ARTIFICIAL INTELLIGENCE IN HEALTHCARE ORGANIZATIONS: UNDERSTANDING THE MULTILEVEL ROOTS OF ORGANIZATIONAL IMPLEMENTATION

Artificial Intelligence (AI) is expected to transform our societies due to its transformative, general-purpose nature. However, the implementation of AI technologies within organizational processes takes place at a path that is much slower than the rapid advancements in technology and research. Despite their large potential to improve health outcomes and operational efficiency, healthcare organizations as well suffer from a fragmented and uneven adoption and implementation of AI technologies in organizational routines. The translation of technical capabilities into routine use is hindered by the distinctive institutional, organizational, and professional features of these organizations. This dissertation contributes to the management and organization literature by investigating the multilevel roots of AI implementation in professional and knowledge-intensive organizations. It pursues a twofold objective: (i) identifying how management and organization research has conceptualized the adoption and utilization of AI and (ii) examining how organizational and professional dynamics shape the integration of AI technologies in organizational routines. First, the dissertation develops a semi-systematic review of 70 top-tier articles, highlighting the sociotechnical nature and the interdependence of system, strategic, organizational, and individual dynamics in AI adoption and utilization. The review provides the conceptual ground to develop the empirical contribution of the dissertation. The research then explores how professional and research-oriented organizations manage AI innovation through a multiple case study of four large research and teaching hospitals in Italy. The study identifies four core tensions – organizing AI exploration, navigating the make-buy-ally continuum, managing internal demand for AI innovation, and balancing internal and external competencies – showing how organizations enact multilevel ambidextrous configurations to govern AI innovation. The second empirical contribution follows a large-scale empirical analysis of 1.5 million AI-based treatment recommendations in nephrology and provides evidence on how physicians engage with algorithmic advice. The findings reveal that professional reliance is strongly anchored in physician’s norms, habits, and routines and highlight a tension between algorithmic and professional rationalities. Overall, the dissertation develops a multilevel account of AI implementation in healthcare as an industry dominated by professional, knowledge-intensive, and research-oriented organizations. It extends the literature on technology and innovation management by illustrating how knowledge arrangements, organizational capabilities, and professional logics jointly shape AI adoption and utilization. In doing so, it outlines implications for policy and management and future research directions on how organizations can navigate the opportunities and challenges of AI-driven organizational transformation.