AI in Service Industries: Striking a balance between Value Co-Creation and Co-Destruction

Purpose – This conceptual paper aims to examine how interactional conditions in human–AI service systems give rise to heterogeneous value configurations over time, and how persistent tensions shape their evolution. Design/methodology/approach – This conceptual study builds on Interactive Value Formation and Paradox Theory to develop a framework focused on agency distribution and practice alignment. Four paradoxical tensions are identified as generative mechanisms, with illustrative examples from banking, recruitment, telecommunications, and hospitality used to illustrate the model’s dynamics. Findings – Value unfolds in four configurations – co-creation, co-destruction, no-creation and compliance – shaped by the interaction of distributed/exclusive agency and aligned/misaligned practices. Paradoxical tensions (automation–augmentation, transparency–opacity, standardization–personalization and autonomy–control) act as generative mechanisms that trigger shifts across configurations over time, helping explain the dynamic and recursive nature of value formation in human–AI service systems. Research limitations/implications – As a conceptual framework, the model is not empirically tested and relies on illustrative examples.Its applicability may vary across organizational contexts. Future research could examine the dynamics identified here through longitudinal and process-oriented studies. The framework extends Interactive Value Formation by accounting for AI-related agency redistribution and paradoxical dynamics. Practical implications – Managers can use the model primarily as a diagnostic lens to surface misalignments and anticipate trade-offs, informing intervention priorities. Originality/value – The study reconceptualizes value formation as a paradox-driven, recursive process; introduces agency distribution and practice alignment asitsstructural foundation; and offers a diagnostic lensto support managerial sensemaking of human–AI interactions.