Co-Pilot

Co-Pilot lets you ask questions about your call and test data in plain English and get back the calls and runs that match — no filters to configure or query language to learn.

How it works

As calls and scenario runs are analyzed, RubricHQ indexes their transcripts and metrics into a searchable knowledge base. Co-Pilot searches that index semantically — by meaning, not just keywords — so you can describe what you’re looking for the way you’d say it out loud.

Ask a question like:

• “Which calls failed identity verification?”
• “Show me runs where the agent talked over the caller.”
• “Find calls where the customer asked to cancel.”
• “Where did the payment not get completed?”

Co-Pilot returns the most relevant calls or runs, ranked by how closely they match.

Asking a question

Open Observability → Co-Pilot and type your question. You can narrow the search with:

• Source — search call observations (ingested or production calls) or scenario runs (your test batches).
• Agent — limit results to a specific agent.
• Limit — how many results to return.

Suggested questions

Co-Pilot surfaces suggested questions generated from your own data — starting points that highlight patterns worth investigating. They also appear on your dashboard, so insights surface on their own rather than waiting to be asked.

Saved queries

Questions you run often can be saved and re-run with one click, so a recurring investigation — for example, “calls that ended in a transfer this week” — is always one click away.