Use the Review tab to inspect unresolved chats that the AI triage system has flagged for attention.
The system reviews unresolved conversations and classifies them into issue types so your team can decide what action is needed.
When a chat is unresolved, the triage agent analyzes the conversation and places it into one of these categories:
The chatbot did not have the information needed to answer the customer’s question.
Action: Add or improve source knowledge.
The answer exists in the knowledge base, but the chatbot only found it after extensive searching.
Action: Improve the relevant source content so it is easier to retrieve.
The chatbot found the right information, but gave an incomplete, unclear, or poor response.
Action: Review the prompt, response behavior, or instructions.
The chatbot found multiple pieces of knowledge that contradicted each other.
Action: Update or remove outdated/conflicting knowledge.
Do not create corrections for every issue. Too many corrections can bloat the knowledge base and confuse the chatbot over time.
Before taking action, check whether the issue is truly caused by missing or incorrect knowledge.
The Review inbox should stay high-signal. Avoid adding unnecessary fixes for chats that were not real issues.
A good Review workflow should ensure that:
Real knowledge gaps are correctly identified
False knowledge gaps are avoided
The inbox does not become noisy or bloated
Each reviewed item leads to the right resolution path
Corrections are only created when truly necessary
Last updated
Was this helpful?
Was this helpful?
