AWS launches Amazon Bedrock Data Automation for financial document processing with custom blueprint system

Amazon Web Services released Amazon Bedrock Data Automation (BDA), a foundation model-powered service for extracting structured data from financial documents including bank statements, tax forms, and contracts.

The service processes documents through custom blueprints—configuration templates that define which data fields to extract, validation rules, and output structure. AWS positions BDA as an alternative to both traditional OCR software and direct foundation model use, claiming lower costs and higher accuracy than using models like Anthropic's Claude for document extraction.

Technical architecture

Amazon Bedrock Data Automation uses foundation models to understand document context and extract structured data. The system includes:

• Custom blueprint configuration for defining extraction patterns
• Visual grounding with confidence scores for explainability
• Built-in hallucination mitigation
• Output in JSON, CSV, and raw data formats

Blueprints specify the document type, data fields to extract, validation rules, and output format. Organizations can use catalog blueprints for common document types or create custom blueprints for specific workflow requirements.

Document processing capabilities

AWS demonstrated BDA on four financial document types:

Bank statements: Extracts transaction data including dates, amounts, descriptions, and reference numbers across multiple pages. The system handles varying formats and feeds data into automated accounting workflows.

W-2 tax forms: Processes standardized tax forms by extracting employer information, employee details, federal tax data, state tax tables, and code-amount pairs. Custom blueprints can group fields into logical structures aligned with downstream tax processing systems.

1099-B forms: Handles investment income reporting documents, which lack built-in blueprints and require custom configuration.

Vendor contracts: Processes unstructured contract documents with custom extraction patterns.

Pricing and availability

Pricing details were not disclosed in the announcement. AWS requires users to have an active AWS account with appropriate IAM permissions and must request model access through the AWS console. The service is available through the Amazon Bedrock console.

Implementation requirements

Organizations need:

• Active AWS account with specific IAM permissions
• Model access approval through AWS console
• Sample financial documents for testing blueprints

For documents with varying formats, AWS notes that multiple custom blueprints may be required. When using the same blueprint on documents with different data, BDA may return slightly different output structures, though the JSON format allows for rule-based downstream processing.

What this means

Amazon Bedrock Data Automation targets the gap between traditional OCR systems that struggle with document variety and direct foundation model use that may be costly for high-volume processing. The custom blueprint system allows organizations to standardize extraction patterns for repetitive document types, potentially reducing manual data entry in financial workflows. However, AWS has not disclosed pricing or provided independent benchmark comparisons to validate cost and accuracy claims against alternatives like Claude or other document AI services.