Agentic AI-Based Indian Legal Notice and Complaint Understanding with Source-Grounded Next-Step Guidance

This paper proposes an Agentic AI-based architecture for understanding Indian legal notices and complaints using Retrieval-Augmented Generation (RAG) and Explainable AI (XAI). The proposed system aims to help ordinary citizens understand complex legal documents by extracting key information, identifying relevant legal provisions, and providing source-grounded explanations and next-step guidance. The architecture consists of an Orchestrator Agent and six specialized agents: Document Analysis, Legal Research, Summarization, Explainability, Next-Step Guidance, and Validation. It combines BM25 keyword search with FAISS dense-vector retrieval and cross-encoder re-ranking to identify relevant Indian statutes and legal references. The proposed generation component uses a LoRA-adapted LLaMA-3.1-8B-Instruct model. The system emphasizes traceability, factual grounding, and validation to reduce unsupported legal claims. It is intended to assist users in understanding legal documents and does not replace professional legal advice. This paper presents the proposed architecture and methodology, along with illustrative hypothetical case studies. Implementation and empirical evaluation remain future work.