Veranstaltungsprogramm

Traditional knowledge management systems (KMS) support the storage, retrieval, and recombination of explicit knowledge but remain limited in facilitating tacit, practice-embedded, and context-dependent knowledge reuse. This limitation is particularly critical in collaborative business processes (CBP), where reuse frequently fails because the original context and situated meaning of knowledge are insufficiently preserved. Two persistent challenges follow: the "Tacit Gap" – the difficulty of transferring tacit, practice-embedded knowledge through repository-based systems – and the "Context Gap" – the loss of contextual grounding during knowledge reuse. With the emergence of generative artificial intelligence (GenAI) and large language models (LLMs), KMS are shifting towards agentic, dialogue-based systems that can act as interactive, context-sensitive intermediaries in organizational knowledge processes. However, organizational applications of GenAI in knowledge management often remain tool-driven, and research has not yet consolidated how GenAI capabilities can be systematically aligned with established knowledge theories.

Addressing this gap, this research in progress asks: How can GenAI role functionalities be systematically designed to support knowledge conversion and knowledge reuse in collaborative business processes? Following a theory-driven, design-oriented research approach, we develop a taxonomy using Nickerson et al.'s iterative method. The taxonomy combines two complementary theoretical perspectives: the SECI model of knowledge conversion (Nonaka & Takeuchi, 1995) and Markus's (2001) typology of knowledge reuse situations. The resulting 4×4 taxonomy specifies 16 distinct GenAI role functionalities, each defined by its knowledge conversion output, its target reuse situation, and its SECI-specific GenAI capability grounding, and each assignable to exactly one cell through two operational coding rules.

To demonstrate technical feasibility, four roles – the Scribe Agent, Semantic Linking Agent, Mentor Agent, and Concept Mining Agent – were instantiated in a prototype, the Intelligent Knowledge Management Assistant System (IKMAS). IKMAS implements a state-machine-based processing flow in which an LLM-based router classifies incoming requests along the taxonomy, executes role-specific agent configurations, and independently verifies outputs, looping back to context enrichment when applicability conditions are missing. Preliminary testing shows that user requests were assigned to the corresponding roles and that role-specific instructions were consistently executed. The completed study is expected to contribute a theoretically grounded taxonomy as design knowledge for GenAI-enhanced KMS, a reference architecture demonstrating its technical operationalization, and a basis for evaluating GenAI as a context-sensitive knowledge intermediary in organizations.