A sharp increase in reported errors collecting evidence for veterans’ disability claims has coincided with the Department of Veterans Affairs’ wider use of automation. The overlap raises questions about how claims are being handled, but published figures do not identify whether an error came from software or staff—and do not establish that automation caused the increase.
An investigation by Navy Times reported that evidence-collection errors rose from more than 66,000 in fiscal 2021 to more than 329,000 in fiscal 2026. The investigation also reported a rise in the share of higher-level reviews returned for additional evidence or reconsideration, from about 23% to 35%. Those figures describe errors found through review; they are not, by themselves, a comparison of automated and human processing.
Different measures answer different questions
Public discussion of VA accuracy has included figures with different denominators. At an April 2026 congressional hearing, VA Deputy Under Secretary Margarita Devlin cited issue-based accuracy near 94%, while Rep. Tim Kennedy cited claims-based accuracy of 83.31%, according to Government Executive. The Veterans Benefits Administration explains that issue-level accuracy measures individual medical issues, while claim-level accuracy treats a claim as a whole. The percentages therefore are not interchangeable, and neither one alone measures evidence-collection errors or identifies their source.
Workload is another important but separate measure. The VBA says it completed more than 2.5 million disability compensation and pension claims in 2024, 27% more than in 2023. The VA’s FY 2026 budget brief says its backlog of claims older than 125 days peaked above 400,000 in early 2024, then declined for the rest of that year and stayed relatively flat in 2025. These figures describe the volume and timing of claims, not why evidence errors increased.
Automation is in the process, but the evidence does not isolate its role
The VA introduced its Automated Decision Support tool in December 2021. Navy Times reported that it helps gather and summarize records for human raters; the VA says it does not determine a veteran’s compensation. The investigation found the agency planned to expand the tool’s use, while also reporting complaints from employees about missed details. The VA has disputed that automation is responsible for the error increase. Without figures separating errors by processing method, the reported totals cannot resolve that disagreement.
The system in which those reviews occur has also changed over time. The Board of Veterans’ Appeals says VA began implementing its multi-lane appeals process in February 2019. That structural change is relevant when comparing later review activity, but it does not explain the particular rise in evidence-collection errors reported by Navy Times.
One veteran’s case illustrates the stakes, not the scale
Navy Times reported that Marine veteran Randy Leger Diaz, who filed for benefits in November 2024, was initially denied compensation for most of the conditions he claimed. The VA later acknowledged evidence-collection errors in his case, and he received monthly compensation for some claims that had first been rejected. His experience shows how an evidence error can matter to an individual claimant; one case cannot establish how common such errors are or whether ADS caused them.
A separate VA Office of Inspector General review, published September 30, 2025, examined denied PACT Act toxic-exposure decisions and said VBA had not yet identified root causes for the errors it reviewed. The review involved site visits in Waco, Texas, and St. Petersburg, Florida; it concerned a different set of decisions and did not assess ADS. It is not evidence about the cause of the current increase, but it underscores why identifying error counts and identifying their causes are distinct tasks.
