Reviewing AI-drafted KB articles

How drafts are created

When the chatbot answers a buyer's question using past resolved tickets (because no existing KB article covered it), it automatically drafts a candidate KB article and queues it for your review. The draft includes:

• A suggested title and slug
• An excerpt summarising the topic
• The full answer the assistant gave (which was grounded in real ticket resolutions)
• A link back to the chatbot conversation that triggered it

Reviewing drafts

1. Go to Admin → KB Gaps.
2. Filter by Origin: Chatbot drafts to see only the AI-drafted candidates.
3. Open a draft to read the suggested article in full.
4. Click Publish to push it into your Knowledgebase, or Dismiss if it's not a real gap.

Published articles enter the Documentation table as drafts so you can edit before going live — no AI content goes public without your approval.

The virtuous cycle

This is the heart of the deflection model: a buyer asks a question, the AI finds the answer in past tickets, the answer is reviewed and published, and the next buyer with the same question gets the answer instantly from the KB without needing to open a ticket at all.