RAG or fine-tuning? Start with the knowledge problem.
Retrieval-augmented generation finds relevant source material before a model writes an answer. It is useful when answers depend on changing documents and need citations. Fine-tuning changes model behavior; it does not automatically guarantee current or correct factual knowledge. For a business knowledge assistant, first assess content quality, access rules, retrieval coverage, and evaluation.