Consent is architecture, not a checkbox

Every engagement platform claims to respect opt-outs. The question worth asking is where the enforcement lives. If it lives in each automation's configuration — a suppression list here, a filter there — it is one misconfigured journey away from a TCPA incident and a very uncomfortable compliance meeting.

The alternative is structural: a consent engine that every outbound message must call synchronously at delivery time, whose answer is final. Channel × purpose matrix (reminders are not marketing), quiet hours as deferral, frequency caps across every module, and hard blocks that nothing can override — not a journey, not an AI agent, not a manual campaign send.

Two properties matter as much as the gate itself. Denials must explain themselves — "blocked: opt-out sms/marketing, recorded Mar 12 via STOP keyword" — because staff who understand the refusal stop trying to work around it. And every evaluation must be logged with the policy version in force, because "why did this send in March" deserves an auditable answer.

When the gate is visible — when a marketer sees the eligible count drop with per-rule exclusions before hitting send — the anxiety disappears from both directions. Marketing sends confidently; compliance sleeps. That is what "consent-safe" means when it is architecture.