Skip to content
ParamergeParamerge
Stephen Lieberman
Through Paramerge
Contact

PAN Lab example

VA claims automation (automated survivor-benefit decisions)

The record no one read: automated survivor-benefit decisions

Rules read scanned death certificates and applications, then write the grant, the payment, and the letter into the record with no human when the rules match - and a thin summary sheet the record keeps as its only evidence. Modeled on the VA's automated survivor-benefit (DIC) decisions and its claims-processing copilots - their shape, not the real systems. Every inside correction channel failed; the only one that ever changed anything was an outside audit, and it runs on a multi-year clock while the automation runs every day.

Stylized model of a documented deploymentCaseworker documentation & copilots

Open this example in PAN Lab v0.1 to apply pressures and levers and watch what the system does.

What this models

This example runs on the VA-claims-automation-class rules-based determination suite network: 6 components and 14 pathways between them. Every context in the Lab is a stylized model, never a reconstruction of any actual deployment, and each assumption behind it carries a provenance label.

Evidence base: 4 assumed · 2 published baseline. In the Lab, the shaded evidence band behind each headline readout draws its width from the least-established class below.

  • assumed

    This models the rules-based-automation pattern with external-audit-only correction documented in the VA claims automation case file — not a reconstruction of the actual DIC automation, summary sheets, or calculators. These tools are deterministic automation and document extraction, not machine learning, and the documented OIG error rates are outcome statistics from statistical samples, not the tools' per-interaction generation rate.

  • baseline

    The end-to-end automated write is present at baseline: when all rules match, rating decisions, awards, and notification letters enter the permanent benefit record with no human involvement, matching the case documentation.

  • baseline

    The Inspector-General-audit node carries a real but periodic review pathway of its own — the external audit was empirically the only correction channel to change behavior — so it enters the dynamics rather than sitting on the picture. Its multi-year latency is this shape's defining correction gap, and the reconciliation of actioned payments against source evidence is drawn on the map but runs dry at baseline.

  • assumed

    Peer pathways are authored on both signs: production-goal norms spread processor to processor, one predefined rule set decides every matching claim so its defect is systematic rather than idiosyncratic, and the internal second look survives at a low level but was documented as less rigorous for automated claims than for manual ones — the check calibrated down exactly where the automation ran.

  • assumed

    The scanned-documents node carries a real extraction inflow into the automation — the optical character recognition (OCR) feed where a thin or garbled summary sheet is born — so it enters the dynamics. What that extraction encodes about any served person's entitlement stays external.

  • assumed

    The entitlements, payments, and appeal rights of served survivors and veterans are documented in the case file and measured outside any diagram like this one. This Lab models institutional propagation, not benefit outcomes, and estimates no benefit change, underpayment, or overpayment for any served person.

What this example does not show

  • The documented harm is to real survivors and veterans - improper payments, underpayments, and legally deficient notification letters. The Lab models institutional propagation, not client outcomes or entitlements, and estimates no benefit change, underpayment, or overpayment for any served person; that harm is documented in the case file and measured outside any diagram like this one.
  • These tools are deterministic rules-based automation and document extraction, not machine learning, and the OIG figures (a roughly 98 percent legal-or-procedural deficiency rate, an at-least-2 percent monetary-error floor, 27 percent inaccurate hypertension determinations, about 24 percent incorrect effective dates) are outcome statistics from statistical samples, never the tool's per-interaction generation rate.

Sources and evidence

What this example rests on, claim by claim. Every entry resolves to the same ledger the Evidence Registry publishes.

  • In a review issued April 30, 2026 (report 25-00153-47), the Department of Veterans Affairs Office of Inspector General found that at least 8,000 of an estimated 8,100 automated Dependency and Indemnity Compensation (survivor-benefit) granting decisions issued from September 2023 through August 2024 - nearly all - contained at least one legal or procedural deficiency, such as incomplete evidence summaries and omitted favorable findings, with most rating decisions listing only the death certificate as evidence. The OIG separately found that at least 2 percent of the decisions (at least 190) carried monetary-impact legal errors totaling at least 2.7 million dollars (2,727,764 dollars in questioned costs); the roughly 98 percent figure is the share with any legal or procedural defect, not the monetary-error rate. The system, phased in beginning May 2020, extracts data from scanned documents and applies predefined encoded rules to grant service-connected death claims end to end with no human involvement when the rules are met; the OIG describes it as rules-based automation and document extraction, not machine learning, and its figures are outcome statistics from a statistical sample rather than a per-interaction rate.

    empirical
    • Government Department of Veterans Affairs Office of Inspector General, Review of Automated Decisions for Veterans' Service-Connected Death Claims (Report 25-00153-47) (2026) https://www.vaoig.gov/reports/review/review-automated-decisions-veterans-service-connected-death-claims
    • Investigative Nieberg, A VA system paid out millions in 'improper' claims (Task & Purpose, 2026) https://taskandpurpose.com/military-life/va-inspector-general-survivor-benefits/
    • Trade press Weston, Audit finds VA automation glitch ruined 98% of veteran survivors' benefits claims (Public Radio East, 2026) https://www.publicradioeast.org/2026-06-19/audit-finds-va-automation-glitch-ruined-98-of-veteran-survivors-benefits-claims

  • The Office of Inspector General reported that VA's internal correction channels did not catch the automated survivor-benefit deficiencies and that the external audit was, empirically, the only channel that changed behavior. In April 2020 a VBA analyst reported through the internal defect-tracking system that automated decisions listed only the death certificate as evidence, and the Pension and Fiduciary Service closed the defect without action; the same deficiency was central to the 2026 findings, and VA removed the long-form guidance from its manual only in March 2025, immediately after the OIG's preliminary briefing - roughly five years later, and the OIG's full public report did not follow until 2026, roughly six years after the ticket. The OIG found the quality-review checklist for automated claims was less rigorous than the review traditional claims receive, and that the PACT Act section 701(b) modernization plan to Congress did not fully disclose that VBA grants these claims end to end without human intervention. Errors persisted as the program expanded: the VA Secretary announced expanded DIC automation in May 2025, and 20 additional automated decisions from September and October 2025 showed similar errors as of November 2025, with one recommendation still open and VBA concurring only in part.

    empirical
    • Government Department of Veterans Affairs Office of Inspector General, Review of Automated Decisions for Veterans' Service-Connected Death Claims (Report 25-00153-47) (2026) https://www.vaoig.gov/reports/review/review-automated-decisions-veterans-service-connected-death-claims

Where this connects

Institutional pressures in this domain

  • Workload surge — Demand outruns staffing; per-case attention shrinks and review becomes triage.
  • Deadline pressure — Statutory or managerial timeliness rules reward fast approval of machine output over slow disagreement.
  • Reviewer bottleneck — One fixed-capacity checking stage sits between AI output and consequence; everything queues behind it.
  • Staff turnover — Experienced skepticism leaves; new staff calibrate their trust on the tool itself.
  • Vendor opacity — The deploying institution cannot inspect the model, data, or update pipeline it is accountable for.
  • Compliance over substance — Paper controls (sign-offs, checklists) satisfy audits while the behavior they describe erodes.

All of them in context on the Caseworker documentation & copilots domain page.

Levers available here and the patterns behind them

Documented case histories