Monitor real calls
Score every real call against plain-language checks, group failures into issues, and get alerted.
Monitoring tells you how an agent does with real callers. After each real call ends, an AI judge scores it against a list of checks. Failed calls are grouped into issues, and alerts tell you when something gets worse.
Test calls, eval runs, and dashboard tests are never counted. Their results are under Evaluation.
Turn on checks
Every agent starts with five basic checks: no made-up facts, follows its rules, answers what the caller asked, no repeating loops, and ends the call properly. They are always on.
Add goal checks for what this agent is for:
- Open the agent and select Monitoring, then Checks.
- Click Draft from instructions. Speko writes 4 to 6 checks from the agent's instructions and workflow.
- Review them and click Turn on all, or turn on the ones you want.
To add your own, type it in plain words, for example Never promises a start date, and click Add check. It is on right away.
Speko drafts checks again when you deploy changed instructions. New drafts wait for you as Not on yet.
Read call results
Each scored call gets one result:
| Result | Meaning |
|---|---|
| Success | Every check that applied passed |
| Handled correctly | Voicemail or no answer, handled as expected |
| Agent mistake | At least one check failed |
| System error | The call broke: the agent never spoke, went silent for 8 s or more, dropped, or lost audio |
A check the judge cannot decide is marked Did not apply and never fails a call.
The Monitoring page shows the success rate, response time, and counts for the last 24 hours, 7 days, or 30 days. Open Calls to filter by result or search, and open any call for its recording, transcript, and the verdict on each check.
Work through issues
Failed calls with the same cause are grouped into one issue, such as one failing check or one kind of system error. Each issue shows how many calls it affected, when it started, and whether it is Rising, Falling, Steady, or New.
Open an issue to read the judge's latest reason and the calls behind it. Click Make test case to turn a failing call into a test case under Evaluation, so the fix can be tested before you deploy it.
If a call was really a test, open it and click Mark as test call to remove it from the numbers.
Get alerted
- Open Monitoring → Alerts.
- Under Send to, add a Slack channel, an email address, or a webhook URL.
- Turn on a ready-made alert.
| Alert | Fires when |
|---|---|
| Success rate drops below 60% | Average over the last 20 judged real calls, once at least 10 are judged |
| A call has a system error | Any real call ends in a system error |
| A new issue appears | A failure type not seen in the last 30 days shows up |
| Responses get slow | 95th-percentile response time goes above 3 s |
Each alert fires at most once every 10 minutes. For Slack, invite the Speko Slack app to the channel first. Webhook alerts are signed with the secret shown on the page.