evidoria

← Back to browse

Good practice Imported

AI-Assisted Disability Claims Processing at the US Department of Veterans Affairs

United States of America · Washington, D.C. · See the United States of America profile · See the Washington, D.C. profile

Evidence: Observational / pre–post Top 32% 67/100 · Ask Evidence Copilot about this practice

Since August 2025 the VA has piloted an AI tool that pre-fills disability-benefit questionnaires from veterans' medical records; average processing time fell 42% (141 to 81 days) and accuracy hit a two-year high of 93.95%, even as a GAO report flagged governance gaps.

81 days
Average disability-claims processing time (early 2026 (down from 141 days in January 2025))
-42%
Processing-time reduction (January 2025 to early 2026)
93.95%
12-month issue-based accuracy rate (two-year high, early 2026)
under 100,000
Pending disability and pension claims backlog (February 2026)
2,000,000+
Disability and pension claims processed in FY2026 (as of June 2026)
$130 million
Proposed FY2027 budget for claims automation and AI
AI-Assisted Disability Claims Processing at the US Department of Veterans Affairs

Details

Maturity
Scaling
Promoter
U.S. Department of Veterans Affairs — Veterans Benefits Administration
Period
2025-2026
Keywords
veterans affairs, social benefits, disability, government services, healthcare administration

Context

In August 2025, the U.S. Department of Veterans Affairs (VA) began piloting an AI tool that uses veterans' medical records to help complete disability benefit questionnaires (DBQs), reducing the need for in-person medical exams, as part of a broader modernisation push at the Veterans Benefits Administration (VBA).

Objectives

The pilot aims to speed up disability and pension claims processing and reduce the claims backlog while maintaining or improving decision accuracy.

Activities

The AI tool pre-fills disability benefit questionnaires directly from veterans' medical records, and the approach has since been extended across multiple VBA regional offices and claim types beyond the initial August 2025 pilot.

Results

Average claims-processing time dropped 42%, from 141 days in January 2025 to 81 days by early 2026; the VBA's 12-month issue-based accuracy rate reached 93.95%, its highest in two years; the backlog of pending disability and pension claims fell below 100,000 by February 2026; and VA processed more than 2 million disability and pension claims in fiscal year 2026 as of June 2026.

Conclusions

A March 2026 U.S. Government Accountability Office report (GAO-26-109137) found gaps in VA's technology-modernisation and AI risk-management practices, congressional Democrats have warned that faster processing combined with concurrent workforce reductions could raise the risk of erroneous decisions, and the proposed FY2027 budget allocates $130 million specifically to claims automation and AI.

Implementation
Indicative cost
High (€500k–€5M)
Time to results
Short (< 1 year)
Staffing & skills
Veterans Benefits Administration (VBA) claims-processing staff and regional offices, VA technology-modernisation teams integrating the tool with veterans' medical-records systems

Conditions for success

  • Integration of the AI tool with veterans' existing medical-records data to pre-fill questionnaires reliably
  • Extension of the tool across multiple VBA regional offices and claim types rather than a single site
  • Continued budget commitment, including the proposed $130 million FY2027 allocation to claims automation and AI

Common failure modes

  • GAO-26-109137 identifies specific gaps in VA's AI risk-management and technology-modernisation governance
  • Lawmakers have flagged insufficient oversight of AI's role in claims decisions amid concurrent staffing cuts, raising concern about erroneous decisions
  • VA has not fully disclosed or documented how AI outputs are used or validated in individual claim decisions

Commonly funded by

National / regional programmes Own resources / municipal budget

Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Attachments

Similar practices you may find useful