Normalized claim
Processing time reduction: 85% decrease
85 percent reduction in processing time
EPA accelerates chemical risk assessments with Amazon Bedrock and Amazon Textract
Use case typeIntelligent document processingUpdated Jun 13, 2026
The United States Environmental Protection Agency (EPA) worked with the AWS Generative AI Innovation Center to build proof-of-concepts that improve chemical study evaluation and data evaluation record creation. The solution uses Amazon Bedrock with Claude 3.7 Sonnet, Amazon Textract, Amazon Bedrock Guardrails, Amazon Bedrock Knowledge Bases, Amazon Titan embeddings, Amazon OpenSearch Service, Amazon S3, and Amazon DynamoDB. Scientists review AI-generated outputs, query source documents through a chatbot, and retain full human control over the final scientific determinations.
- Organization
- United States Environmental Protection Agency
- Industry
- Public Sector
- Location
- United States
- Published
- July 2025
Reported outcomes
4 months to seconds
DER creation timeTime & speed
−85%processing time reduction85%accuracy versus manual results−99%DER creation cost reduction
Strategic outcomes
Speed & agilityAccelerated chemical assessmentsNew product / capabilityAutomated evidence-based document reviewBetter decisions & insightImproved cited multi-document searchRisk & complianceMaintained human-controlled scientific determinations
Why do we believe this?Outcome claims, sources, and evidence checks
Normalized claim
DER creation time: 4 months to seconds
tasks that previously took four months now take seconds
Normalized claim
Accuracy versus manual results: 85%
85 percent accuracy rate
Normalized claim
DER creation cost reduction: 99% decrease
reduce cost of the DER creation process by about 99 percent
Why do we believe this deployment?Customer identity, provider attribution, maturity, and source checks
- Customer
- United States Environmental Protection Agency
- Provider
- AWS
- Maturity
- PoC
- Linked source
- AWS Public Sector Blog
The POC reduced DER creation time from about four months to seconds
Customer identity supportedSource describes one deploymentMaturity supported
Primary read
Use case focus
Showing 3 of 3
- 1Intelligent document processing
- 2Scientific document review
- 3Retrieval-augmented search
- EPA scientists manually evaluate thousands of chemical studies against detailed criteria.
- The work is labor-intensive, limited by staffing shortages, and slowed by cumbersome document review and statutory deadlines for FIFRA applications.
- Research papers are uploaded to Amazon S3 and processed with Amazon Textract.
- Amazon Bedrock with Claude 3.7 Sonnet generates responses for standard evaluation criteria.
- Scientists use a chatbot to query the document and confirm evidence, while human-in-the-loop review keeps experts in control.
- Amazon Bedrock Knowledge Bases provide retrieval with citations and multi-document search across the FIFRA dataset.
- The pipeline processes documents in batches to reduce inference cost.
- The EPA reported about an 85 percent reduction in processing time for chemical assessments.
- The POC reduced DER creation time from about four months to seconds.
- The EPA reported about 85 percent accuracy versus manual results.
- The post says the approach could reduce DER creation cost by about 99 percent.
Architecture
Documents are ingested into Amazon S3, extracted with Amazon Textract, and evaluated by Claude 3.7 Sonnet on Amazon Bedrock. Amazon Bedrock Guardrails are used for responsible output control. Amazon Bedrock Knowledge Bases, Amazon Titan embeddings, Amazon OpenSearch Service, and Amazon DynamoDB support retrieval, search, and record storage. Scientists review outputs in a human-in-the-loop workflow and can query source documents through a chatbot. The FIFRA workflow also includes batch processing to reduce inference cost.
Implementation partners1
Sources & evidence1
Evidence: Medium50/100Evidence strength
- Customer explicitly identified
- Deployment status explicitly supported
- Quantified outcome available
- Technical implementation details available
Type: Blog PostPublished: Jul 14, 2025Publisher: AWSEvidence: VendorConfidence: Medium
AI-generated summary. Verify important details with the linked sources before relying on this case.
Explore related AI use cases
Customer evidence
Provider evidence
Market context
AI patterns
Was this useful?
Community
Comments
Loading comments...