AWS Bedrock Data Automation: Intelligent Document Processing

Amazon Web Services (AWS) has launched Amazon Bedrock Data Automation, a feature designed for intelligent document processing (IDP) at scale. This managed service leverages generative AI to extract information from various unstructured data sources, including documents, images, audio, and video, without requiring extensive data annotation or model training. Key benefits include simplified workflows, improved accuracy, and reduced complexity compared to traditional named entity recognition (NER) methods. Bedrock Data Automation excels in handling diverse data types like numeric scores and free-form text, capabilities lacking in classic NER. The service automates document parsing, context management, and model selection, enabling developers to focus on business logic rather than IDP implementation details. It's particularly useful for creating product feature tables, extracting metadata, and analyzing legal documents, customer reviews, and news articles. For organizations needing more customization, the solution offers alternative paths using self-hosted foundation models (FMs) or Amazon Textract for OCR, although this requires more operational overhead. The solution, deployable via AWS Cloud Development Kit (CDK), integrates with other AWS services like Step Functions, Lambda, S3, and ECS to create a scalable and robust IDP pipeline. A user-friendly web application simplifies interaction, allowing users to upload documents, define attributes for extraction, and select parsing modes (Amazon Bedrock Data Automation, Textract, or custom Bedrock FMs). Cost analysis shows that while using custom Bedrock FMs can be cost-effective, Bedrock Data Automation offers a superior balance of accuracy, ease of use, and scalability for most use cases. Current limitations include document size restrictions (up to 20 pages for custom attributes with Bedrock Data Automation as of June 2025) and the need to split larger documents for processing. Future improvements will focus on expanding model support, increasing document size limits, and enhancing extraction accuracy.

(Source: https://aws.amazon.com/blogs/machine-learning/intelligent-document-processing-at-scale-with-generative-ai-and-amazon-bedrock-data-automation/)