Retrieval-augmented generation has become a standard pattern for grounding large language models in enterprise data. A typical implementation converts documents into embeddings, stores them in a vector database, retrieves the most similar chunks for a query, and adds those chunks to the model prompt. This works well for document lookup, policy search, support content, and other tasks where semantic similarity is the main requirement. Enterprise knowledge, however, is rarely organized as isolated passages. It is distributed across applications, databases, APIs, documents, ownership hierarchies, product catalogs, and operational records. Once questions require relationships, provenance, time, or multi-step reasoning, vector retrieval alone becomes unreliable. The next stage of enterprise AI therefore depends on combining RAG with enterprise knowledge graphs.