AI-enabled HR systems in Bangladeshi firms correlate with better employee outcomes—higher satisfaction, engagement, performance and wellbeing, and lower turnover intent—because they boost perceived P–O fit and empowerment; these benefits grow with technology trust and are weakened by privacy concerns.

AI-enabled HRM systems in business organizations: effects on workforce outcomes and the roles of trust and privacy

Hriday Chandra Shil, S. K. Md. Anik Hassan Rabby, Md Mostafizur Rahman, Nashita Mumtahina, Md Anamul Islam, Faria Zafreen, Asraful Islam, Md Al Fassi · September 10, 2026 · Frontiers in Artificial Intelligence

openalex correlational medium evidence 7/10 relevance Full text usable extracted full text DOI Source PDF

Structured author observations

Linked only from stored provider relations; the raw author line above is never matched by name.

OpenAlex

Latest observation: September 13, 2026

Hriday Chandra Shil provider ID
S. K. Md. Anik Hassan Rabby provider ID
Md Mostafizur Rahman provider ID
Nashita Mumtahina provider ID
Md Anamul Islam provider ID
Faria Zafreen provider ID
Asraful Islam provider ID
Md Al Fassi provider ID

In a two-wave survey of 487 employees across 61 Bangladeshi firms, AI-enabled HRM practices were associated with higher job satisfaction, engagement, performance and wellbeing and lower turnover intention, relationships partially mediated by perceived person–organization fit and psychological empowerment and moderated positively by technology trust and negatively by privacy concerns.

Citation observations

Cumulative provider counts captured on specific dates; providers are never combined.

0 cumulative citations

OpenAlex · Observed September 13, 2026

View corpus context

0 cumulative citations

Semantic Scholar · Observed September 13, 2026

View corpus context

Introduction The rapid expansion of artificial intelligence (AI) in human resource management has substantially reshaped how organizations attract, recruit, develop, evaluate, and retain their workforce. Research examining the combined employee-level effects of AI-enabled digital HRM (AI-DHRM)—through psychological mediating processes and under technology-related boundary conditions—remains sparse, particularly in emerging economies. Drawing on Social Exchange Theory, the Job Demands-Resources model, and UTAUT2, this study conceptualizes AI-DHRM as a reflective higher-order construct comprising four functionally distinct sub-dimensions, specifies a dual mediation model integrating an affective-relational pathway (person-organization [P-O] fit perception) and a motivational-agentic pathway (psychological empowerment), and identifies technology trust and privacy concern as critical boundary conditions. Methods A time-lagged, two-wave survey was administered to full-time employees of 61 AI-HRM-adopting organizations in Bangladesh, yielding 487 matched responses. AI-DHRM practices, mediators, moderators, and controls were measured at Time 1; all five outcomes were measured five weeks later at Time 2. Data were analyzed in IBM AMOS 26.0 using a two-stage structural equation modeling approach, with bias-corrected bootstrapped mediation ( n = 5,000 resamples), the index of moderated mediation, and Johnson-Neyman analysis. Results AI-DHRM practices exerted significant positive effects on job satisfaction (β = 0.41, 95% CI [0.29, 0.53]), work engagement (β = 0.38, 95% CI [0.26, 0.50]), job performance (β = 0.35, 95% CI [0.25, 0.45]), and employee wellbeing (β = 0.33, 95% CI [0.21, 0.45]), and significantly reduced turnover intention (β = −0.29, 95% CI [−0.41, −0.17]). Both P-O fit perception and psychological empowerment partially mediated these relationships across all five outcomes. Technology trust strengthened, and privacy concern attenuated—but did not reverse—the AI-DHRM-mediator pathways. Moderated mediation was confirmed across all ten conditional indirect effects. Discussion The findings establish AI-DHRM as an integrated system that employees experience as a coherent organizational investment, transmitted through two complementary psychological channels and conditional on trust and privacy perceptions. Organizations should design AI-HRM as a coherent bundle and treat trust-building and privacy-by-design as prerequisites to deployment. Findings rest on self-report data from a single country and warrant replication.

Summary

Main Finding

AI-enabled digital HRM (AI-DHRM), modeled as a unified higher‑order system of four AI-enabled HR functions, yields clear employee-level benefits: higher job satisfaction, greater work engagement, better job performance, improved employee wellbeing, and lower turnover intentions. These effects operate partly through two psychological channels—person–organization (P–O) fit perceptions (affective‑relational pathway) and psychological empowerment (motivational‑agentic pathway)—and are strengthened by employees’ trust in the technology and weakened (but not reversed) by privacy concerns.

