The Department of Veterans Affairs has disclosed 367 artificial intelligence use cases operating across the agency, including 215 classified as high-impact systems supporting healthcare, benefits processing, records management, communications and internal operations.
The recently released 2025 VA AI Inventory catalogs cases of AI use across the department, from clinical tools used during patient care to systems that help process benefits, automate records management and support customer service functions.
Among the systems identified is Ambient AI Scribe, a clinical documentation tool used during medical appointments.
According to VA, the technology listens to clinician-patient conversations and generates clinical notes, reducing administrative workload and allowing providers to spend more time focused on veterans.
VA’s disclosure also details widespread use — by more than 50,000 personnel — of commercial AI products across the department, including Microsoft Copilot Chat, Microsoft Teams Premium, Grammarly GitHub Copilot and other platforms for support tasks such as meeting transcription, document drafting, report summarization, information retrieval and software development.
Nearly Every Major VA Function Features AI
Health and medical applications represent the largest category of use cases, followed by systems supporting benefits administration, information technology, service delivery, financial management and administrative operations.
Applications identified in the inventory assist clinicians with imaging analysis and diagnostic workflows, identify disease risk, summarize records, analyze feedback, process forms and provide information through chat-based tools. VA also classifies its VA.gov chatbot as a high-impact use case.
VA says the chatbot is a tool that helps veterans locate information, complete tasks and access support services through self-service interactions.
Another high-impact use case, TERA Memorandum Automation, helps claims processors complete toxic exposure memoranda by pre-populating answers using information already contained in veterans’ records.
VA reports the system achieved a 98.12% accuracy rate and was used to assist with more than 181,000 forms, saving an estimated 54,581 work hours since deployment.
“VA is using artificial intelligence to improve how veterans access care, benefits and services, while also helping employees work more efficiently and effectively,” said VA Press Secretary Quinn Slaven in a statement to Military Times.
According to Slaven, the inventory reflects “steady growth, stronger governance and expanding real-world impact, particularly in health care, benefits processing and operational efficiency.”
Inventory Provides A Detailed View Of Governance Requirements
For currently deployed high-impact systems, the inventory tracks whether agencies completed pre-deployment testing, impact assessments, independent reviews, monitoring processes, operator training, fail-safe mechanisms and appeal procedures.
Individual entries also identify potential risks associated with specific applications. For several health care systems, descriptions note that inaccurate outputs could contribute to downstream clinical or operational impacts if not detected and corrected, while also stating that providers receive training and oversight intended to mitigate those risks.
VA says all high-impact systems are required to complete an AI Impact Assessment and Risk Mitigation Plan before being put to work.
“Deployed high-impact AI use cases are continuously monitored,” said Slaven. “And any use case that cannot be brought into compliance is discontinued.”
Some AI systems previously reported by VA are now classified as retired. Those systems include healthcare, training and operational applications. The inventory does not specify why individual systems were deactivated.
Performance Data Varies Across Systems
The level of detail provided in the inventory varies from one system to another. Some entries provide measurable performance outcomes, testing results or operational metrics. Others provide only basic descriptions of intended functions and governance requirements.
The U.S. Government Accountability Office, which developed an AI Accountability Framework for federal agencies, says meaningful oversight requires documentation that can be independently reviewed.
“GAO’s AI Accountability Framework highlights the importance of documentation and a paper trail that an independent third party can follow,” GAO officials told Military Times. “As AI inputs and operations are not always visible, a central tenet to oversight is, ‘Show me, don’t just tell me.’”
GAO said organizations should be able to demonstrate who is responsible for AI systems, how those systems operate, whether they rely on quality and reliable data and how limitations are identified and addressed.
The agency also emphasized the importance of monitoring metrics such as accuracy and error rates and establishing thresholds that help determine when corrective action or retirement of a system is warranted.
Veterans Advocates Want Human Accountability
Chris Macinkowicz, deputy director of the Veterans of Foreign Wars National Veterans Service, said the organization’s concerns focus less on whether VA should use artificial intelligence and more on how those systems are managed after deployment.
“AI needs to be a tool, not an employee,” Macinkowicz told Military Times.
He said human review remains essential, particularly when systems are involved in processes affecting veterans.
“We need to make sure human eyes are on it afterwards,” he said. Macinkowicz also stressed the importance of continued evaluation after deployment.
“We need to have assurances from VA that they’re going to go back after these AI agents are deployed to make sure that they’re still accurate a month down, six months down, a year down [the road],” he said.
GAO officials said organizations should establish metrics that help determine when corrective action is needed and when an AI system should be retired because it no longer meets its intended objectives.
As VA expands the use of AI across healthcare, benefits processing and other operations, GAO officials said meaningful oversight depends on agencies being able to demonstrate, through documentation and measurable performance data, that systems continue operating as intended.
