A voice agent QA report that lists what to fix
Hundreds of test calls come back as a few findings, each with the lines that prove it.
One finding
This is taken from the sample report, which uses a made-up insurer and made-up calls.
Read the whole sample reportHigh Gave account details before security passed, 5 calls
01:04AgentAnd can you confirm the first line of your address?
01:08CallerEr, 22 Station Road?
01:12AgentThanks. Your claim for the burst pipe is open.
Judge: The address on file is 14 Mill Lane. Claim details were read out after a failed answer.
Fix: Hold every account detail until all security checks pass.
What every report holds
Read the score first. Open the quotes when you need them.
- 1
The score
Calls passed against your pass mark, and reply times: the median, and the slowest 1 in 20 (p95).
- 2
The callers
Which kinds of caller failed most, so you know where the agent is weakest.
- 3
The findings
Failed calls that share a cause, grouped into one finding and ranked by risk.
- 4
The evidence
For each finding, the quoted lines, the time in the call and the judge’s reason.
- 5
The fix
A suggested change for each finding, plain enough to hand to whoever owns the prompt.
- 6
The regression set
The failed calls saved, so the next run checks the fix and keeps checking it.
Run a test on your own agent.

- test calls
- 506
- passed
- 96.8%
- median reply
- 0.94 s
Holds the new day when a caller changes their mind.