Generative AI as a catalyst for HRM practices: mediating effects of trust

The paper examines how generative AI tools (e.g., ChatGPT, Gemini, Copilot) are transforming work in IT-enabled settings and influence employee outcomes. It addresses the research gap on adoption and impacts of generative AI in workplaces, focusing on user perceptions (usefulness, ease of use), trust, organizational commitment, employee engagement, and performance. Key issues posed include: whether intention and user perception of generative AI drive acceptance and use; whether trust is vital; and how organizational commitment, engagement, and performance interrelate in the generative AI context. Objectives are to determine and validate user perceptions influencing use of generative AI tools; test whether trust mediates the link between user perception and organizational commitment; and assess how organizational commitment affects engagement and, in turn, how engagement affects performance. The study is grounded in TAM, TRI, and SOR theories, positing user perception (ease of use, usefulness) and technology readiness factors (optimism, innovativeness) as antecedents, with trust as a mediator toward organizational outcomes.

The review highlights that generative AI can automate routine tasks, enhance productivity, and improve human-computer interaction, positively affecting organizational performance, engagement, commitment, and growth. Despite interest, research on adoption intentions and impacts in HR contexts remains limited across sectors. Concerns persist about privacy, ownership, job displacement, and potential negative effects on engagement and commitment. Existing frameworks such as UTAUT (including UTAUT2/3), TAM, TRI, and SOR have been applied to study technology adoption, emphasizing perceived usefulness, ease of use, optimism, innovativeness, and trust. Trust is identified as crucial in human–AI interaction, influenced by reliability, prior experience, data security, and privacy. Based on the literature, the authors develop a theoretical framework integrating TAM (ease of use, usefulness), TRI (optimism, innovativeness), and SOR, modeling trust as a mediator between user perception and organizational commitment, and linking organizational commitment to employee engagement and engagement to performance. Hypotheses propose positive effects of optimism and innovativeness on user perception; user perception on trust; trust on organizational commitment; organizational commitment on engagement; engagement on performance; and a mediating effect of trust between user perception and organizational commitment.

Design: Cross-sectional survey using a structured questionnaire with two sections: demographics and constructs from TAM-SOR-TRI. A 7-point Likert scale (1=strongly disagree; 7=strongly agree) was used. Content validity was established via expert review (four professionals in management and IT and an English language consultant). A pilot with 100 IT employees (software engineers, testing specialists, team leaders, project managers) preceded the main study. Sampling and data collection: Convenience sampling targeting IT employees who use generative AI tools. Data were collected via Google Forms from January to March 2024. Of 552 responses, 52 were removed for poor quality or insufficient completion; 500 valid responses remained. Sample size and power: Justified using multiple criteria (e.g., 50 + 5x items; with 33 items, minimum ~215). Power analysis in SPSS (alpha=0.05, SD=1.17) yielded power=0.955 with effect size 0.82 at N=500. Cochran’s formula for unknown population suggested 385; N=500 exceeded requirements. Measures: Nine reflective constructs (33 items total): Optimism (3 items; TRI), Innovativeness (4; TRI), Perceived Usefulness (3; TAM), Perceived Ease of Use (3; TAM), Trust (5; based on McKnight et al., Glikson & Woolley, others), Organizational Commitment (5; Allen & Meyer; Meyer et al.), Vigor (3; UWES), Dedication (3; UWES), Performance (5; Pradhan & Jena). Higher-order constructs: User Perception (Ease of use + Usefulness) and Employee Engagement (Vigor + Dedication). Analysis: EFA (IBM SPSS) extracted 9 components explaining 81.961% cumulative variance. KMO=0.929; Bartlett’s test p<0.001, indicating factorability. CFA/SEM (AMOS v28) assessed measurement and structural models, including higher-order constructs. Reliability: Cronbach’s alpha 0.897–0.948. Convergent validity: loadings >0.6; AVE >0.5. Discriminant validity: Fornell-Larcker satisfied; HTMT ratios <0.85. Model fit: Lower-order 9-factor model fit well: CMIN/df=1.99; CFI=0.967; GFI=0.938; TLI=0.925; IFI=0.987; NFI=0.958; SRMR=0.033; RMSEA=0.024; PClose=1.000. Higher-order model (User Perception; Employee Engagement) also fit well: CMIN/df=2.905; CFI=0.947; GFI=0.928; TLI=0.925; IFI=0.957; NFI=0.938; SRMR=0.037; RMSEA=0.048. Common method bias: Harman’s single-factor test via CFA indicated poor fit for a single-factor solution. Latent common method factor test showed a small chi-square difference (Δχ²=4.10; df diff=1) between models with and without the method factor (χ²=915.700, df=459 vs χ²=919.800, df=460), suggesting minimal CMB impact. Structural model: Squared multiple correlations (R²): User Perception=0.63 (explained by Optimism, Innovativeness); Trust=0.38 (by User Perception); Organizational Commitment=0.30 (by User Perception and Trust); Employee Engagement=0.50 (by Organizational Commitment); Employee Performance=0.54 (by Employee Engagement). Mediation analysis tested Trust mediating User Perception → Organizational Commitment.

