Escalations

HumanRail surfaces decisions where the AI isn't confident enough to proceed — distinct from risk-based approval gates.