The Role of Artificial Intelligence in Recruitment and Selection

AI in Recruitment: Support, not Substitution

According to Jolandie Overbury, a registered Industrial Psychologist and Member of the SIOPSA Interest Group, People Assessment in Industry (PAI), AI is currently used in a deliberately limited capacity. Its role is to support analytical and administrative processes - for example, generating draft interpretive assessment reports - while final interpretation and decision making remain firmly in human hands. Every output is reviewed, contextualised and validated by a qualified professional before it informs any hiring decision.

A similar pattern is evident at Capitec. Joanna Maingard, Assessment Lead, notes that AI tools are being introduced incrementally, with early adoption focused on productivity and workflow optimisation rather than candidate evaluation. This phased approach reflects an organisational preference to test value and manage risk before extending AI into more complex or consequential domains.

This cautious sequencing is echoed in academic research. A 2025 systematic review of AI in recruitment found that organisations derive the greatest benefit when AI is first deployed for low risk, efficiency driven tasks such as résumé screening, scheduling and administrative support, while human judgment is retained for interpretive and evaluative functions. In practice, the most defensible recruitment models are those where AI augments expertise rather than attempts to replace it (Dadaboyev, et al, 2025).

Where Confidence Gives Way to Concern: Psychometric Scoring

The debate becomes sharper when AI moves from support into scoring and interpretation.

This observation captures one of the most persistent concerns surrounding AI enabled psychometric assessment: whether these tools can meet the foundational standards required for fair and defensible selection (Sunday Times, 2025).

According to Overbury, the use of AI in candidate scoring is currently a “no go.” Her concern is not theoretical. Psychometric tools used in selection must demonstrate validity, reliability and freedom from bias and many AI based assessment products cannot yet do so convincingly (Dukanovic & Krpan,2025).

Recent empirical research supports this scepticism. A 2025 study published in Frontiers in Psychology compared AI driven personality inference tools with traditional psychometric assessments and found that while some AI tools showed partial construct validity, they consistently failed to demonstrate predictive validity for real world job outcomes. In selection contexts, predictive validity is non negotiable. Without it, the defensibility of AI based scoring tools is fundamentally undermined (Dukanovic & Krpan,2025).

Compounding this challenge, a comprehensive review of fairness in AI driven recruitment highlights a growing disconnect between rapid commercial adoption and the slower pace of academic validation. Many AI assessment tools are deployed without sufficient evidence of reliability, job relevance or robust bias mitigation, exposing organisations to ethical, legal and reputational risk (Rigotti, & Fosch Villaronga,2024).                                                                           

AI powered Video Interviews: The Rise & Risk

Few applications of AI in recruitment generate as much debate as AI powered video interviews.

According to Roxanne Swanepoel, Group Head of Talent at Mr Price Group, these tools are appealing in theory. They promise speed, standardisation and scalability, particularly in high volume recruitment. In practice, however, they raise significant concerns.

AI based video interviews analyse verbal and non verbal cues such as facial expressions, tone of voice and body language to support early stage screening. While this may appear objective, Swanepoel cautions that such systems often struggle to interpret behaviour accurately and sensitively.

Overbury highlights a critical contextual risk: behaviours such as avoiding direct eye contact, commonly interpreted by AI systems as low confidence, may be culturally appropriate or respectful in South African contexts. When AI systems are trained on datasets that do not adequately reflect this diversity, candidates may be unfairly disadvantaged.

Research into algorithmic bias reinforces this concern. The landmark Gender Shades study demonstrated that commercial facial analysis systems produced error rates of up to 34.7% for darker skinned women, compared to less than 1% for lighter skinned men. Although the study focused on facial recognition, its implications extend directly to AI driven hiring tools that rely on similar computer vision technologies (Buolamwini & Gebru, 2018).

Cost, Access and Practical Barriers

Beyond questions of fairness and validity, practical constraints also shape AI adoption.

According to Swanepoel, AI enabled assessment platforms require significant investment - not only in software, but in infrastructure, integration and ongoing vendor support. In a cost constrained market, organisations are understandably hesitant to commit resources to technologies that have not yet demonstrated clear, defensible value. AI may be advancing rapidly, but uptake remains shaped by organisational readiness, risk tolerance and the ability to justify return on investment, particularly when hiring decisions carry long term consequences for both individuals and institutions (Dadaboyev et al, 2025).

Conclusion: Efficiency is not Authority

AI will continue to advance and to reshape how organisations recruit and select talent. But insights from South African practitioners draw a clear line: efficiency does not equal authority. Until AI‑based tools can meet the psychometric standards required for fair, reliable and defensible selection, their role should remain advisory, not decisive. 

In hiring - where decisions shape careers, cultures and lives - judgment remains a human responsibility.

References:

• Dadaboyev, S. M. U., et al. (2025). Role of artificial intelligence in employee recruitment: Systematic review and future research directions. Springer.
• Patel, A. & Mahomed, N. (2025). “The People Puzzle, Minds Over Machines: AI in the Workplace”, Business Day / Sunday Times.
• Patel, A. and Mahomed, N., ‘The People Puzzle, Minds Over Machines: AI in the Workplace’, The Sunday Times, Business Day, 2025.
• Dukanovic, D., & Krpan, D. (2025). Comparing chatbots to psychometric tests in hiring. Frontiers in Psychology.
• Buolamwini, J., & Gebru, T. (2018). Gender Shades: Intersectional accuracy disparities in commercial gender classification. Proceedings of the 1st Conference on Fairness, Accountability and Transparency.
• Mujtaba, D. F., & Mahapatra, N. R. (2025). Behind the screens: Uncovering bias in AI driven video interview assessments.
• Jolandie Overbury, Industrial Psychologist, SIOPSA Interest Group.
• Roxanne Swanepoel, Group Head of Talent, Mr Price.
• Joanna Maingard, Assessment Lead, Capitec.