Retaining Tacit Knowledge Through Dialogue: The AI Moderator in Work Processes

Abstract

Tacit knowledge is regarded as a central difficult-to-access resource for organizational learning. Drawing on Michael Polanyi (1985), this kind of knowledge is closely tied to individual action, experience, and contexts, and largely resists documentation capture. In complex work settings, tacit knowledge constitutes a basis for performance and quality assurance. Against the backdrop of demographic change and skills shortages, the challenge of securing, transferring, and making this knowledge usable for subsequent generations is intensifying: around one quarter of the workforce in Germany will reach retirement age within the next ten to fifteen years (Statistisches Bundesamt, 2025). This paper presents an AI-based assistance approach developed at the Fraunhofer Institute for Factory Operation and Automation IFF to support workers when additional guidance is needed, particularly during complex or non-routine tasks. The assistance enables hands-free interaction and goes beyond AI-driven dialogue management by addressing upstream and downstream processes, including preparing existing knowledge for AI use, structured content modelling, and linguistic adaptations. At its core, KIMO (short for AI Moderator) is a dialogue-based AI system that supports employees within their work processes. KIMO facilitates reflection on one’s actions and fosters the articulation of experiential knowledge that remains implicit in everyday practice. The approach builds on narrative methods that prompt reflection and the telling of experience-based stories, making implicitly applied heuristics and decision logics more visible (Erlach & Thier, 2004). These methods are complemented by an instructional dialogue design aligned with Bloom’s taxonomy of learning objectives (2004). Unlike conventional knowledge management systems, which focus on retrospective documentation and explicit knowledge repositories, KIMO emphasizes AI-based interaction at the moment when knowledge emerges. Through its dialogic mode, person-bound experiential knowledge is externalized, structured, and transformed into organizationally usable knowledge structures. The paper discusses the potential of dialogue-based AI systems for sustainable knowledge transfer and the long-term preservation of organizational memory. It concludes by outlining implications for the design of AI-enabled knowledge management systems, including required information and content infrastructures (e.g., a structured knowledge base).