{"resourceId":"wa-dshs-snap-ai-review-pilot-proposal","versions":[{"version":"external-ee945a20aeb0d45544a3c9c5dc285bcfdcff072329e6cfd4e0722f115581099c","resource":{"id":"wa-dshs-snap-ai-review-pilot-proposal","title":"WA Department of Social and Health Services eyes cuts, modernization for 2027","organization":"Washington State Department of Social and Health Services (reported by TVW / Washington State Standard)","sector":"State human services / SNAP","geography":"Washington, United States","publishedAt":"2026-10-19","publicationDate":"2026-10-19","eventDate":null,"sourceName":"Washington State Standard (first published by TVW)","sourceLabel":"Washington State Standard (first published by TVW)","sourceUrl":"https://washingtonstatestandard.com/2026/10/19/wa-department-of-social-and-health-services-eyes-cuts-modernization-for-2027/","evidenceClass":"independent-reporting","outcomeClass":"emerging","topics":["governance-procurement","operating-model","knowledge-work"],"finding":"Washington DSHS is proposing a budget decision package for an AI pilot that identifies SNAP cases for additional review to reduce the payment error rate, which rose to 6.98% in 2025 (below the 10.62% national figure). Secretary Angela Ramirez said AI would not make eligibility decisions but would be a tool for decision makers, as the agency also plans unfilled vacancies, system consolidation, contract renegotiation and narrower eligibility in some programs.","sledRelevance":"A large state HHS agency is budgeting for AI-driven SNAP case selection as H.R. 1 ties state cost share to payment error rates above 6%; a near-term procurement signal.","evidence":"Interview-based report of a proposal; no design, vendor, cost or evaluation plan described. Error-rate figures attributed to USDA.","architectureImplications":"Case-selection models would draw on SNAP case data and must integrate with the state's eligibility and QC review workflow.","governanceImplications":"The secretary states AI will not decide eligibility; the legislature must approve the decision package.","securityPrivacyImplications":"Targeting cases for extra review touches sensitive household data; wrongful-delay concerns raised in Michigan apply here too.","caveats":"Proposal only; budget not approved; no technical detail.","streamIds":["health-human-services"],"roles":{"sales":"Sales framing: Washington is a live example of SNAP error-rate pressure turning into a budget request for AI case review. The opportunity is helping DSHS shape a pilot with clear human-review and fairness controls before procurement, not selling a finished model. Avoid claims that AI will lower the error rate by a set amount; the source quantifies no expected impact, and the package still needs legislative approval.","engineering":"Engineering framing: a SNAP case-selection model would score cases for supervisory or QC review using eligibility system data. Key design questions are which features are used, how scores route to reviewers, and how the model is validated against QC findings. The stated boundary that AI does not make eligibility decisions should be enforced in the workflow, with reviewers able to see why a case was flagged.","delivery":"Delivery framing: if funded, ownership should sit with DSHS economic services and QC teams. Delivery should baseline current case-review yield, pilot in limited offices, and monitor whether flagged cases fall disproportionately on particular groups. With vacancies left unfilled, plan reviewer capacity realistically so flags do not create backlogs that delay benefits."},"roleDetails":{"sales":{"customerProblem":"Washington's SNAP payment error rate rose to 6.98%, near the H.R. 1 cost-share threshold, while the agency faces budget cuts.","stakeholders":["DSHS Economic Services Administration","SNAP QC and payment accuracy staff","DSHS CIO","Legislative budget staff"],"discoveryQuestions":["What is the current error-finding yield of manual case reviews?","Which error types drive the rate?","How will flagged cases be reviewed and by whom?","What fairness checks will the pilot include?"],"valueHypothesis":"Bounded hypothesis: better case selection could raise the share of reviewed cases that contain errors, if reviewer capacity and fairness controls are in place.","potentialEngagement":"Pilot-design support for the proposed SNAP case-review AI, including baseline metrics and human-review workflow.","unsupportedClaims":["AI will bring the error rate below 6%","The pilot is funded","AI will reduce staffing needs"]},"engineering":{"fit":"Fits case-prioritization analytics feeding human review.","architectureIntegration":"Scores cases from the eligibility system and routes them into existing supervisory or QC review queues.","prerequisites":["Historical QC findings for training and validation","Eligibility system data access","Reviewer workflow tooling"],"constraints":["Pending legislative approval","USDA FNS rules on SNAP processes","Reviewer capacity under unfilled vacancies"],"security":"Limit model features to what is necessary; log every score and review decision.","proofOfValue":["Compare error-finding yield of AI-selected versus current case selection","Check flag rates by demographic group","Measure added processing delay for flagged cases"]},"delivery":{"work":"Design and run a limited SNAP case-review pilot with baseline and fairness monitoring.","dependencies":["Budget approval","QC historical data","Reviewer staffing"],"ownership":"DSHS Economic Services Administration with SNAP QC.","skillsAdoption":"Reviewers learn to treat flags as prompts, not conclusions.","governanceCheckpoints":["Legislative approval","Fairness review before expansion","Client impact review"],"acceptanceCriteria":["Proposed: higher error-finding yield than baseline","Proposed: no material disparity in flag rates without documented justification","Proposed: flagged cases processed within SNAP timeliness standards"],"risks":["Wrongful delays or denials","Reviewer backlog","Racial or language disparities"]}},"retrievedAt":"2026-10-29T22:06:40-05:00","enrichedAt":"2026-10-29T22:20:00-05:00","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Arrives alongside unfilled vacancies and a no-raise first contract year, so AI is partly framed as capacity relief for a constrained workforce.","procurementImplications":"If funded, this would be a 2027 procurement for SNAP case-review analytics.","operatingModelImplications":"Pairs retrenchment with modernization; the secretary acknowledged that least harm does not mean no harm for clients.","sourceVerification":{"openedUrl":"https://washingtonstatestandard.com/2026/10/19/wa-department-of-social-and-health-services-eyes-cuts-modernization-for-2027/","referenceExcerpt":"Ramirez stressed that AI would not make eligibility decisions but be a new powerful tool for decision makers.","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}