VA exploring multiple generative AI pilots, responding to ‘high demand’ from workforce

Other gen AI projects within the agency include an AI-powered code-completion tool and a “meta pilot” centered on search, summarization and content drafting.
VA Secretary Dennis McDonough testifies before a Senate Appropriations Subcommittee on Military Construction on May 4, 2022 in Washington, D.C. (Photo by Kevin Dietsch/Getty Images)

The Department of Veterans Affairs plans to pilot several applications of generative AI, a spokesperson for the agency told FedScoop. As part of a “generative AI meta pilot,” the agency plans to test the technologies on open-source large language models that have now been brought into the VA’s already authorized-to-operate platforms. 

Generative AI applications are in “high demand” at the agency, VA spokesperson Gary Kunich told FedScoop. Amid that interest, the department now plans to test the use of generative AI to summarize feedback from users and employee surveys run by the Veterans Experience Office, assist employees in human resources and the Veterans Benefits Administration with reading through internal policies, and draft contracting documents, including statements of work and descriptions of jobs. 

There are other generative AI projects, too. The agency is pursuing an ongoing pilot with GitHub Copilot, a generative AI-powered code-completion tool, as well as a generative AI “meta pilot” focused on search, summarization, and content drafting. Current AI tech sprints within the agency related to ambient dictation for clinical encounters and community care document processing also involve generative AI.

These use cases have not yet been listed on the agency’s public AI inventory, which agencies are required to produce annually by an executive order. As FedScoop reported earlier this year, agencies still vary widely in their approach to generative AI. 


But the VA has also ditched at least one generative AI pilot. Last year, the agency experimented with using generative AI in the form of a chatbot but later decided to abandon the project, according to Kunich. The current chatbot used by the agency remains rules- and retrieval-based, though it takes advantage of natural language understanding and processing. 

Current guidance for the technology at the agency, which FedScoop obtained through a public records request, states that no “web-based, publicly available generative AI service has been approved for use with VA sensitive data.” Department staff are also instructed to limit the sharing and saving of data on these systems and to check the output of large language models for accuracy. You can read that document here.

Rebecca Heilweil

Written by Rebecca Heilweil

Rebecca Heilweil is an investigative reporter for FedScoop. She writes about the intersection of government, tech policy, and emerging technologies. Previously she was a reporter at Vox's tech site, Recode. She’s also written for Slate, Wired, the Wall Street Journal, and other publications. You can reach her at Message her if you’d like to chat on Signal.

Latest Podcasts