AI Foundations
Agentic Retrieval
Agentic Retrieval refers to an AI agent's direct, tool-based access to complete documents. Unlike RAG, which splits documents into chunks and retrieves only the most similar ones, the agent can open, navigate, and read documents in context.
How Agentic Retrieval works
Agentic Retrieval refers to an AI agent’s direct, tool-based access to complete documents. Unlike classic RAG, which splits documents into text chunks and retrieves only the most similar ones, the AI agent can open, navigate, and read documents in context.
Contrast with RAG
RAG hits its limits with complex, legally structured documents such as general insurance terms and conditions, because clauses reference one another and isolated text snippets lead to incorrect answers. Agentic Retrieval avoids this problem: the agent reads the complete document in context instead of relying on isolated snippets.
Agentic Retrieval at LoyJoy
LoyJoy offers Agentic Retrieval as an AI agent module with document tools. One example is answering questions from policy terms for insurers: the AI agent works directly with the complete policy document and correctly answers coverage and benefit questions, even when clauses reference one another.