Self-learn review
Run a weekly review of AI-suggested knowledge base improvements and approve the best ones in minutes.
The review queue
After every human handoff, DGbot analyses the conversation and proposes knowledge base improvements. These appear in the Self-learn section as suggestion cards.
Open Self-learn in adminThe Self-learn review queue. Each card shows the original question, the agent's answer, and the AI's proposed FAQ entry.
Each card shows:
- The original visitor question — what the visitor asked that the bot couldn't answer
- The agent's response — what the human agent actually said to resolve it
- Proposed FAQ entry — the AI's suggestion for a new Q&A pair to add to the knowledge base
How to review
Click View conversation on any card to read the full exchange. Context helps you judge whether the suggestion is accurate and worth adding.
Ask yourself:
- Is the proposed answer accurate?
- Would this answer have resolved the original conversation?
- Is there any sensitive or confidential information in the answer?
- Is the question phrased how visitors actually ask?
- Approve — add the Q&A to your knowledge base immediately
- Edit — refine the wording before approving (click the pencil icon)
- Reject — discard the suggestion (it's removed permanently)
Approve, edit, or reject each suggestion. Approved suggestions are added to a dedicated Self-learn FAQ source.
Weekly routine
A 15-minute weekly review is all it takes to keep the self-learn queue clear and your knowledge base improving continuously.
Suggested routine:
1. Open Self-learn (takes 30 seconds)
2. Skim all cards — reject obviously wrong or confidential ones immediately
3. Approve the clear wins without editing — most suggestions are ready to use
4. Edit and approve the ones that need minor wording tweaks
5. Review a batch of conversations directly if the queue is thin this week
Bots that are reviewed weekly typically reach 80%+ resolution rates within 60 days. Bots that are never reviewed plateau at whatever quality they had at launch.