The correct answer is B because Amazon Augmented AI (Amazon A2I) allows developers to integrate human review workflows into ML systems. It supports defining confidence thresholds, such that only low-confidence predictions are sent to human reviewers.
From AWS documentation:
" Amazon A2I provides built-in human review workflows for ML predictions. You can configure confidence thresholds to determine when human review is triggered, enabling continual adjustment based on accuracy needs. "
This supports use cases where business decisions (like financial approvals) require manual oversight for edge cases.
Explanation of other options:
A. Amazon Personalize is used for recommendation systems, not human review.
C. Amazon Inspector is for security assessment and vulnerability scanning.
D. AWS Audit Manager is used for compliance audits, not ML prediction reviews.
Referenced AWS AI/ML Documents and Study Guides:
Amazon A2I Developer Guide – Confidence Threshold Configuration
AWS ML Specialty Guide – Human-in-the-Loop Systems
AWS Responsible AI Documentation – Review and Escalation Workflows
