VA Inspectors Use AI to Speed up Healthcare Audits

The Department of Veterans Affairs Office of Inspector General has begun using Microsoft Copilot Chat to assist in auditing VA healthcare facilities. Investigators use the tool to review large volumes of data, including interview transcripts, documents, and physical observation notes. The goal is to streamline the process of compiling reports on whether veterans are receiving proper medical and mental health care.

Since March, inspection reports have included a standard disclosure stating that the AI tool was used to review inspection data. The reports clarify that human investigators verify the accuracy of the AI's output against source materials and take full responsibility for the final content. This approach is designed to ensure that the software acts as a helper rather than a decision-maker, maintaining human oversight over all critical findings.

Human oversight remains central to the process

A memo from the Office of Management and Budget in April 2025 defined high-impact AI use as situations where software output serves as the principal basis for decisions affecting rights or safety. The VA watchdog states that its use of Copilot Chat does not meet this definition because investigators do not rely on the AI for final decisions. Instead, the tool is used to help staff manage the massive amount of information collected during site visits.

According to a spokesperson for the watchdog, data entered into the tool remains within the secure environment of the VA Office of Inspector General. It is not shared with other departments or external entities. This containment is part of the agency's effort to manage the risks associated with using artificial intelligence in sensitive government operations, ensuring that confidential inspection data stays protected.

Standardized prompts aim to improve consistency

Former Inspector General Michael Missal, who oversaw the initial adoption of the tool, explained that the primary goal was to make the work more efficient and faster. He noted that investigators often examine similar issues across different facilities, so using standardized prompts helps ensure consistency. This method acts as a quality assurance measure, reminding staff to check for specific details that might otherwise be overlooked during a rushed review.

Missal said the tool was tested and purchased under his watch, with usage beginning in 2024. He had planned to expand its use to other parts of the office, including audits and investigations. However, a review of other reports suggests that the healthcare inspections unit was the first to adopt the technology, highlighting a targeted approach to integrating AI into specific workflows rather than a blanket deployment across all agency functions.

Efficiency gains come with operational trade-offs

While the AI assistant helps process data quickly, it introduces a new layer of complexity to the audit process. Investigators must now verify the AI's generated output against original sources, which adds a step to the workflow. This verification process is crucial to prevent errors or hallucinations in the reports. The trade-off is that while the tool speeds up data review, it requires rigorous human checking to maintain the integrity of the findings.

The use of Copilot Chat represents a shift in how federal agencies handle large datasets. By leveraging AI for initial data processing, the VA aims to free up investigator time for deeper analysis and on-site work. However, the reliance on standardized prompts means that the scope of the AI's assistance is limited to specific tasks. This focused application helps manage risks while still capturing the benefits of automated data review. As reported by GN technics/ai (en-US), this marks a significant step in the operational integration of AI tools within federal oversight mechanisms.