{"resourceId":"dutchess-county-eva-snap-virtual-assistant","versions":[{"version":"external-ee945a20aeb0d45544a3c9c5dc285bcfdcff072329e6cfd4e0722f115581099c","resource":{"id":"dutchess-county-eva-snap-virtual-assistant","title":"Dutchess County DCFS Pilots AI-Powered Virtual Phone Assistant to Improve SNAP Customer Service","organization":"Dutchess County Department of Community and Family Services","sector":"County social services / SNAP","geography":"Dutchess County, New York, United States","publishedAt":"2026-08-07","publicationDate":"2026-08-07","eventDate":null,"sourceName":"Dutchess County Executive news release","sourceLabel":"Dutchess County Executive news release","sourceUrl":"https://www.dutchessny.gov/Departments/County-Executive/47531.htm","evidenceClass":"government-evaluation","outcomeClass":"emerging","topics":["knowledge-work","accessibility-workforce","operating-model"],"finding":"Dutchess County is piloting two EVA components for SNAP: a 24/7 phone assistant in more than 30 languages that answered nearly 1,200 calls since July 1, saving more than 23 staff hours, and a module that recites the federally required SNAP Rights & Responsibilities script after interviews (900+ clients since mid-June). The county reports SNAP 30-day application timeliness improved from 74% (July 2025) to 83% (July 2026) during the test period. Seventeen New York counties use EVA components.","sledRelevance":"A county social services district shows a lower-risk AI pattern: informational phone answering and scripted compliance recitation that free caseworker time without touching eligibility decisions, relevant as ABAWD work-rule changes raise SNAP workloads.","evidence":"County press release on an update presented to the legislature; metrics are self-reported; the timeliness gain is attributed to EVA without controlling for other factors.","architectureImplications":"Telephony-based multilingual voice assistant; scripted module with verbal attestation captured for QA.","governanceImplications":"Delivering a federally required script by AI with client attestation raises the question of state and federal acceptance, which the release does not address.","securityPrivacyImplications":"Callers may share personal details; the release does not describe data handling.","caveats":"Press release; small reported time savings (23 hours); the timeliness change has multiple possible causes. Published Aug 7, 2026, before the main window, but newly relevant given October 2026 funding for caseworker workload tools and rising ABAWD workloads.","streamIds":["health-human-services"],"roles":{"sales":"Sales framing: counties under ABAWD and H.R. 1 workload pressure want quick, low-risk relief. Dutchess shows AI phone and script modules that do not touch eligibility decisions, with self-reported timeliness gains. Lead with language access and staff time, propose a measured pilot, and do not attribute the 74% to 83% timeliness change solely to AI; the county offers no controlled comparison.","engineering":"Engineering framing: this is telephony-integrated voice AI with multilingual answering and a scripted rights-and-responsibilities module that records client attestation. Check call data retention, how attestations are stored with the case record, escalation to humans, and the accuracy of program information across 30+ languages.","delivery":"Delivery framing: start with informational calls and scripted modules, measure call containment, staff time and timeliness against baseline, and review translation quality. Ownership sits with the social services district's customer service or SNAP operations lead. Confirm with state oversight that AI delivery of required scripts meets program rules before full rollout."},"roleDetails":{"sales":{"customerProblem":"County SNAP offices face growing call volume and interview workload from ABAWD and federal changes.","stakeholders":["Commissioner of social services","SNAP operations manager","County IT / telephony","Language access coordinator"],"discoveryQuestions":["What share of calls are informational?","How long does reciting required scripts take per interview?","Which languages do callers need?","Has the state approved AI delivery of required notices?"],"valueHypothesis":"Bounded hypothesis: offloading informational calls and scripted recitations can return staff time to case processing.","potentialEngagement":"A phone-assistant and scripted-module pilot with baseline call and timeliness metrics.","unsupportedClaims":["AI caused the timeliness improvement","Large staff savings (reported savings were 23 hours)","AI handles eligibility questions reliably"]},"engineering":{"fit":"Fits informational and scripted, non-decision interactions.","architectureIntegration":"Telephony integration, multilingual voice AI, attestation capture stored for QA.","prerequisites":["Telephony system access","Approved knowledge content","Attestation storage linked to cases"],"constraints":["State and federal acceptance of AI-delivered scripts","Translation accuracy","Escalation to staff"],"security":"Call recording retention, PII minimization, access controls.","proofOfValue":["Call containment and escalation rate","Translation accuracy spot checks","Timeliness compared with prior year and peer counties"]},"delivery":{"work":"Pilot phone and scripted modules, measure, then expand.","dependencies":["Vendor contract","State approval","Content maintenance"],"ownership":"County SNAP operations / customer service.","skillsAdoption":"Staff learn escalation handling; content owners keep information current.","governanceCheckpoints":["State approval for the script module","Translation QA","Client feedback review"],"acceptanceCriteria":["Proposed: defined call containment rate","Proposed: staff minutes saved per interview measured","Proposed: no rise in complaints about information accuracy"],"risks":["Outdated information","Non-acceptance of AI-recited scripts","Callers unable to reach humans"]}},"retrievedAt":"2026-10-29T22:03:30-05:00","enrichedAt":"2026-10-29T22:20:00-05:00","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"30+ language support may improve language access; frees staff from reciting scripts.","procurementImplications":"A shared vendor across 17 New York counties suggests a cooperative or piggyback procurement path.","operatingModelImplications":"Moves routine information and compliance scripts to AI so staff can focus on interview notes and processing.","sourceVerification":{"openedUrl":"https://www.dutchessny.gov/Departments/County-Executive/47531.htm","referenceExcerpt":"EVA has handled nearly 1,200 calls, saving more than 23 staff hours","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}