A Decision-centric Knowledge Architecture for Knowledge Retention

Abstract

Organizations depend on expert judgement to interpret complex operational situations and maintain consistent decision-making under conditions of uncertainty, regulatory complexity and contextual variation. In SMEs as well as multinational enterprises, the reasoning structures underlying such decisions remain only partially documented and are embedded in the experience of individuals. When experienced employees leave an organization, operational decision capability may be lost. Traditional knowledge management approaches primarily focus on documenting explicit knowledge and procedural information and therefore struggle to preserve organization-specific interpretive decision structures. This paper develops and evaluates a Decision-Centric Knowledge Architecture (DCKA) designed to reconstruct and operationalize expert decision logic. The paper follows a Design Science orientation and addresses the question of how organization-specific decision structures can be modelled in a way that supports traceable interpretive reconstruction. DCKA integrates structured knowledge extraction, graph-based representation of interconnected decision structures and AI-supported interaction grounded in explicit decision structures. DCKA was originally implemented in the payroll department of an international organization facing the retirement of a highly experienced domain expert. Expert knowledge was reconstructed through structured interviews and transformed into an interconnected knowledge graph capturing contextual dependencies, decision signals, evaluation logic, governance constraints and exception patterns. The knowledge graph was connected to a large language model through a controlled retrieval process enabling conversational access to organization-specific decision structures. Evaluation showed that DCKA reconstructed operational decision structures with sufficient plausibility to support expert transition, onboarding and organizational review. In addition, exploratory applications in research, consulting and educational contexts indicated transferability beyond the original payroll domain. The findings suggest that DCKA can support continuity of operational decision-making through traceable reconstruction of implicit interpretive structures. The paper positions decision-making capability as a central object of knowledge retention and extends knowledge management research beyond document-centric approaches.