Key Points

Conceptualization
- AI-DHRM is treated as a reflective higher‑order construct with four sub‑dimensions:
- AI-driven recruitment & selection (AIDRS; loading λ = 0.78)
- AI-enabled learning & development (AILD; λ = 0.82)
- AI-powered performance management (AIPM; λ = 0.79)
- AI-assisted compensation management (AICM; λ = 0.74)
- Theoretical grounding: Social Exchange Theory (SET), Job Demands‑Resources (JD‑R) model, and UTAUT2 (technology acceptance/trust & privacy).
Main estimated effects (time‑lagged design; standardized betas and 95% CIs)
- Job satisfaction: β = 0.41 (95% CI [0.29, 0.53])
- Work engagement: β = 0.38 (95% CI [0.26, 0.50])
- Job performance: β = 0.35 (95% CI [0.25, 0.45])
- Employee wellbeing: β = 0.33 (95% CI [0.21, 0.45])
- Turnover intention: β = −0.29 (95% CI [−0.41, −0.17])
Mediation
- Both P–O fit perception and psychological empowerment partially mediated AI-DHRM’s effects across all five outcomes (bias‑corrected bootstrapped mediation).
Moderation & conditional indirect effects
- Technology trust amplifies AI-DHRM → mediator paths.
- Privacy concern attenuates those paths (attenuation does not fully negate positive effects).
- Moderated mediation confirmed for all tested conditional indirect effects (two mediators × five outcomes).
Practical interpretation
- Employees experience AI‑HRM as a coherent organizational investment. Benefits are contingent on psychological framing (fit, empowerment) and technology perceptions (trust/privacy).
Caveats noted by authors
- Self‑report measures, single‑country (Bangladesh) sample, and a two‑wave design (5‑week lag) limit causal claims and generalizability.

Data & Methods

Sample & design
- Time‑lagged two‑wave survey of full‑time employees in 61 organizations in Bangladesh.
- 487 matched respondent pairs (Time 1 and Time 2, five weeks apart).
- Time 1: AI‑DHRM practices, mediators (P–O fit, psychological empowerment), moderators (technology trust, privacy concern), controls.
- Time 2: five outcome measures (job satisfaction, engagement, performance, wellbeing, turnover intention).
Measures
- AI‑DHRM operationalized as a higher‑order latent factor with four reﬂective subscales (AIDRS, AILD, AIPM, AICM).
- Well‑established scales for mediators and outcomes (paper provides measurement details).
Analysis
- Two‑stage structural equation modeling in IBM AMOS 26.0.
- Bias‑corrected bootstrapped mediation (n = 5,000 resamples).
- Index of moderated mediation and Johnson–Neyman analysis to probe conditional indirect effects.
Robustness
- Authors report model fit and supplementary tests including individual sub‑dimension analyses (noted that the higher‑order specification is justified both conceptually and empirically).

Implications for AI Economics

Labor productivity and performance gains
- Positive standardized effect on job performance (β = 0.35) implies meaningful employee‑level productivity improvements associated with AI‑HRM adoption, though translating β into firm‑level output requires linking to objective productivity metrics and monetary valuations.
Retention and turnover costs
- Negative effect on turnover intention (β = −0.29) suggests cost savings from reduced voluntary separations; economists can incorporate expected reductions in recruiting/training costs when modeling ROI of AI‑HRM investments.
Value of trust and privacy investments
- Technology trust materially amplifies benefits; privacy concern attenuates them. From an economic standpoint, investments in trust‑building (transparent algorithms, explainability, governance) and privacy‑by‑design likely raise net returns to AI‑HRM by increasing realized employee benefits and reducing friction costs.
Design as a system and complementarities
- Treating AI‑HRM as a coherent bundle (recruitment, L&D, performance, compensation) implies complementarities across HR functions. Economic models should account for bundled adoption effects and potential increasing returns when multiple AI‑HRM modules are deployed together.
Policy and regulation considerations
- Privacy concerns reduce realized gains; stricter data protection and algorithmic accountability regimes may increase compliance costs but could raise employee trust and hence economic benefits—cost–benefit analysis should include these indirect pathways.
Distributional and labor market effects
- AI‑HRM may alter wage setting (more transparent, data‑driven compensation) and affect within‑firm equity; economists should study distributional impacts (who gains—tenure groups, skill levels) and potential effects on labor demand/composition.
Research and measurement recommendations for economists
- Use objective performance and retention data to quantify monetary returns.
- Conduct cost‑effectiveness analyses comparing AI‑HRM investment costs (technology, governance, privacy protections) with labor productivity and turnover savings.
- Model equilibrium labor market impacts in emerging economies where AI‑HRM diffusion may change hiring, training, and match quality dynamics.
- Investigate heterogeneity: how effects vary by industry, firm size, worker skill, and country institutional context.
Practical takeaway for firms and policymakers
- Deploy AI‑HRM as an integrated system and prioritize trust‑building and privacy protections to maximize economic returns; without these, potential benefits are attenuated even if not fully negated.

