Building a Global HR Policy Assistant with GraphRAG: A Beginner’s Guide

A step-by-step guide to building a global HR policy assistant using GraphRAG and Neo4j. The core technique introduced is 'classifier nodes' (e.g., BenefitType, BenefitCategory) that connect extracted entities into a navigable knowledge graph instead of isolated document fragments. The guide covers a custom multi-phase pipeline: PDF ingestion, text chunking, embedding, two-phase LLM entity extraction (document-level and chunk-level), entity resolution, index creation, and post-processing to create classifier nodes. It then shows how to deploy a Neo4j Aura Agent on top of the graph to answer employee HR questions using a combination of vector similarity search and graph traversal. Example queries demonstrate how the agent handles contextual questions about leave entitlements and medical insurance across countries and organizational bands.