Most AI agent deployments are black boxes. You know conversations are happening. You can see message counts on a dashboard. But if a customer says "your bot gave me wrong information," you have no way to pull up that conversation and see what actually happened.

The Conversation History Viewer gives you that visibility.

What it shows

Every session your agent has had is stored and searchable. Each session entry shows date and time, number of messages exchanged, whether the conversation ended in escalation or resolution, the user's feedback rating, and any messages the agent flagged as low-confidence.

Open a session and you see the full transcript: every user message, every agent response, the confidence score on each AI reply, and which knowledge base sources were cited.

Filtering to what matters

By outcome. Filter to escalated sessions only — these are the conversations where the agent failed to resolve the issue.

By feedback. Filter to thumbs-down sessions. These are users who explicitly told you something was wrong. This is the highest-signal filter.

By unanswered. Filter to sessions containing at least one low-confidence response.

By keyword. Search for a specific term to see every conversation where that topic came up.

Handling customer complaints

When a customer contacts support to say the bot gave them bad information, find that session by date range or feedback and see exactly what happened. If the agent was wrong, use the session to identify the knowledge base gap and fix it immediately.

Customers who complain and see a fast, specific response — "we found the issue, updated our knowledge base, here's the correct answer" — are often more satisfied than customers who never had the problem.