Limitations to keep in mind when applying these findings: results are based on self‑report data in a single emerging‑economy context with a relatively short time lag, so economists should seek replication using objective outcomes, longer horizons, and cross‑country samples before generalizing magnitudes for policy or firm‑level investment decisions.

Assessment

Paper Typecorrelational Evidence Strengthmedium — The study uses a relatively large sample (487 matched responses) and a time-lagged design with SEM, mediation bootstrapping, and moderated mediation tests which strengthen causal inference compared with cross-sectional work; however, all measures are self-reported, organizations were not randomly sampled, and unobserved confounding and common-method bias remain plausible, limiting causal conclusiveness. Methods Rigormedium — Appropriate and modern statistical techniques were used (higher-order construct modeling, two-stage SEM, bias-corrected bootstrapping, moderated mediation, Johnson–Neyman analyses), and temporal separation of measures mitigates some simultaneity concerns; nonetheless reliance on single-source self-report data, potential non-random sampling, limited information on control variables and robustness checks (e.g., reverse causality tests, objective performance measures) reduce rigor. SampleTime-lagged two-wave survey of 487 matched full-time employees working in 61 organizations in Bangladesh that have adopted AI-enabled HRM; Time 1 measured AI-DHRM (higher-order construct comprising recruitment, learning & development, performance management, compensation management), mediators (P–O fit, psychological empowerment), moderators (technology trust, privacy concern), and controls; Time 2 (five weeks later) measured five employee outcomes (job satisfaction, work engagement, job performance, wellbeing, turnover intention). Data are self-reported. Themeshuman_ai_collab org_design adoption IdentificationTime-lagged, two-wave observational survey: predictors (AI-DHRM, mediators, moderators, controls) measured at Time 1 and outcomes measured five weeks later at Time 2; analysis via two-stage structural equation modeling with bias-corrected bootstrapped mediation (5,000 resamples), index of moderated mediation and Johnson–Neyman probes. No experimental assignment or exogenous instrument; causal claims rely on temporal ordering, theoretical priors, and statistical mediation/moderation. GeneralizabilitySingle-country study conducted in Bangladesh (an emerging economy), limiting external validity to other cultural, regulatory, and economic contexts., Non-random sample of organizations/employees; industry composition and organizational size/AI maturity not fully described, which constrains applicability across firm types., All measures are self-reported and short (five-week) lag may not capture longer-term effects., Findings pertain to employee perceptions/outcomes rather than objective productivity or firm-level economic metrics.

Claims (7)

Claim	Direction	Outcome	Confidence & Evidence	Details
AI-DHRM practices significantly increased employee job satisfaction. Worker Satisfaction	positive	Employee job satisfaction	Reading fidelity high Study strength medium	n=487 β = 0.41, 95% CI [0.29, 0.53] 0.3
AI-DHRM practices significantly increased employees' work engagement. Worker Satisfaction	positive	Work engagement	Reading fidelity high Study strength medium	n=487 β = 0.38, 95% CI [0.26, 0.50] 0.3
AI-DHRM practices significantly increased employee job performance. Output Quality	positive	Employee job performance	Reading fidelity high Study strength medium	n=487 β = 0.35, 95% CI [0.25, 0.45] 0.3
AI-DHRM practices significantly improved employee wellbeing. Worker Satisfaction	positive	Employee wellbeing	Reading fidelity high Study strength medium	n=487 β = 0.33, 95% CI [0.21, 0.45] 0.3
AI-DHRM practices significantly reduced employees' turnover intention. Turnover	negative	Employee turnover intention	Reading fidelity high Study strength medium	n=487 β = −0.29, 95% CI [−0.41, −0.17] 0.3
Person-organization fit perception and psychological empowerment partially mediated the relationships between AI-DHRM practices and all five employee outcomes. Organizational Efficiency	positive	Five employee outcomes: job satisfaction, work engagement, job performance, employee wellbeing, and turnover intention	Reading fidelity high Study strength medium	n=487 0.3
Technology trust strengthened, while privacy concern attenuated but did not reverse, the AI-DHRM-to-mediator pathways. Worker Satisfaction	mixed	Strength of AI-DHRM relationships with person-organization fit perception and psychological empowerment, conditional on technology trust and privacy concern	Reading fidelity high Study strength medium	n=487 0.3