This research formulates a multi-level routine dynamics framework to elucidate how service organizations manage paradoxical tensions resulting from the adoption of artificial intelligence (AI). AI offers the potential for enhanced efficiency and innovation; however, it frequently presents conflicting demands—such as control versus flexibility or automation versus learning—that complicate conventional change management strategies. To tackle this issue, the study amalgamates perspectives from paradox theory, routine dynamics, and dynamic capabilities via a conceptual, theory-building methodology. The framework delineates four categories of AI-related paradoxes—performing, organizing, learning, and belonging—and aligns them with a hierarchical system of routines. It shows how navigating paradoxes happens over and over again on different levels, such as individual, routine, and organizational, and how routines have both ostensive and performative aspects. The paper also presents a "paradox-responsive scorecard" as a decision-support instrument to assess the efficacy of organizations in managing conflicting AI outcomes. The primary contribution of the study is to provide a systematic elucidation of how service organizations can oversee and facilitate paradox navigation across varying capability levels to improve decision-making processes during AI implementation.

Keywords

Change management, service operations, AI, paradox theory, routine dynamics