Revealing the Unspoken: Using LLMs to Mobilize and Enrich Tacit Knowledge in Event Logs of Knowledge-Intensive Processes

Large Language Models (LLMs) excel in understanding, generating, and processing human language, with growing adoption in process mining. Process mining relies on event logs that capture explicit process knowledge; however, knowledge-intensive processes (KIPs) in domains such as healthcare and product development depend on tacit knowledge, which is often absent from event logs. To bridge this gap, this study proposes a LLM-based framework for mobilizing tacit process knowledge and enriching event logs. A proof-of-concept is demonstrated using a KIP-specific LLM-driven conversational agent built on GPT-4o. The results indicate that LLMs can capture tacit process knowledge through targeted queries and systematically integrate it into event logs. This study presents a novel approach combining LLMs, knowledge management, and process mining, advancing the understanding and management of KIPs by enhancing knowledge accessibility and documentation.