Monitoring Agent Conversations and Escalations

Review agent conversations, handle escalations, identify unanswered questions, and feed gaps back into your knowledge base to continuously improve your agents.

This guide is for admins and owners who have deployed AI agents and want to monitor their performance. You will learn how to review conversations, handle escalations, identify unanswered questions, and close knowledge gaps so your agents improve over time.

What you will accomplish: Review agent conversations, handle escalations, identify unanswered questions, and feed gaps back into your knowledge base to continuously improve your agents.

1

Open Your Command Center

Operations Overview is your command center for all deployed agents. It shows conversation review and unanswered questions in one place; escalation handling itself happens in the Escalations tab in Operations (Step 4). Navigate to Operations Overview from the Workspace panel: Agents column, then Operations Overview.

2

Check Key Metrics at a Glance

The Overview tab gives you a high-level view of how your agents are performing across all deployments. Review conversation volume to see how many users are interacting with your agents.

Check average response time and satisfaction metrics to gauge quality. Monitor the escalation rate. A high rate means your knowledge base has gaps that need attention.

3

Review Agent Conversations

Switch to the Inbox tab to see recent agent interactions. Filter by agent, date range, or status to find specific conversations.

Click any conversation to read the full thread and see which sources the agent cited.

4

Resolve Escalated Conversations

Escalations are conversations where the agent could not answer or the user expressed dissatisfaction. Each escalation is an opportunity to improve. Escalation handling lives in the Escalations tab in Operations, a dedicated three-pane view separate from this Operations Overview.

In Escalations, claim an unclaimed escalation, review the conversation context, and respond directly to the user, or create a knowledge base document to address the gap. Resolve the escalation once it has been addressed.

5

Identify Knowledge Gaps

Customer Gaps surfaces questions your agents could not answer. These are potential knowledge gaps that, once filled, make your agents smarter. Open RequestsCustomer Gaps to see questions your agents could not answer.

Click any gap to create a gap draft or resource request to address it. Prioritize high-frequency unanswered questions first. These affect the most users and deliver the biggest improvement when resolved.

6

Close the Loop

When you identify a pattern of unanswered questions, create content to address them. As the knowledge base grows, agent quality improves automatically. Group similar unanswered questions to find recurring themes.

Create or upload documents that address these themes. Each new document teaches your agents to handle those questions. Repeat this cycle regularly. Consistent improvement compounds over time.

Pro Tip: Set up a weekly 15-minute "agent health check." Review the top 5 unanswered questions, create one piece of content, and watch your agent's success rate climb. Small, consistent improvements compound.

Now that you are monitoring your agents and closing gaps, the next step is to use analytics to measure your overall knowledge quality and identify areas for targeted improvement.

Was this helpful?