Savannah River Site Harnesses AI to Boost Efficiency in Liquid Waste Cleanup

Engineer Tannar Pettigrew uses Savannah River Mission Completion technology to advance the liquid waste mission at Savannah River Site.

AIKEN, S.C. — The U.S. Department of Energy’s (DOE) Office of Environmental Management (EM) liquid waste contractor at the Savannah River Site (SRS) is employing artificial intelligence (AI) to drive one of the nation’s key environmental cleanup missions forward.

Savannah River Mission Completion (SRMC) has developed AskSAM, an innovative AI assistant rooted in SRMC documentation and processes that is transforming daily operations and creating new efficiencies.

As the centerpiece of SRMC’s digital transformation, AskSAM empowers users to configure custom AI agents — including collaborative group agents designed to streamline work execution — and access a suite of predefined agents with a single click.

AskSAM answers questions and builds tools, such as an agent for human resources information, as well as training agents that help instructors develop exam materials. Many operational technology applications are now connected to AskSAM, enabling users to ask plain-language questions and receive answers drawn directly from technical systems. Users can even create their own AI agents within AskSAM to automate routine tasks.

SRMC President and Program Manager Thomas Burns Jr. is impressed with how AI has touched all corners of SRMC’s production.

“AI is truly a gamechanger for our staff, freeing them up from the routine tasks so that they can devote as much time as possible to advancing the liquid waste mission and working on the most complex issues we face,” Burns said. “We’ve been on a mission to work with AI tools as an enhancement to our entire staff, from engineers to the business office, and the results have been fantastic.

SRMC is also applying AI and machine learning directly to technical engineering tasks central to the liquid waste mission. These systems help engineering and technical teams quickly resolve operational challenges, resulting in increased facility reliability, availability and throughput.

For example, engineers have historically estimated heat generation, modeled radionuclide inventories and performed tank material balances using manual, legacy processes — often in large spreadsheets governed by nuclear safety standards, DOE directives and rigorous quality assurance. Now, SRMC uses large language models, a type of AI capable of working with both text and math, to assist with these tasks.

The benefits are substantial: For typical engineering analyses, labor hours have dropped by 40% to 70%, while safety and quality controls remain firmly in place.

AI serves as an assistant — not an authority. Engineers define the problem, provide data and assumptions, review all outputs and retain full responsibility for final results and approvals. AI-generated content is subject to technical review, independent verification and approval processes.

-Contributor: TJ Lundeen