17,587,687
Living veterans
VetPop2023 estimate used in the FY2025 compensation report.
Official Source
01
National operating picture
The observatory separates agency-published facts from calculations, planning scenarios, and unresolved data gaps. Every major figure links to an official source and a stated limitation.
6,338,253
Compensation recipients
Veterans receiving disability compensation as of September 30, 2025.
Official Source
600,639
Pending rating-bundle claims
Weekly VBA claims inventory displayed August 10, 2026.
Official Source
69,840
Claims older than 125 days
The rating-related backlog subset displayed August 10, 2026.
Official Source
31,406
FY2027 VBA FTE
Requested full-time equivalents from all funding sources.
Official Source
367
Individual VA AI use cases
Inventory entries across every lifecycle stage; not 367 deployed systems.
Official Source
Four evidence classifications
Official
Published directly by VA, GAO, Congress, or another identified authoritative institution.
Calculated
Mathematically derived from official values with the formula disclosed.
Scenario
An editable planning assumption used to test policy choices.
Unknown
A material fact the available public record does not establish.
02
Core findings
1. Disability-compensation demand is institutionally large and medically complex
The FY2025 compensation report identifies 6.34 million compensation recipients, equal to 36.04 percent of the report’s 17.59 million living-veteran estimate. VA recognized 46.50 million service-connected disabilities—an average of 7.34 per recipient. Official
VA reported 1.85 million recipients at a 100-percent combined disability rating. That is 29.15 percent of compensation recipients and, by calculation, 10.50 percent of the living-veteran estimate. Calculated
2. The FY2027 claims plan has little operating reserve
VA projects 7,205,120 incoming rating and non-rating matters and 7,365,039 completions. Planned production exceeds projected receipts by 159,919 matters—only 2.22 percent of receipts—while year-end inventory remains 923,890 matters. Official Calculated
The implied planned utilization is 97.83 percent. A five-percent receipt surge would exceed planned production by 200,337 matters; a ten-percent surge would create a 560,593-matter shortfall. Calculated
| Workload | Receipts | Production | Year-end inventory |
|---|---|---|---|
| Rating | 2,921,957 | 2,970,929 | 453,047 |
| Non-rating | 4,283,163 | 4,394,110 | 470,843 |
| Total | 7,205,120 | 7,365,039 | 923,890 |
3. The FY2027 Board plan increases pending appeals
The budget projects 113,433 appeals received, 102,000 decisions, and 194,191 appeals pending at year-end. The Board requests 1,163 FTE—137 fewer than the FY2026 enacted estimate and 235 fewer than FY2025 actual staffing. Official
Using FY2025 observed average output only as a sensitivity assumption, the Board would need at least 1,280 FTE to keep pace with FY2027 receipts, 1,562 FTE to reduce inventory by 25,000, or 1,844 FTE to reduce it by 50,000. Scenario
4. Accredited representation exists at scale, but capacity is not measured adequately
GAO identified 13,670 accredited individuals as of November 2024: 8,141 VSO representatives, 5,008 attorneys, and 521 claims agents. GAO also reported that three-fourths of approximately one million open disability claims in August 2024 used an accredited representative. Official
Those counts do not reveal who actively practices veterans law, accepts new clients, serves rural areas, provides Spanish-language service, or can accommodate disabled claimants. GAO found an accreditation backlog of almost 2,200 claims-agent applications and an ADF workforce of 12 paralegals and six attorneys. Official Unknown
5. The “367 AI systems running” claim is inaccurate
VA’s 2025 inventory contains 367 individual AI use cases: 138 pre-deployment, 21 pilot, 134 deployed, and 74 retired. It classifies 215 as high-impact, including 90 that were deployed. An inventory use case is not necessarily a unique system. Official Calculated
Within the VBA subset, this project identified 27 use cases: 12 pre-deployment, two pilot, eight deployed, and five retired. Eleven are classified high-impact; four of those are deployed. Calculated
03
Reproducible capacity model
Change every scenario input. Official receipts and planned production are preloaded; utilization, productivity, and cost are not official staffing standards.
