Agentic Artificial Intelligence for Information Fusion

Abstract

Organizations across healthcare, enterprise analytics, supply chains, and cybersecurity struggle to integrate heterogeneous data sources for timely, reliable decision-making. Traditional information fusion relies on static, pipeline-oriented architectures inadequate for dynamic, distributed, and real-time environments. This study presents a PRISMA-guided systematic review integrating agentic decision theory with organizational information systems perspectives, including the Technology Acceptance Model, Task-Technology Fit, and Sociotechnical Systems Theory. A search of four databases yielded 1,205 records, from which 380 unique publications were identified. A six-dimension taxonomy covering architectures, learning paradigms, coordination mechanisms, decision coupling, trust modeling, and uncertainty representation is developed. Hybrid cloud-edge and federated architectures with multi-agent reinforcement learning offer favorable scalability and robustness trade-offs. Bias-aware design, human-in-the-loop accountability, and explainable outputs are necessary for responsible deployment.