When resistance becomes knowledge: governing health-care AI through federated learning and explainable AI

Purpose Digital transformation in health care is frequently delayed by resistance, which is typically conceptualized as a barrier to AI adoption. This study aims to reframe resistance as a knowledge-generating signal, examining how it shapes organizational learning and knowledge governance during AI implementation in high-stakes, regulated settings. Design/methodology/approach The study draws on a canonical action research program conducted within the eHealth Network: Artificial Intelligence and Innovative ICT Tools Oriented toward Digital Diagnostics project. Empirical material was collected across nephrology, hepatology and diabetology through 55 interdisciplinary meetings, 17 semistructured interviews and extensive project documentation, enabling longitudinal analysis of how knowledge is produced, negotiated and institutionalized during AI implementation. Findings The findings show that resistance operates as diagnostic signal, revealing misalignments across conceptual, representational and evidential knowledge proximities. These tensions trigger the development of boundary infrastructures, including shared vocabularies, standardized test datasets and interdisciplinary routines, through which knowledge is created, validated and governed across professional communities. The analysis identifies two socio-technical mechanisms that mediate these knowledge tensions: federated learning, which functions as a knowledge-governance capability balancing data sovereignty with institutional learning and explainable AI, which, when institutionalized through clinical explanation rounds, supports collective sensemaking and embeds interpretability into everyday decision-making routines. Originality/value The study contributes to the knowledge management and information systems literature by reconceptualizing resistance as a generative constraint for governing knowledge in AI-enabled organizations. It advances understanding of how socio-technical mechanisms can transform resistance into an asset for organizational learning under epistemic pluralism, offering insights applicable beyond health care to other regulated and knowledge-intensive domains.