{"resourceId":"indiana-dcs-adviser-ai-kinship-search-paperwork-20261027","versions":[{"version":"external-893d782766b163f6b2418e6338a83f855dec65712c88ff7369114a53161ebb49","resource":{"id":"indiana-dcs-adviser-ai-kinship-search-paperwork-20261027","title":"Indiana child welfare adviser vets AI for paperwork, staffing forecasts and kinship searches, ruling out AI determinations","organization":"Indiana Capital Chronicle; Gov. Mike Braun's special advisor on abuse and neglect, Adam Krupp (former Indiana DCS director)","sector":"Child welfare","geography":"Indiana, United States","publishedAt":"2026-10-27","publicationDate":"2026-10-27","eventDate":null,"sourceName":"Indiana Capital Chronicle","sourceLabel":"Indiana Capital Chronicle (States Newsroom)","sourceUrl":"https://indianacapitalchronicle.com/2026/10/27/child-welfare-advisor-wrapping-up-his-work-this-year/","evidenceClass":"independent-reporting","outcomeClass":"emerging","topics":["knowledge-work","governance-procurement","operating-model"],"finding":"Adam Krupp, who left the Indiana DCS director post in April for a special advisor role to Gov. Mike Braun on abuse and neglect, is completing a comprehensive evaluation of the state's child welfare system and has been vetting AI's cost, implementation timelines and uses. He said AI should not make child welfare determinations but could reduce paperwork and data entry, forecast staffing from caseload trends, and mine large case files for possible kinship placements, a search he said can take a case manager about a week but AI 60 seconds.","sledRelevance":"A state child welfare system is likely to receive formal AI recommendations in an upcoming assessment, framed around administrative relief and kinship search rather than risk scoring.","evidence":"News report based on an interview and weekly status reports. The kinship-search time comparison is the adviser's claim, not a measured result. No tool, vendor or pilot is named.","architectureImplications":"Kinship search across years of unstructured case notes needs retrieval and entity extraction over the case management system, with strict access controls.","governanceImplications":"Explicitly excluding AI from determinations is a useful boundary, but kinship identification still affects placement decisions and needs human verification and audit.","securityPrivacyImplications":"Case files contain highly sensitive child and family data; mining them for names requires role-based access, logging and data minimization.","caveats":"Plans and opinions, not deployments; the assessment report is not yet published; claims are unquantified apart from an anecdotal time comparison.","streamIds":["health-human-services"],"roles":{"sales":"Sales framing: Indiana's child welfare leadership is signaling interest in administrative AI and kinship search, not predictive risk tools. An appropriate discovery conversation would focus on case file search, documentation and workforce forecasting. Any engagement should wait for, and align with, the forthcoming assessment recommendations. Do not repeat the week-versus-60-seconds claim as a measured result; it is an anecdote. Interpretation: kinship search is a distinct, bounded use case worth testing.","engineering":"Engineering framing: kinship search means retrieval and named-entity extraction across many years of unstructured notes from many caseworkers. Engineers should plan for deduplication of names, relationship inference with confidence levels, and citations back to the source note so workers can verify. Access must follow case assignment, and outputs should be leads for caseworkers, not placement recommendations. Staffing forecasts are a separate analytics workload.","delivery":"Delivery framing: the owner would be DCS permanency and practice leadership, with IT and legal. Delivery should start with a small pilot on closed or consented cases, measure time to identify candidate relatives and how many leads caseworkers verify, and keep a human approval step. Training should stress that AI output is a starting point. Targets should be set after a baseline, not taken from the anecdote."},"roleDetails":{"sales":{"customerProblem":"Caseworkers spend days combing long case files for possible kinship placements and lose field time to paperwork and data entry.","stakeholders":["DCS director and deputy for permanency","Caseworker supervisors","DCS IT and data teams","Agency legal counsel","Governor's office"],"discoveryQuestions":["How long does a kinship search take today, measured?","Which case management system holds the notes, and how structured are they?","What uses has the assessment recommended?","What review is required before contacting a possible relative?"],"valueHypothesis":"Bounded hypothesis: AI-assisted case file search could shorten time to identify potential kin, subject to caseworker verification and a measured baseline.","potentialEngagement":"A scoped pilot design for AI-assisted kinship lead identification on a limited set of cases, including baseline time measurement and accuracy review.","unsupportedClaims":["That AI finds kin in 60 seconds as a measured result","That AI should make placement or safety decisions","Any savings or outcome improvements"]},"engineering":{"fit":"Fits retrieval and extraction over unstructured child welfare case notes; not a decision tool.","architectureIntegration":"Index case notes from the case management system, run entity and relationship extraction, and return cited leads into the caseworker workflow.","prerequisites":["Access to historical case notes","Role-based access aligned to case assignment","Ground truth from past kinship placements"],"constraints":["Confidentiality laws for child welfare records","Name ambiguity and data quality","No tool or vendor selected yet"],"security":"Restrict queries to assigned cases, log every search, and keep data within approved government cloud boundaries.","proofOfValue":["Recall of known kin on closed cases","Time to first verified lead versus baseline","False-lead rate reviewed by supervisors"]},"delivery":{"work":"Pilot kinship-lead search on closed cases, then on a small number of active cases with supervisor review.","dependencies":["Assessment recommendations","DCS legal approval","Case management data access","Caseworker participation"],"ownership":"Indiana DCS permanency and practice leadership, with DCS IT as technical owner.","skillsAdoption":"Caseworkers need training on verifying AI-surfaced leads and on the limits of extraction from old notes.","governanceCheckpoints":["Legal review of data use","Supervisor approval before relative contact","Post-pilot accuracy and equity review"],"acceptanceCriteria":["Measured reduction in time to first verified lead against baseline","Every lead cites its source note","No AI output used as a placement decision"],"risks":["Missed or wrong relatives","Privacy breach of case files","Scope creep toward risk scoring"]}},"retrievedAt":"2026-10-30T22:07:44-05:00","enrichedAt":"2026-10-30T22:20:00-05:00","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Framed as returning caseworker time to families; workforce forecasting could affect staffing decisions and should be shared with staff.","procurementImplications":"Any procurement should follow the published assessment and specify that tools support, not make, determinations.","operatingModelImplications":"AI-assisted search would add a verification step to permanency practice and require supervisors to own lead quality.","sourceVerification":{"openedUrl":"https://indianacapitalchronicle.com/2026/10/27/child-welfare-advisor-wrapping-up-his-work-this-year/","referenceExcerpt":"He noted immediately that he doesn’t mean allowing AI to make child welfare determinations.","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}