Every required disclosure and prohibited phrase, checked against not just whether it happened but when.
Compliance risk lives in the edge cases: escalations, improvised promises on a sales call, support agents under pressure. Sampling misses them and a keyword check can't tell if a disclosure came before or after the moment it mattered.
A late disclosure isn't the same as an on-time one. Compass checks both whether it happened and when.
Every flag carries a confidence score. Low-confidence cases route to human review, not a false alarm.
Escalations, improvised moments, and exceptions get the same scrutiny as routine interactions.
Every flag and every clear pass traces to a specific moment in the conversation. When someone asks how you know, there's an answer.
AI agents get checked against the same disclosures and timing rules as human agents.
Required disclosures, identity checks, escalation triggers, pricing commitments, or start from a built-in policy.
When a requirement isn't met, Compass shows the specific moment in a conversation and what should have happened, not a vague flag.
One missed disclosure is a coaching conversation. The same miss across half your team is a process problem. Compass surfaces both.
Fire rate trends show whether a gap is a one-off or becoming a habit.
Each required behavior is tracked as its own signal, not a single pass/fail check.
Outcome correlation shows how much a signal matters, with the data behind it.
