Conversations

Conversations are the evidence behind dashboard metrics. Use them to understand what the customer heard, how the agent responded, and why the call reached its final outcome.

The list combines supported channels in one operational view. Filters help narrow the dataset before opening a record, and exports use the currently filtered set rather than every conversation in the workspace.

Review a conversation

1. Open Conversations and search by contact, phone number, use case, or agent.
2. Filter by channel, status, use case, or sort order. Supported channels include inbound, outbound, calls, Echo, and WhatsApp.
3. Select a conversation to open the Conversation and Details tabs.
4. Review messages or transcript content alongside summary and conclusion data.
5. Inspect context and session metadata when debugging Echo or messaging conversations.
6. Export the currently filtered conversation set when it needs offline analysis.

Conversation details

Available detail depends on the channel and whether the interaction completed successfully. A record can include participant information, timestamps, status, transcript or messages, generated summary, conclusion, recording, use-case and agent references, and channel-specific metadata. A missing recording does not necessarily mean the conversation record failed; recording availability depends on the call path and processing state.

For Echo sessions, context can include the page and referrer, device information, session identifiers, and attachments provided during the conversation. For calling workflows, lead and outcome data help connect the transcript to the originating use case.

What to look for

• Repeated questions may indicate missing knowledge or unclear instructions.
• Very short calls can reveal greeting, latency, or phone-number issues.
• Incorrect customer details usually point to lead mapping or variable problems.
• Correct conversations with missing downstream updates indicate an integration or webhook issue.
• Echo sessions include page URL, referrer, device context, attachments, and visitor/session identifiers when available.

Diagnose systematically

Start with the final status and timeline, then read the transcript around the first unexpected turn. Compare the agent response with its instructions and retrieved knowledge. If the conversation is correct but an external record is missing, inspect the relevant integration or webhook delivery rather than changing the prompt.

Customer data: Treat transcripts and recordings as sensitive customer data. Apply your organization’s retention and access policies before sharing them.

Exports can contain the same sensitive information as the application view. Store them only in approved systems and remove them when the investigation or reporting task is complete.

Open Conversations to inspect recent activity.