Transforming Human Resource Management with Artificial Intelligence in Recruitment, Performance, and Retention

Abstract

This paper examines how artificial intelligence (AI) is reshaping human resource management (HRM) across recruitment, performance evaluation, and retention. Motivated by the trade-offs between hiring speed and quality and the high costs of employee turnover, this study reviews the latest evidence on AI’s impact. We trace the evolution of AI use in HR (screening, matching, interviews, onboarding, performance tracking, attrition prediction) and frame it through theories such as resource-based view, human capital, signaling, and socio-technical systems. A conceptual framework maps an end-to-end AI-enabled talent pipeline and links AI to outcomes (efficiency, hire quality, diversity, retention) while considering risks (fairness, privacy, explainability) and governance (human oversight). Using a systematic narrative review (2015-2025) and case examples (e.g. Unilever’s AI hiring, IBM’s attrition model), we synthesize results on time-to-hire, candidate funnels, quality-of-hire, and retention metrics. Key findings include evidence of faster hires and cost savings (Unilever: 90% time reduction, £1M savings) but also limitations in demonstrating improved long-term retention. Public sentiment is wary: many Americans oppose AI making final hiring decisions, though some see AI as more consistent. We analyze regulatory and ethical issues: “high-risk” classification under the EU AI Act (2024) imposes duties on data quality, bias audits, transparency; new U.S. rules (NYC LL144) require bias impact ratios; and live litigation (Workday class action) signals compliance risks (Reuters., 2024). An implementation guide recommends target use-cases, data/model pipelines, oversight processes, and ROI study designs aligned with NIST and ISO standards. We discuss trade-offs (speed vs fairness, predictiveness vs explainability), strategic implications for HR leaders, and identify research gaps in longitudinal outcomes, causal methods, and fairness optimization.