How can I see what my AI agent did on a call or workflow run, and why a step failed?

Go to Activity, where everything historical lives: calls, messages with their delivery status, meetings, and balance activity for billing. Each agent also has its own activity section.

A workflow execution is called a run. Open a run to see its status (for example step failed), go through it step by step on a timeline, and see exactly which step failed and why, along with the full execution log. For example, a call forwarding step fails if no phone number was set for it, and an SMS step fails if there is no phone number to send to.

Phone calls are listed under Calls. Open a call to play the recording and read the transcript. The recording shows not only what was said but also what the AI was doing and why: you can see its tool calls, so if the agent hung up or did not follow its prompt, you can see the reason.

Web widget sessions are listed separately from phone calls, under the agent's web sessions, because the widget can take typing and voice. A test you run in the builder in chat mode is listed there as a session with the mode Test. Open the session to see the whole conversation.

Certain kinds of failed steps can be retried from the run. A phone call itself cannot be retried.

You can group the data to drill down, for example by direction, month, status or your own grouping. The same grouping works across reporting, such as your balance details grouped by service, or by date and then service.

If you need help from our team, send us the link to the recording (copy the URL from your browser address bar) so we can look at it with you.