Copilot Use Cases

Build an Agent

Watch the Video Tutorial

• Upload a workflow document (a lending qualification flow in this example) to Regal Copilot, and it drafts a first version of the agent with a summary of what it built, which you can inspect in the editable canvas.
• Run a test call to check whether the agent takes actions at the right moments, then either edit manually or ask Copilot to fix issues, like moving the set caller intent step later in the conversation, and compare versions to confirm the change.
• Keep repeating the test, review, and revise loop until the agent behaves the way you want, then use test and simulation (covered in a separate tutorial) to validate at scale and catch regressions.

Generate Test Cases & Run Simulations

Watch the Video Tutorial

• In the agent's Test Cases tab, ask Copilot to generate end-to-end test cases (like the quote intake assistant in the demo), each one a full realistic call path with success criteria you can review before running anything.
• Run simulations either from the test case tab or by asking Copilot to do it in-session, then check which scenarios passed and which failed. In the demo, existing policy calls and the full quote flow passed, while reaching the wrong person and transferring to a specialist struggled.
• Ask Copilot to analyze the failures and suggest prompt fixes, have it apply them to your draft, and rerun until edge cases hold up, all without manual testing.

Improve Agents from Call Transcripts

Watch the Video Tutorial

• Open Copilot from the agent and either target one issue in a specific transcript (like reducing repetition, where the agent stopped explaining the monthly cleaning offer twice and now asks a short confirmation question instead) or ask for a broader health check across the last 5 or 10 calls, or the last month.
• For the targeted fix, review Copilot's summary of what changed and open the Canvas UI to inspect the updated logic and test the audio. For the broad review, Copilot surfaces patterns across calls (in the demo, a priority issue with how the agent phrases something to customers) so you can decide what to fix first.
• Ask Copilot to implement the top-priority fix, click apply this edit, then review the change summary and the draft agent and test it before finalizing.

Review Observability Metrics

Watch the Video Tutorial

• Open the observability dashboard and ask Copilot for the exact cut you need when it isn't shown, like the last 14 days grouped by agent, and it builds a dynamic view with metrics such as call volume and end-to-end latency by agent. You can keep changing the date range or grouping to narrow in on outliers.
• Treat that overview as a starting point. When something looks off, like latency spikes, contact interruptions, or a high repetition rate under an LLM issue, ask Copilot which calls were affected and it returns a list from a recent window with links to the individual transcripts.
• Open the transcripts to understand the pattern, and ask Copilot to pinpoint where in the call an issue like repetition happens so you can debug faster and decide whether the agent's behavior needs to change.

Use cases and good-to-knows