- All hypothesized relationships were supported. - Antecedents of User Perception: Optimism → User Perception β=0.416, t=9.176, p<0.001 (stronger effect than Innovativeness); Innovativeness → User Perception β=0.081, t=3.964, p<0.001. - User Perception → Trust: β=0.891, t=8.906, p<0.001. - Trust → Organizational Commitment: β=0.101, t=2.290, p<0.05. - Organizational Commitment → Employee Engagement: β=0.512, t=7.994, p<0.001. - Employee Engagement → Employee Performance: β=0.750, t=8.048, p<0.001. - Mediation: User Perception had a significant direct effect on Organizational Commitment (β=0.758, t=6.218, p<0.001) and a significant indirect effect via Trust (indirect β=0.090, t=7.456, p<0.001), indicating partial mediation by Trust. - Explained variance (R²): User Perception=0.63; Trust=0.38; Organizational Commitment=0.30; Employee Engagement=0.50; Employee Performance=0.54. - Measurement quality: Cronbach’s alpha 0.897–0.948; AVE >0.5; HTMT <0.85. - Model fit indices indicated excellent fit for both lower-order and higher-order models (e.g., CFI up to 0.967; RMSEA as low as 0.024).

Findings show that technology readiness factors (optimism, innovativeness) improve user perception (ease of use, usefulness) of generative AI. Positive user perception strongly builds trust in AI tools, which in turn enhances organizational commitment. Higher organizational commitment fosters employee engagement, leading to improved performance. Trust partially mediates the effect of user perception on organizational commitment, underscoring trust’s central role in converting favorable perceptions into organizationally relevant attitudes. Results corroborate prior evidence that AI adoption can strengthen engagement and performance and highlight optimism and innovativeness as key antecedents of continued AI use. The integrated TRI–TAM–SOR framework effectively explains how individual technology readiness and perceptions translate into trust, commitment, engagement, and performance in IT workplaces using generative AI.

The study integrates TRI, TAM, and SOR to explain adoption and downstream HR-related outcomes of generative AI in IT settings. User perception (ease of use, usefulness) and technology readiness (optimism, innovativeness) drive trust; trust enhances organizational commitment; commitment elevates engagement; and engagement boosts performance. Both lower- and higher-order models demonstrated excellent psychometric properties and fit. Practically, organizations should communicate benefits, provide guidance and training, and prioritize usability and reliability of AI tools to build trust and improve HR outcomes. The work offers a research agenda to examine generative AI from multiple theoretical perspectives and informs HRM practices such as talent acquisition, training, and workforce management.

- Geographic scope: Sample restricted to IT employees in Hyderabad (Indian metro), limiting generalizability to other regions and cultures. - Sampling: Convenience sampling may introduce selection bias; respondents were tech-savvy users of generative AI. - Self-report measures: Potential for common method bias and response biases, though checks indicated minimal CMB. - Cross-sectional design: Causal inferences are limited; longitudinal studies are recommended to track evolving relationships as AI tools change. - Rapid technological evolution: Findings may require updates as generative AI capabilities and usage practices evolve. - Model scope: Focused on TRI–TAM–SOR; other frameworks (e.g., diffusion of innovation, IS service quality, UTAUT2) and moderators (e.g., HR practices, organizational factors, gender) warrant integration in future research.