Leveraging Large Language Models for Information Extraction in Project Risk Management

Abstract

Effective risk management is crucial but challenging in modern projects due to inherent complexities and the dynamic emergence of risks within informal, unstructured data sources. Traditional approaches often fail to proactively identify risks, creating significant detection gaps. This paper introduces a novel architecture leveraging Large Language Models (LLMs) tailored explicitly to address information extraction (IE) in project risk management (PRM). Using a Design Science Research (DSR) approach, we develop and evaluate an architecture that integrates diverse unstructured data, facilitating continuous, proactive, and context-aware risk identification. The proposed architecture incorporates aggregation, orchestration, and specialized risk agents, allowing for nuanced, timely extraction and structuring of risk indicators. Through iterative development and expert validation, our artifact demonstrates substantial potential to enhance proactive risk management, bridging critical gaps between informal risk emergence and formal identification processes.