An internal policy assistant searches only the current approved policies that the employee may access. It answers with a short explanation and links to the exact sections used. If sources conflict or do not answer the question, it says so and routes the question to the policy owner.
Academy / What RAG is—and when not to use it
What RAG is—and when not to use it
Understand how AI can answer from your own documents, why source quality matters and when a simpler search is the better choice.
SHORT ANSWER
What RAG is—and when not to use it
RAG retrieves relevant material before an AI generates an answer. Use it when people need natural-language answers from a governed body of documents with citations. Do not use it to hide poor source ownership, replace an exact database lookup or invent answers when no approved source exists.
Common failures are indexing obsolete documents, ignoring permissions, retrieving fragments without enough context, showing no citations and evaluating only fluent answers instead of source support. Better prompts do not repair those data and governance problems.
Collect twenty real questions and the approved source passage for each answer. Include several questions that should remain unanswered. Test retrieval and permissions before testing generated prose.
- Retrieval
- Selecting potentially relevant source material for a question.
- Grounded answer
- An answer whose important statements are supported by the retrieved approved sources.
READY TO REMOVE A BOTTLENECK?
Discuss your workflow →