Formula
Required capacity = annual receipts ÷ target utilization + dedicated inventory reduction. Additional FTE = additional capacity ÷ marginal matters per FTE, rounded up.
Claimant-side parity scenario
This model estimates a multidisciplinary workforce for a selected share of complex rating matters. It does not create a legal entitlement to counsel and does not assume every professional is an attorney.
Possible roles include accredited VSO staff, attorneys, claims agents, paralegals, medical-evidence coordinators, interpreters, and accessibility specialists operating under lawful supervision and conflict controls. Scenario
04
Appeals sensitivity model
The calculation uses FY2025 Board decisions divided by FY2025 Board FTE—88.67 decisions per FTE—as an observed average. It does not account for future case mix or occupational composition. Scenario
| Policy objective | Annual decisions required | Minimum whole FTE | Difference from FY2027 request |
|---|---|---|---|
| Keep pace with FY2027 receipts | 113,433 | 1,280 | +117 |
| Reduce pending inventory by 25,000 | 138,433 | 1,562 | +399 |
| Reduce pending inventory by 50,000 | 163,433 | 1,844 | +681 |
05
Accredited-representation capacity
| Measure | Count | Reporting date | What remains unknown |
|---|---|---|---|
| VSO representatives | 8,141 | November 2024 | Active caseload, availability, language, geography, accessibility |
| Attorneys | 5,008 | November 2024 | Veterans-law practice share and new-client availability |
| Claims agents | 521 | November 2024 | Active caseload and geographic coverage |
| Claims-agent applications pending | Almost 2,200 | November 2024 | Current inventory and age distribution |
| ADF adjudicatory staff | 12 paralegals; 6 attorneys | November 2024 | Current staffing and post-OPM workforce analysis |
06
VBA artificial-intelligence registry
The table below is a reproducible subset of VA’s official 2025 inventory. It includes every row assigned to the Veterans Benefits Administration. Missing fields remain missing; the project does not infer them.
Showing 27 of 27 VBA use cases.
| Use case ID | Use case name | Lifecycle stage | Impact classification | Topic area |
|---|---|---|---|---|
VA-24-2143 |
Call Center Knowledge Navigator | Pilot | Not high-impact | Government Benefits Processing |
VA-24-2184 |
National Training Team | Schools — FAQ Gen AI Dashboard | Retired | High-impact | Not reported |
VA-24-2266 |
National Training Team | Schools NLP FAQ Dashboard | Retired | Not high-impact | Not reported |
VA-24-2545 |
Privacy Act Automation | Deployed | Not high-impact | Administrative Functions |
VA-24-3291 |
Payment Redirect Fraud (PRF) Model | Deployed | High-impact | Procurement & Financial Management |
VA-24-3578 |
Synthetic Data Creation | Retired | Not high-impact | Not reported |
VA-24-3619 |
Billie GPT | Retired | Not high-impact | Not reported |
VA-24-3824 |
electronic Virtual Assistant (e-VA) | Retired | High-impact | Not reported |
VA-24-4722 |
Automated Decision Support | Deployed | High-impact | Government Benefits Processing |
VA-24-4763 |
Mail Automation Services | Deployed | High-impact | Government Benefits Processing |
VA-24-4804 |
Master Claims Assistance Tool (M-CAT) | Pilot | Not high-impact | Government Benefits Processing |
VA-24-790 |
Pension Optimization Initiative (POI) | Deployed | High-impact | Government Benefits Processing |
VA-25-1348 |
Smart Claim Check | Pre-deployment | Not high-impact | Government Benefits Processing |
VA-25-1471 |
Articulate 360: AI Assistant | Deployed | Not high-impact | Human Resources |
VA-25-1512 |
Loan Guaranty Lender’s Handbook Chatbot | Pre-deployment | Not high-impact | Service Delivery |
VA-25-1594 |
Automated Ratings Summarization | Pre-deployment | High-impact | Government Benefits Processing |
