When it runs
The review runs after a conversation ends, or after a period of inactivity if the visitor walks away. It works for website chat, browser voice and Telegram, and never makes the visitor wait.
The default reviewer
With no setup, every conversation gets:
- a 0–100 goal score
- an outcome: completed, incomplete, cancelled, follow-up or escalated
- sentiment on a five-point scale from very negative to very positive
- a summary, key points and tags
- action items with low, medium or high priority, plus an owner and due date when the conversation names them
- knowledge gaps: questions the agent couldn't answer from your knowledge base
The reviewer sticks to the transcript and leaves out anything it has no basis for.
A custom reviewer
Pick a sub-agent as the agent's post-conversation analyst. Define the fields you want filled, and give it tools such as a datastore, knowledge search or connected-app actions so it can save a record or follow up. If it returns nothing usable, the default reviewer fills in.
Where results show up
Each conversation shows its review. Scores, sentiment, tags and knowledge-gap counts also roll up into your analytics.
Related docs