Significant (Outcome Severity)
Errors cause meaningful, recoverable harm
Significant is an outcome severity level: errors cause meaningful, recoverable harm — real cost, but correctable without the irreversibility of catastrophic stakes.
Examples: fraud declines on legitimate transactions, wrongful content moderation, material but reversible financial errors, hiring decisions with appeal paths.
Most high-volume enterprise automation (P fraud scoring, content moderation) lives in significant cells — requiring calibrated models, staffed escalation paths, drift monitoring, and periodic audit rather than naked automation.
Misclassifying catastrophic work as significant to reduce friction is a governance failure — same family as understating severity in task classification (When the Commitment Boundary Needs Reinforcing).