# Does one frustrated message trigger an alert?

No: a single frustrated message writes a finding, not an alert, and only survives to a case if triage rules it a real, recurring problem rather than one person having a bad moment.

Crossing the threshold on one message writes a finding, a record that one trace tripped the classifier. Nothing pages anyone yet. Triage looks at findings before anyone is notified, separating a real pattern from one bad moment, which happens on the best agent in the world.

Where frustration earns attention is in aggregate: when findings pile up across the same call site, that's a spike, not a coincidence, and it becomes one case instead of a pile of individual flags nobody can act on. A case is what carries attribution and, if it clears the alerting rules, reaches a human.

The upside of scoring every trace instead of a sample is that the spike is visible from the first few messages, not after enough people have given up and gone quiet.

---

Source: https://tessary.ai/answers/tessary-frustration/does-one-frustrated-message-trigger-an-alert
More on Tessary frustration: https://tessary.ai/answers/tessary-frustration
From Tessary, agent reliability for AI agents in production: https://tessary.ai