RAG
RAG is the technique that lets an AI answer from your documents instead of its general training: the system first retrieves the relevant passages from your knowledge, then has the model answer using them, ideally with sources shown.
Why it matters
Out of the box, an LLM knows nothing current or private about your business, and will improvise when asked. RAG is the standard fix, and it's what makes "ask questions of our own knowledge" tools trustworthy enough to use: answers grounded in your actual documents, with receipts.
How it works
Your documents are indexed for meaning-based search. A question retrieves the most relevant passages, which are handed to the model alongside the question: "answer from this." Quality lives and dies on the retrieval and the source material: stale or contradictory documents produce confidently stale answers.
What to do about it
The classic first build: point RAG at your policies, proposals and product docs so the team stops re-asking each other. Budget as much effort for curating the documents as for the tech.
