Build a Live Supervisor Dashboard for AI Support Calls

Description and use

A Telnyx Edge Compute application lets supervisors watch AI support calls live, inject coaching nudges, and join the conversation via WebRTC, eliminating post‑call errors.

Problem

Voice AI quality checks happen post‑call, allowing errors to reach customers before review.

Outcome

The live supervisor dashboard enables real‑time coaching, policy enforcement, and seamless human escalation, reducing post‑call errors and improving customer experience.

Workflow

• 1. Configure assistant to use a WebSocket endpoint for event streams.
• 2. Deploy the Live Support Coach Room on Telnyx Edge Compute.
• 3. Run the caller simulator to generate conversation events.
• 4. Supervisor opens the dashboard to see active rooms and live transcripts.
• 5. Policy engine evaluates events and injects coaching nudges via conversation.item.create frames.
• 6. Supervisor clicks "Escalate to supervisor" to initiate an outbound call and join the AI conversation with a WebRTC softphone.
• 7. Supervisor monitors and participates in the call in real time.

Evidence

Most voice AI quality checks happen after the call ends. A transcript is stored, someone reviews a small sample later, and any mistake has already reached the customer.

Source

The application uses two browser tabs: The caller simulator creates a conversation and submits caller or assistant turns. The supervisor dashboard lists active rooms, streams the transcript, shows policy flags, and exposes an escalation button.

Source

When the supervisor clicks Escalate to supervisor, the room creates an outbound call to the configured supervisor device, joins that call leg to the active AI conversation with ai_assistant_join, and answers the call in the browser through the Telnyx WebRTC SDK.

Source