VA-25-2036 |
Education Call Center (ECC) Next Gen POC | Pre-deployment | Not high-impact | Service Delivery |
VA-25-2089 |
AI Personal Assistance POC | Pre-deployment | Not high-impact | Administrative Functions |
VA-25-2373 |
Generative Annotations | Pre-deployment | Not high-impact | Government Benefits Processing |
VA-25-3480 |
Smart Pension Automation | Pre-deployment | High-impact | Government Benefits Processing |
VA-25-3670 |
ASKA: Automated Support and Knowledge Assistance | Pre-deployment | Not high-impact | Other |
VA-25-393 |
Predictive Claims Processing Capability POC | Pre-deployment | High-impact | Government Benefits Processing |
VA-25-5163 |
Customer Sentiment | Deployed | Not high-impact | Human Resources |
VA-25-5578 |
AICES | Pre-deployment | High-impact | Government Benefits Processing |
VA-25-5852 |
VBA Mail Management Services (MMS) - Modification | Pre-deployment | Not high-impact | Government Benefits Processing |
VA-25-6965 |
Smart Ratings Recommendation | Pre-deployment | High-impact | Government Benefits Processing |
VA-25-733 |
Auto Doc ID | Deployed | Not high-impact | Administrative Functions |
Download detailed VBA AI subset as JSON
Separate budget documents state that Automated Decision Support is planned for more than 90 percent of VBA claim volume by the end of calendar year 2027, while VBA employees retain final-decision authority. Inventory status and operational coverage are distinct measures.
07
Petition-ready policy framework
This preview does not publish a final legal petition or claim that the preliminary staffing scenarios are adjudicated facts. It converts the evidence into six reviewable policy requests.
Capacity reserve and transparency
Require VA to publish a workload-to-staffing model, forecast error bands, occupational assumptions, and a funded surge reserve rather than planning at near-saturation.
Board stabilization
Prevent annual appeals receipts from exceeding decision capacity and fund a time-limited inventory-reduction plan with docket-specific timeliness targets.
Accreditation workforce plan
Publish the completed workforce analysis, application aging, active-representative capacity, language coverage, accessibility capacity, and geographic service gaps.
Claimant-side parity
Fund independent, accredited, multidisciplinary assistance for complex, rural, homeless, visually impaired, cognitively impaired, and Spanish-primary claimants.
High-impact AI due process
Require human sign-off, immutable audit records, model and vendor disclosure sufficient for oversight, demographic testing, accessibility testing, override reporting, notice, and an effective challenge process.
Issue-level outcome data
Publish grants, partial grants, denials, reasons, review outcomes, remands, reversals, time to finality, and deaths before final resolution without exposing personal information.
08
Methodology and correction controls
Unit discipline
A rating-bundle claim is not the same as an individual medical issue, examination, appeal, or veteran. The site labels the unit used by each source.
Projection discipline
FY2027 budget values are displayed as requests or projections, never as completed outcomes.
No false precision
Workforce and cost outputs are scenario results. Users can change every nonofficial assumption.
Reproducibility
Public JSON and CSV files contain the inputs, formulas, source locations, retrieval dates, limitations, and document checksums.
Correction procedure
A correction must identify the disputed statement, authoritative source, exact location, and requested change. Material corrections appear in the changelog.
Privacy
No individual claim file, medical record, VA file number, Social Security number, address, credential, or proprietary Ayala logic belongs in this repository.
