AI-Driven Retrieval-Augmented System Enhances Health Insurance Support

A new AI system based on Retrieval-Augmented Generation (RAG) has been developed to enhance interaction with health insurance information. Unlike previous models that separate conversational assistance, policy recommendation, and document retrieval, this approach integrates these tasks into one unified architecture. The system features a chatbot for answering general insurance questions, a RAG-driven policy recommendation engine that utilizes both structured and unstructured data, and a document retrieval module for clause-level searches within uploaded policies. A key innovation is the evaluator agent, which simulates human-like judgment to assess the quality of responses across parameters such as relevance, accuracy, clarity, and helpfulness, thereby creating an automated feedback loop to improve system performance over time. Experimental results show high semantic retrieval accuracy (BERTScore F1 up to 0.84), strong recommendation metrics (Hit@5=1.0, Recall@5=0.833), and effective clause retrieval (BERTScore F1=0.8443). The domain-specific application of RAG combined with modular architecture and quality assessment reduces risks of inaccurate AI outputs and enhances policy transparency. This development signifies a step forward in AI-assisted insurance services, potentially improving user experience and compliance by delivering precise and contextually relevant policy information. The research data supporting this system is available upon reasonable request from the authors.