The application must provide an on-demand workflow to assess bias drift for models deployed to real-time endpoints. Which action meets this requirement?

Invoking an AWS Lambda function to launch a SageMaker Clarify job directly addresses the need for an on-demand workflow to assess bias drift. SageMaker Clarify is specifically designed for bias detection and explainability, offering pre-built metrics and a structured approach to evaluate models. Lambda provides the serverless, on-demand execution environment required to trigger this analysis programmatically from an application. The option to pull the sagemaker-model-monitor-analyzer container is incorrect because this container is primarily for model quality and data quality monitoring, not specifically for bias detection. AWS Glue Data Quality is for data quality checks within ETL pipelines, not for real-time model bias assessment. Manually comparing bias metrics using SageMaker notebooks is not an on-demand, automated workflow suitable for continuous monitoring.