The Dark Side of AI: When Automation Reduces Employee Engagement and Output

Artificial intelligence (AI) and automation are increasingly adopted by organizations to improve efficiency, reduce operational costs, accelerate decision-making, and enhance productivity. However, a growing body of research suggests that AI does not always generate positive outcomes for employees or organizations. This study examines the dark side of AI by exploring the conditions under which automation may reduce employee engagement and output. Using a desk-based qualitative research design, the paper synthesizes prior studies on AI adoption, technostress, job insecurity, emotional exhaustion, psychological capital, perceived organizational support, work engagement, and individual productivity. Drawing on Conservation of Resources Theory, Job Demands–Resources Theory, Self-Determination Theory, and socio-technical systems theory, the study proposes that AI automation can become harmful when it is perceived as a threat to job security, autonomy, competence, professional identity, and psychological resources. The thematic analysis indicates that AI may reduce employee engagement and sustainable output through several mechanisms: increased technostress, emotional exhaustion, job insecurity, role ambiguity, work intensification, reduced autonomy, over-reliance on algorithmic recommendations, and depletion of psychological capital. Evidence from recent studies further suggests that organizational support, employee involvement, digital training, psychological capital, and transparent AI implementation can mitigate these negative effects. The findings contribute to the AI and work literature by shifting the discussion from a purely productivity-oriented view of automation to a more balanced understanding of its psychological and organizational consequences. Practically, the study suggests that organizations should not evaluate AI adoption only through cost reduction, speed, or output volume, but also through employee engagement, well-being, autonomy, learning, and long-term work quality. AI creates sustainable value only when it is implemented as a human-augmentation tool rather than merely as a labor-substitution mechanism.