TLDR: The U.S. Environmental Protection Agency (EPA), in collaboration with the AWS Generative AI Innovation Center, is leveraging generative AI to dramatically accelerate its document processing workflows. New proof-of-concepts presented at the AWS Summit Washington, DC 2025 demonstrate potential reductions in processing time by up to 99% and significant cost savings, enhancing the agency’s mission to protect human health and the environment.
The United States Environmental Protection Agency (EPA) is making significant strides in modernizing its operations by integrating generative artificial intelligence (AI) into its document processing workflows. This initiative aligns with the EPA’s broader ‘Powering the Great American Comeback’ program, which aims to position the U.S. as a global leader in AI innovation.
At the AWS Summit Washington, DC 2025, scientists from the AWS Generative AI Innovation Center (GenAIIC) and the EPA unveiled two groundbreaking proof-of-concepts (POCs). These demonstrations showcased how generative AI can drastically reduce manual effort and empower EPA scientists in their critical work of evaluating studies related to pesticides and chemicals, ultimately improving time to market and saving taxpayer dollars.
The EPA’s Office of Chemical Safety and Pollution Prevention (OCSPP) and Office of Pesticide Programs (OPP) are at the forefront of this technological adoption. The intelligent document processing (IDP) solutions, powered by Amazon Web Services (AWS), have yielded impressive results:
Dramatic Time Reduction: One POC demonstrated a potential 99 percent reduction in processing time for certain tasks, transforming processes that previously took four months into mere seconds. Another general improvement showed an 85 percent reduction in processing time, allowing tasks that once took months to be completed in hours or days.
High Accuracy: The AI-generated evaluations have achieved high accuracy rates. For low- and medium-complexity document batches, the system maintained high accuracy, while another application achieved an 85 percent accuracy rate when compared to manually derived results.
Significant Cost Savings: The process is highly cost-effective. Schoeff, an unnamed source in the original report, indicated that the document evaluation report (DER) creation process could see a cost reduction of approximately 99 percent. Furthermore, processing 250 research documents with nine evaluation prompts costs roughly $40 in Amazon Bedrock usage, which is less than one hour of staff time.
The EPA emphasizes its ‘AI for science’ approach, ensuring that AI tools assist staff rather than making policy decisions. To maintain scientific integrity, the AWS GenAIIC and EPA have implemented key principles:
Transparency: The system provides clear confidence scores for AI-generated content, such as text extraction, and explicitly highlights responses that have undergone human review.
Verification Tools: EPA scientists are equipped with a chatbot integrated directly with the documents, allowing them to ask questions and verify generated responses, ensuring the reliability of the information.
Also Read:
- Pega and AWS Announce Five-Year Strategic Partnership to Accelerate Enterprise Legacy System Modernization with Generative AI
- Amazon S3 Undergoes Major AI Transformation to Power Next-Generation Data Infrastructure
This collaborative effort, spearheaded by individuals such as Dan Ford, Sandra El Ashry, Brian Zambrano, Steve Goley, Natasha Tchir, Newel Hirst, and Sameer Mohamed, marks a pivotal moment in the EPA’s journey towards a more efficient and technologically advanced future in environmental protection.
