playbooks/1 min read
RAG for support — grounding answers in your catalog
Retrieval keeps an agent's answers tied to your real products, policies, and docs. A practical look at building retrieval that support teams can trust.
A support agent that "sounds confident" but invents a return policy is worse than no agent at all. Retrieval-augmented generation (RAG) ties every answer back to sources you control.
Ground first, generate second
Before the model writes a word, it retrieves the relevant policy, product, or doc. The answer is then built from that context — not from memory.
What to index
- Product catalog and stock
- Shipping, returns, and warranty policies
- Help-center articles and macros
If it isn't in a source the agent can cite, it shouldn't be in the answer.
Keep it fresh
Re-index when content changes. Stale retrieval is just a slower way to be wrong.