AI promises to make HR leaner and more personalized—automating hiring, standardizing reviews and tailoring development—but the evidence base lacks robust field evaluations and flags serious privacy and ethical risks, underscoring the need for regulation and more empirical research.

Main Finding

This systematic literature review (2020–2024; PRISMA-guided) finds that AI has strong potential to transform core HRM functions—particularly recruitment, performance evaluation, and talent management—by raising efficiency, objectivity, and personalization. However, the empirical evidence is thin: many claimed benefits lack robust, real-world validation and are counterbalanced by serious risks (algorithmic bias, privacy/monitoring concerns, digital divide) and adoption frictions. Responsible, regulated, hybrid human–AI approaches and further empirical research are needed to realize net social and economic gains.

Key Points

• Scope and method
  • Systematic review of literature published 2020–2024 using PRISMA; searched Google Scholar, Scopus, ScienceDirect.
• Main functional effects
  • Recruitment: AI streamlines sourcing, screening, and outreach; can reduce candidate pools dramatically (one study reports ~75% reduction), lower time-to-hire and processing costs, and identify passive candidates via online behaviour analysis.
  • Performance evaluation: AI enables real-time performance monitoring and more standardized/objective metrics; supports individualized development plans.
  • Talent management: AI tools can personalize training, forecast skill gaps, and support retention strategies.
• Risks and limitations highlighted
  • Algorithmic bias: training on historical data can reproduce workforce demographics and discrimination (e.g., Amazon case cited).
  • Loss of soft-skill and novelty detection: AI excels at hard skills and pattern matching but can miss unconventional/high-potential candidates and nuanced qualities.
  • Privacy and surveillance: continuous monitoring raises ethical and legal concerns about employee privacy and workplace trust.
  • Digital divide and adoption barriers: uneven access and skills across geographies/worker groups impede equitable benefits.
  • Limited empirical evidence: many studies are conceptual, pilot-scale, or descriptive; there is a lack of large-scale causal estimates of economic impacts.
• Governance & mitigation
  • Emerging solutions: explainable AI (XAI), algorithmic audits, fairness toolkits.
  • Regulatory context: GDPR and the EU AI Act are shaping adoption, particularly in regulated industries; authors call for stronger, more universal governance frameworks.
• Theoretical framing
  • Technology Adoption: TOE (Technology-Organizational-Environmental) and TAM (perceived usefulness/ease of use) explain adoption drivers/constraints.
  • HRM impact: AMO (Ability-Motivation-Opportunity) frames how AI should complement ability development, motivation, and opportunities to boost performance.

Data & Methods

• Review design: Systematic literature review following PRISMA guidelines.
• Inclusion window: Studies published 2020–2024.
• Data sources: Google Scholar, Scopus, ScienceDirect (primary academic databases reported).
• Types of evidence in corpus: conceptual papers, descriptive analyses, case studies, pilot evaluations; relatively few large-scale or causal empirical studies.
• Methodological limitations reported by authors:
  • Heterogeneous quality and methodologies across reviewed papers.
  • Scarcity of longitudinal and experimental evidence to assess long-term effects.
  • Potential publication bias toward positive or prescriptive accounts of AI benefits.

Implications for AI Economics

• Productivity and vacancy-cost effects
  • Short-term firm-level productivity gains likely via faster candidate matching, lower time-to-hire, and reduced administrative HR costs. These reduce vacancy costs and may raise output per worker if matched quality improves.
• Labor market frictions & search
  • AI can reduce search frictions (better targeting of passive candidates), altering job matching dynamics and potentially raising match efficiency. However, if algorithms entrench existing hiring patterns, they may exacerbate mismatches for nonstandard candidates.
• Wage and inequality dynamics
  • Complementarity with human capital: returns to AI adoption will depend on worker skills and complementarities (AMO). Firms that invest in re-skilling and integration may capture gains; workers lacking digital skills risk displacement or downward pressure on wages.
  • Potential for increased polarization: demand for algorithm-savvy HR professionals and data-literate roles may rise, while routine HR tasks decline.
• Market structure and vendor dynamics
  • Concentration risk: dependence on third‑party HR AI vendors creates switching costs and potential market power, affecting pricing and innovation diffusion.
• Externalities and regulatory costs
  • Algorithmic bias and privacy harms create negative social externalities that are not internalized without regulation—leading to potential litigation, remediation costs, and trust losses that can reduce adoption benefits.
  • Compliance with GDPR/EU AI Act and algorithmic audit requirements will impose compliance costs but may raise social welfare by mitigating harms.
• Policy and measurement priorities for economists
  • Need for causal evidence: randomized trials, field experiments, and quasi-experimental designs (e.g., difference-in-differences, instrumental variables) to measure impacts on hiring outcomes, wages, turnover, productivity, and equity.
  • Key metrics to track: time-to-hire, vacancy durations, recruitment costs per hire, match quality (retention, performance), wage trajectories of hires, algorithmic disparity measures across demographic groups, and firm productivity measures.
  • Policy levers: support digital equity (infrastructure and training subsidies), mandate transparency/audits for HR algorithms, encourage hybrid human–AI workflows, and monitor market concentration in HR AI platforms.
• Practical guidance for economists studying AI in HRM
  • Treat AI adoption as a bundle of technologies and complementarities (training, process change); model heterogeneity across firm size, sector regulation intensity, and worker skill composition.
  • Quantify both private returns to firms and social distributional effects to inform policy balancing innovation and equity.

Suggested next research steps (concise) - Conduct large-scale, multi‑firm experiments or natural experiments measuring causal effects of specific HR AI tools. - Estimate distributional impacts across worker skill groups and regions to assess inequality consequences. - Study vendor market structure and its interaction with firm adoption and bargaining power.