The short answer
A language model writes fluent text from patterns it learned. Asked for your price, it may write a believable number that is wrong. That is often called a hallucination.
Grounding, often built with retrieval-augmented generation (RAG), first retrieves relevant facts from your data and then writes the answer using only those facts. If nothing relevant is found, the right answer is “I don’t know”.
What GorZesk does about it
GorZesk’s assistants retrieve from your catalog, inventory and orders before answering, keep tenant data separate, and hand off when confidence is low. Answers about prices and stock come from your records.
Limits to know
- Grounding reduces wrong answers; it does not remove them. Data can be missing or stale, so testing and review still matter.
Common questions
Is RAG the same as grounding?
RAG is a common technique for grounding: retrieve first, then generate from what was retrieved.
Can I see where an answer came from?
GorZesk answers from your own catalog and order records, which you can inspect in the dashboard.
Related topics
- AI assistantsWhat is an AI assistant for business?A plain explanation of business AI assistants, what they can do, and where a person should stay in charge.Read
- Trust and safetyAI safety for business chatbotsThe practical risks of AI chatbots (wrong answers, data leaks, prompt injection) and how to reduce them.Read
Last updated 2026-09-25. Something unclear or wrong? Tell us.