RAG (Retrieval-Augmented Generation)
An AI pattern that retrieves relevant documents first, then generates an answer grounded in them.
RAG is the architecture behind most trustworthy support AI. Instead of answering from the model's memory alone, it first retrieves the most relevant passages from your knowledge base or ticket history, then asks the model to answer using only that grounded context. This is what keeps an AI reply tethered to your actual policies rather than a plausible invention.
It is a technique, not a metric, though retrieval quality and grounding rate are the things to watch.
What it hides: RAG is only as good as the knowledge it retrieves. Point it at a stale, contradictory, or gap-ridden knowledge base and it will confidently generate wrong answers from bad sources — garbage in, fluent garbage out. The knowledge-base hygiene problem doesn't disappear; RAG makes it more consequential.