Open data downloads
09
Public source and evidence ledger
Static documents are recorded with SHA-256 checksums. Dynamic webpages are marked as dynamic and must be rechecked before quotation.
| Source | Reporting period | Used for | Location and integrity | Limitation |
|---|---|---|---|---|
| Detailed Claims Data VA-CLAIMS-WEEKLY |
Weekly status displayed August 10, 2026 | Current pending rating-bundle claims; Claims older than 125 days; Issue-level and claim-level accuracy | Claims Inventory, Claims Backlog, and Accuracy status reports. Dynamic webpage | Dynamic page. Values change weekly. Claim-based accuracy is the share of sampled cases that were error-free; it is not a wrongful-denial rate. |
| FY 2025 Annual Benefits Report: Compensation VA-ABR-2025-COMP |
Fiscal year 2025; recipients as of September 30, 2025 | Veteran population; Compensation recipients; Service-connected disabilities; 100-percent combined ratings | PDF pages 4, 5, 8, and 10. SHA-256: b78b0ef4f76c82b5dec067ac7f54f48821f7b72626e15170561ca9877f13fd29 | Counts recognized service-connected disabilities, not every condition claimed or experienced. Veteran population is based on VetPop2023. |
| FY 2027 Congressional Submission: Budget in Brief VA-BUDGET-2027-BIB |
FY 2025 actual, FY 2026 enacted/estimated, FY 2027 request | VA and VBA staffing; Board staffing and appeals projections; Budget totals | PDF pages 1, 8, and 35–38. SHA-256: 83dba990470148f01f9090730ddac932a8809b12775f0fcf0a00c34b378ea09e | FY 2027 values are requested or projected, not enacted outcomes. |
| FY 2027 Congressional Submission, Volume III: Benefits and Burial Programs and Departmental Administration VA-BUDGET-2027-V3 |
FY 2024–2027 workload and budget data | Rating and non-rating receipts, production, and inventory; Automated Decision Support expansion; Direct-labor FTE context | PDF pages 77–79 and 59–71. SHA-256: 0792f98b5038a3f1cb54aa8e857076ca098bfcf707250860ea3bcb89d23a0ead | Projected claims include PACT Act workload, while the direct-labor FTE table cited in the document excludes Toxic Exposures Fund FTE. This observatory therefore does not treat that table as a complete productivity denominator. |
| 2025 VA AI Use Case Inventory VA-AI-2025 |
Inventory current as of December 2025 | AI lifecycle counts; High-impact counts; VBA-specific AI registry | All Use Cases worksheet, rows 3–369. SHA-256: f6bfdb02859539de48aeba1003afda18bd0fa92223df33763bc872e292b8ad88 | An inventory entry is a use case, not necessarily a unique software system. Lifecycle status must be considered; retired and pre-deployment entries are not running production systems. |
| Veterans Benefits: More Thorough Planning Needed to Help Better Protect Veterans Assisted by Representatives GAO-25-107211 |
Data primarily through November 2024 | Accredited representative counts; Claims-agent application backlog; ADF staffing | Report pages addressing representative counts and ADF workforce resources. SHA-256: ef8d2a30cf96b996a2bb1c9a726696c56243ee3ff5f219fd7d6d1c159c6f8db2 | Accreditation does not establish active practice, new-client availability, caseload, language capacity, geography, or accessibility capability. |
| Web Content Accessibility Guidelines (WCAG) 2.2 WCAG-22 |
Current Recommendation | Accessibility target and testing protocol | W3C Recommendation. Dynamic webpage | This project targets WCAG 2.2 AA. Automated tests alone cannot establish full conformance. |
10
Accessibility statement
Accessible HTML is the authoritative publication format. The project targets WCAG 2.2 Level AA but does not claim certified conformance. It uses semantic headings, landmarks, keyboard-operable controls, visible focus, high contrast, large targets, reduced-motion support, descriptive table captions, and text equivalents instead of chart-only communication.
Automated testing cannot replace disabled-user review. Manual acceptance review remains required with TalkBack and Chrome on Android, NVDA with Firefox and Chrome on Windows, VoiceOver with Safari, keyboard-only navigation, and browser zoom at 200 and 400 percent.
Report an accessibility defect through GitHub Issues without including personal claim information.