vocabulary: label: Amazon Augmented AI Vocabulary description: Core terms, actions, and workflows for the Amazon Augmented AI human-in-the-loop review service. tags: - Amazon Augmented AI - Human In The Loop - Machine Learning - AI Review - AWS created: "2026-04-19" modified: "2026-04-19" resources: - name: HumanLoop description: A human review task created when an ML prediction requires human judgment, associated with a flow definition. properties: - HumanLoopName - HumanLoopArn - HumanLoopStatus - FlowDefinitionArn - CreationTime - HumanLoopOutput related: - FlowDefinition - HumanLoopInput - HumanLoopOutput - name: HumanLoopInput description: The input data provided to human reviewers in a human loop, serialized as a JSON string. properties: - InputContent related: - HumanLoop - name: HumanLoopOutput description: The S3 location where human reviewer responses are stored after a loop completes. properties: - OutputS3Uri related: - HumanLoop - name: DataAttributes description: Attributes that specify the nature of the content in a human loop to ensure appropriate handling. properties: - ContentClassifiers related: - HumanLoop actions: - name: StartHumanLoop description: Start a human review loop for an ML prediction requiring human judgment. resource: HumanLoop method: POST path: /human-loops - name: ListHumanLoops description: List all human review loops for a specific flow definition. resource: HumanLoop method: GET path: /human-loops - name: DescribeHumanLoop description: Get the status and output details of a specific human review loop. resource: HumanLoop method: GET path: /human-loops/{HumanLoopName} - name: DeleteHumanLoop description: Delete a completed or stopped human review loop. resource: HumanLoop method: DELETE path: /human-loops/{HumanLoopName} - name: StopHumanLoop description: Stop an in-progress human review loop. resource: HumanLoop method: POST path: /human-loops/stop workflows: - name: HumanReviewML description: End-to-end workflow for routing ML predictions to human reviewers using Amazon A2I. steps: - action: StartHumanLoop description: Route the ML prediction to human reviewers - action: DescribeHumanLoop description: Poll the human loop status until completion - action: ListHumanLoops description: Review completed human loops and retrieve results personas: - name: ML Engineer description: Integrates human review into ML pipelines to improve model accuracy and handle edge cases. actions: - StartHumanLoop - ListHumanLoops - DescribeHumanLoop - StopHumanLoop - DeleteHumanLoop - name: Operations Manager description: Monitors the human review queue, manages reviewer capacity, and tracks SLA compliance. actions: - ListHumanLoops - DescribeHumanLoop - StopHumanLoop