Agent Activity and Quality
Agent Activity and Quality helps your team turn real customer interactions into better service. Use it to understand what an agent received and did, confirm whether an action succeeded, find repeated problems, and decide whether instructions, knowledge, connections, or handover behavior should change.
AI Logs
Navigation: Open Misc → AI Logs when the page is available to you.
AI Logs provide the detailed activity trail needed to answer “What happened in this interaction?” Use search and filters to narrow activity by use case, success status, source, failure reason, Knowledge Base retrieval state, or a linked conversation. Production activity comes from customer-facing operation, while Test Chat comes from built-in testing.
Open a row to review the customer input, agent output, timing, status, and available context. Treat prompts and responses as customer data.
Use AI Logs to answer questions such as:
- Did the agent receive the expected customer request?
- Did the activity succeed or fail?
- Was relevant Knowledge Base content available?
- Was the issue seen in testing, production, or both?
- Does the same failure occur across several interactions?
Use the page to investigate customer-visible behavior and operational failures.
QA Feedback
QA Feedback captures a clear human judgment on an agent outcome so teams can recognize good behavior and prioritize specific improvements. It can be opened from Voice Agent cards, recent-call actions, Email Agent reports, or other feedback links. Tabs can separate email and voice feedback.
Use feedback to record a clear judgment and short reason. Focus on observable behavior:
- correct or incorrect information;
- appropriate tone;
- complete confirmation of customer details;
- successful handover;
- unnecessary reply;
- missed or incorrect tool action.
Avoid pasting unnecessary personal customer data into free-text feedback.
Investigate an issue
- Identify the exact agent, channel, and time.
- Open the channel-specific record: conversation, email, or call log.
- Follow its activity link when available.
- Compare the response with the assigned Knowledge Base and agent instructions.
- Correct the source of the issue.
- Reproduce it in a test.
- Monitor a small number of new live interactions.