AI in HR speeds hiring and trims costs in Indonesia, but only where firms are digitally mature and large; benefits flow partly through stronger HR analytics and shrink for complex roles.

Main Finding

AI adoption in HRM significantly improves both recruitment efficiency (faster, lower-cost hiring) and selection efficiency (higher throughput and selection quality). These benefits are partly mediated by firms’ HR analytics capability and are larger in organizations with higher digital maturity and greater size, but diminish for roles with high job complexity.

Key Points

• Primary effects
  • AI adoption → positive, significant effects on recruitment efficiency and selection efficiency (correlations: r = 0.37 and r = 0.32, p < 0.01).
  • AI adoption → HR analytics capability (r = 0.48, p < 0.01).
• Mediation
  • HR analytics capability partially mediates the AI → recruitment and AI → selection relationships (AI generates data/automation; analytics capability extracts predictive, operational value).
• Moderation
  • Digital maturity amplifies AI’s effect on recruitment efficiency.
  • Firm size amplifies AI’s effect on selection efficiency (scale/scope economies).
  • Job complexity weakens AI’s effect on selection efficiency (complex roles require deeper human judgement).
• Governance and context
  • Authors note regulatory pressures (e.g., EU AI Act classifying recruitment systems as high-risk), variation in organizational readiness, and emergence of authenticity/assessment-validity issues (AI-generated applications).
• Reliability & validity
  • Constructs show solid psychometric properties (Cronbach’s α 0.80–0.91; CR 0.86–0.93; AVE ≥ 0.55).
  • Multi-source design (survey + operational HR metrics) and CMB checks used to reduce bias.

Data & Methods

• Sample
  • n = 200 recruitment/selection units across Indonesian firms (purposive sampling with sector and size quotas).
• Measures
  • AI adoption: intensity of tools (resume screening, chatbots, automated scheduling, video analytics, candidate matching, offer optimization).
  • HR analytics capability: data quality, analytics skills, governance, tools, integration into decision-making.
  • Moderators: organizational digital maturity, job complexity, firm size (log employees / categories).
  • Outcomes (objective HR metrics): time-to-fill, time-to-hire, cost-per-hire, sourcing yield, selection ratio, assessment throughput, offer acceptance, early attrition (≤ 6 months). Outcomes were standardized so higher = better efficiency.
• Design & analysis
  • Multi-source (perceptual survey + objective HR records) to mitigate common-method bias.
  • SEM (PLS-SEM and CB-SEM fit checks) with bootstrapping (5,000 resamples) to test direct, mediating, and moderating effects.
  • Robustness checks, attention checks, missingness <5%, Harman/marker-variable/full-collinearity VIFs used to diagnose CMB.
• Key descriptive stats
  • AI adoption mean = 4.92 (SD 0.86); HR analytics capability mean = 4.75 (SD 0.90); digital maturity mean = 4.88 (SD 0.82); job complexity mean = 4.10 (SD 0.95).
  • Recruitment and selection efficiency standardized (mean = 0, SD ≈ 0.9).

Implications for AI Economics

• Complementarity and capability investments
  • AI capital is complementary to HR analytics capability; returns on AI investments are substantially higher when firms build analytics skills, data quality, governance, and integration—implying complementarity between analytics human capital and AI capital.
• Scale economies and heterogeneity of returns
  • Larger firms realize stronger selection-efficiency gains (throughput and fixed-cost dilution). Expect heterogenous ROI across firm size—important for models of diffusion, firm-level adoption thresholds, and aggregate productivity effects.
• Diminishing returns in complex tasks
  • For high-complexity roles, marginal efficiency gains from AI are smaller. Economic models should account for task complexity as a friction limiting automation-driven productivity growth in more cognitively demanding occupations.
• Policy and compliance costs
  • Regulatory regimes (e.g., EU AI Act) impose compliance, governance, and transparency costs that will affect net benefits and adoption timing. Policy-induced compliance costs may differentially impact smaller firms and change adoption equilibria.
• Labor-market dynamics
  • Faster hiring and higher throughput can reduce vacancy durations and matching frictions, potentially raising labor market tightness and reallocating recruiter effort toward high-value, human-centric tasks. However, increased throughput may also amplify downstream screening for authenticity and verification, generating new evaluation tasks.
• Measurement and evaluation
  • Using operational HR metrics (time-to-fill, cost-per-hire, early attrition) provides tangible productivity measures for cost–benefit analyses of AI—useful inputs for macro and micro economic models of AI adoption impact.
• Research and policy directions
  • Future economic research should quantify welfare implications (net of compliance costs), examine distributional effects across firm sizes and occupations, and estimate long-run productivity spillovers from scale adoption versus localized experimentation.

Limitations to note for interpretation - Cross-sectional survey design limits causal inference despite multi-source measures. - Indonesia-focused sample limits generalizability; regulatory and market contexts differ internationally. - Paper reports significance and correlations; full path coefficients and effect sizes should be consulted in the article for precise economic magnitudes.

If you want, I can extract the reported path coefficients and R² values (if provided later in the paper) and produce a short table of estimated effect sizes and implied elasticities for use in economic modeling.