From Walking Libraries to Answer Engines: Optimizing Knowledge Internalization in Public Financing with RAG-Based AI Chatbots

Abstract

This research arises from the critical need to optimize knowledge internalization within the SECI Model, especially in organizations handling complex and regulated information such as public financing. The study aims to identify the limitations of the existing Knowledge Management System (KMS), implement a Retrieval-Augmented Generation (RAG)-based AI Chatbot prototype, and evaluate its effectiveness in facilitating knowledge internalization. A qualitative case study approach was employed using the Design Science Research Method (DSRM). Data were collected through in-depth interviews and prototype testing involving 10 purposively selected respondents from the Public Financing Division of PT SMI. Findings indicate that the conventional KMS is suboptimal, leading to high dependency on social learning and the "walking library" phenomenon, where critical knowledge is siloed within specific individuals rather than being institutionalized. The implemented AI Chatbot prototype successfully accelerated the learning curve for new employees, reduced cognitive load – the mental effort required to search for and synthesize information, and received positive evaluation based on the KMS Success Model dimensions. The study implies that organizations should transition their knowledge-sharing culture from individual dependency to smart system-based interactions.