How RFI Know generates answers
Where suggested answers come from and how to judge them
Three sources, in order of trust
1. Your past answers. When a new question closely matches one you've answered before — in a completed questionnaire you imported, or in previous work on the platform — RFI Know reuses that answer. This keeps your responses consistent across every questionnaire you send out.
2. Your documents. For questions without a close historical match, the AI drafts an answer grounded in your uploaded policies and procedures, and cites the specific documents and passages it relied on so you can verify every claim.
3. Nothing else. If your knowledge base has no relevant support for a question, RFI Know says so — a relevance gate filters out weak matches rather than letting the AI guess. Your data is never mixed with other customers', and answers are never generated from another organization's content.
Confidence scores
Every generated answer carries a confidence score:
- 70%+ — strong support from your knowledge base; usually needs only a quick read.
- 40–69% — partial support; review and adjust.
- Below 50% — flagged with a low-confidence warning. Limited knowledge-base support was found, so treat the draft as a starting point, not an answer.
It improves as you use it
Every answer you accept or write becomes part of your organization's answer library. The second questionnaire is faster than the first, and the tenth is mostly review.