A multi module a.i. system for intelligent health insurance support using retrieval augmented generation.

Area of Science:

• Artificial Intelligence
• Health Insurance Informatics
• Natural Language Processing

Background:

• Interacting with health insurance information is complex.
• Existing AI solutions often address conversational assistance, policy recommendation, and document retrieval in isolation.
• A unified approach is needed to streamline user interaction with health insurance data.

Purpose of the Study:

• To develop a Retrieval-Augmented Generation (RAG)-based multi-module AI system for health insurance information.
• To integrate conversational AI, policy recommendation, and document retrieval into a single architecture.
• To introduce an AI evaluator agent for automated quality assessment of responses.

Main Methods:

• Developed a unified AI system with integrated chatbot, RAG-powered policy recommender, and clause-level document retrieval.
• Utilized structured and unstructured policy data for the recommendation engine.
• Implemented an evaluator agent to assess response quality (relevance, accuracy, clarity, helpfulness).

Main Results:

• Achieved strong semantic retrieval performance (BERTScore F1 up to 0.84).
• Demonstrated robust policy recommendation (Hit@5 = 1.0, Recall@5 = 0.833).
• Showcased effective clause retrieval from policy documents (BERTScore F1 = 0.8443).

Conclusions:

• The proposed RAG-based modular AI system effectively streamlines health insurance interactions.
• The inclusion of a quality-assessment agent reduces hallucination and improves transparency.
• This approach offers enhanced, user-focused support in the health insurance domain.