Clear communication, staff consultation and targeted training during AI rollouts are linked to higher job satisfaction and self-reported performance by strengthening employees' sense of meaningful work, especially among those with positive views of AI.

Main Finding

Employee-centered AI implementation practices (transparent communication, consultation, and training) positively affect white-collar employees’ job satisfaction and self-reported job performance both directly and indirectly by increasing work meaningfulness. The mediated (indirect) effect through meaningfulness is stronger for employees who hold more positive attitudes toward AI.

Key Points

• Definition: Employee-centered AI implementation (ECAII) = transparent communication about AI, employee consultation/involvement, and targeted training initiatives.
• Direct effects: ECAII practices → higher job satisfaction; ECAII practices → higher job performance.
• Indirect effects: ECAII practices → greater perceived work meaningfulness → higher job satisfaction and performance.
• Moderation: Employees’ personal attitudes toward AI amplify the indirect pathway — workers with more positive AI attitudes experience stronger meaningfulness-mediated benefits.
• The study extends High-Involvement Management and meaningful-work theories to AI-enabled workplace change and offers practical guidance for HRM during AI adoption.

Data & Methods

• Sample: Survey of 168 Italian white-collar employees who actively use AI in their work.
• Design: Cross-sectional survey data (self-reported measures).
• Measures: ECAII practices (communication, consultation, training), work meaningfulness, job satisfaction, job performance, attitude toward AI.
• Analyses: Structural equation modeling to estimate direct and mediated relations; latent moderated mediation analyses to test moderation by AI attitudes.
• Limitations (methodological): relatively small and country/occupation-specific sample; reliance on self-reported performance and cross-sectional data limits causal inference.

Implications for AI Economics

• Firm-level returns to AI adoption: Employee-centered implementation increases the likelihood that AI generates productivity and performance gains, suggesting that complementarities between AI capital and human-management practices matter for realized returns on AI investments.
• Human capital and training investment: Targeted training and involvement are important complements to AI deployment; economics models of AI adoption should account for managerial/HR inputs as part of adoption costs and sources of heterogeneity in outcomes.
• Labor supply and retention: Higher job satisfaction and perceived meaningfulness under ECAII imply lower turnover risk and potentially reduced hiring/training costs — important when computing net benefits of AI-driven automation/augmentation.
• Heterogeneous effects and behavioral frictions: Workers’ attitudes toward AI moderate benefits, indicating demand-side heterogeneity (acceptance/resistance) that can affect adoption diffusion, equilibrium wages, and distributional outcomes. Models should incorporate psychological and cultural factors beyond skill composition.
• Policy and regulatory relevance: Public incentives for AI diffusion could be more effective if paired with support for worker-centered practices (training subsidies, guidelines for transparent implementation) to maximize social returns and mitigate transitional costs.
• Data and measurement priorities: For economic evaluations, objective productivity metrics and longitudinal designs are needed to quantify net productivity effects and dynamic labor-market consequences; this study’s findings motivate inclusion of managerial practices as variables in empirical work on AI’s economic impacts.

Suggestions for further research relevant to AI economics: replicate with larger and more diverse samples, use objective performance and firm-level productivity data, employ longitudinal/quasi-experimental designs to identify causal effects, and quantify cost–benefit trade-offs of ECAII at firm and sector levels.