talent and productive capabilities. The objective should be to ensure that AI Capital and other relevant skills are recognised and rewarded fairly, regardless of an applicant's background, while improving access to both employment and better-paid career opportunities.
8. Finally, what would you like employers and policymakers to take away from this research, and are there any areas you believe need further investigation?
The central message I would like employers and policymakers to take away is that AI-related skills can improve employment opportunities, but technological upskilling alone cannot eliminate labour-market inequalities. The findings demonstrate that investing in AI Capital can strengthen the recruitment prospects of underrepresented women, yet substantial disadvantages persist in both interview opportunities and access to better-paid jobs. For employers, the priority should be to introduce effective equality, diversity and inclusion policies that translate into measurable improvements in recruitment outcomes. This requires regular HR training on discrimination and unconscious bias, transparent shortlisting criteria, structured assessments and systematic monitoring of recruitment inequalities. Trade unions also have an important role to play, particularly in scrutinising recruitment practices, challenging discriminatory outcomes and ensuring transparency and accountability when AI-powered hiring technologies are introduced.
For policymakers, a key priority should be inclusive digital education and AI training, ensuring that disadvantaged and underrepresented communities have affordable and accessible opportunities to develop AI Capital. My earlier research demonstrated that AI Capital is associated with improved employment prospects and access to higher-paid jobs. Therefore, unequal access to AI training risks creating new digital divides and reinforcing existing income inequalities. Investment in AI skills must consequently be accompanied by stronger anti-discrimination measures and effective enforcement of employment equality standards. Regarding future research, we need to understand more precisely how employers evaluate AI-related qualifications across different demographic groups, whether skills-based hiring can reduce discriminatory recruitment practices, and how AI-powered screening technologies influence employment inequalities. Further longitudinal research is also needed to examine whether acquiring AI Capital improves not only interview opportunities but also actual hiring, wages and long-term career progression, particularly among underrepresented communities. The objective should extend beyond creating a more AI-skilled workforce to fostering a more inclusive labour market in which everyone has a fair opportunity to acquire, demonstrate and benefit from AI Capital. Technological progress should serve as a driver of social mobility and economic inclusion, while helping to reduce existing inequalities and prevent new forms of labour-market disadvantage.
“AI training can help open doors, but it should not be viewed as a substitute for fair recruitment practices.”
Professor Nick Drydakis, Anglia Ruskin University
Key takeaways
- AI-related qualifications can improve interview prospects, but do not on their own remove recruitment inequalities.
- Employers should use transparent, structured assessment and monitor outcomes across demographic groups and salary levels.
- Accessible AI training and robust safeguards around AI recruitment tools are both needed for a fairer labour market.
