MULTI-AGENT RAG ARCHITECTURE FOR END-TO-END HEALTHCARE ADMINISTRATIVE AUTOMATION

Abstract

Healthcare administrative processes involve numerous interconnected activities including prior authorization, claims adjudication, medical coding, policy verification, provider eligibility assessment, and reimbursement validation. Conventional workflow automation systems rely on isolated rule-based components that often struggle with complex decision-making and rapidly changing healthcare regulations. MultiAgent Retrieval-Augmented Generation (RAG) combines autonomous artificial intelligence agents with enterprise knowledge retrieval to enable collaborative reasoning and intelligent workflow automation across healthcare administrative functions. This paper proposes a Multi-Agent RAG architecture integrating specialized AI agents, semantic retrieval, vector databases, enterprise knowledge management, workflow orchestration, and cloud-native MLOps governance for end-to-end healthcare administrative automation. The proposed framework enables autonomous coordination among specialized agents responsible for policy retrieval, coding validation, clinical reasoning, claims adjudication, and workflow management while grounding every decision using authoritative enterprise knowledge. Experimental evaluation demonstrates improvements in workflow efficiency, retrieval accuracy, administrative productivity, regulatory compliance, operational scalability, and decision consistency. The proposed architecture provides a production-ready solution for intelligent healthcare administration. Keywords— Multi-Agent AI, RetrievalAugmented Generation, Healthcare Administration, Workflow Automation, Large Language Models, Semantic Search, MLOps, Enterprise AI.