A forms-processing application on EC2 and RDS stores uploaded scanned forms in S3 and sends notifications via Amazon SNS for a human team to validate and extract data, then call an external API. You must automate form processing with accurate extraction, minimal time to market, and minimal ongoing operational overhead. Which solution best meets these goals?

The correct solution leverages Amazon Textract and Amazon Comprehend for automated, accurate data extraction with minimal time to market and operational overhead. Textract is a fully managed service for OCR and data extraction from documents, and Comprehend can analyze the extracted text for insights. AWS Step Functions orchestrates the workflow, and AWS Lambda provides serverless compute, minimizing operational burden. Storing output in S3 is cost-effective. Building custom OCR libraries on EKS or EC2 instances (Option 1) increases development time, complexity, and operational overhead, and may not achieve the required accuracy without significant effort. Training and hosting custom AI/ML models on EC2 (Option 2) or SageMaker (Option 3) also introduces significant development time, expertise requirements, and operational overhead compared to using fully managed, pre-trained services like Textract and Comprehend. ElastiCache (Option 3) is a caching service, not ideal for persistent storage of extracted data.