28Can we use the model's own confidence to auto-approve invoice extractions and send only the uncertain ones to a human?▼hardNewOpenAIGoogle◆ premiumTurn confidence scores into a defensible approval policy. Work through selective accuracy, uncertainty, correlated errors and review capacity before deciding which invoice fields can bypass a human.Practise this question →Read answer →
24Our fraud model has a 0.96 ROC-AUC, but reviewers cannot handle its alerts. Would you launch it?▼hardNewGoogleMicrosoftAmazon◆ premiumChoose an operating threshold using prevalence, error costs and review capacity. A good ranking score can still produce an unusable queue.Practise this question →Read answer →