Improving AI Response Quality
Improving quality is a simple loop: find gaps, update Responses, then re-test.
1. Find the biggest gaps
Use these signals together to identify what needs work:
- Inbox: review escalations or unclear responses.
- Audit: look for low-scoring metrics by topic or intent.
- Test Coverage: check failed or missing scenarios.
If you see repeated questions without strong answers, that is a good candidate for a new Response or a more detailed workflow.
2. Update Responses
Open Connectors -> Knowledge and update the Response or add a new one.
Tips for higher quality:
- Keep responses short and factual.
- Add Guidance for tone or formatting without rewriting content.
- Use Exact responses for language that must not change.
3. Use guidance at the right level
| Where to set it | Where to find it | Scope |
|---|---|---|
| Agent Guidance | Agents -> Settings -> Other | Applies to all responses |
| Response Guidance | Connectors -> Knowledge | Applies only to that response |
Use agent-level guidance for global rules and response-level guidance for intent-specific instructions.
4. Validate and deploy
- Re-run scenarios in Test Coverage.
- Audit a fresh sample after changes go live.
- Track improvements in Analytics.
If results do not improve, revise the response, adjust guidance, or add a workflow step for additional data.
Next steps
Last updated on