{"resourceId":"texas-treo-sam-aws-govcloud-dir-discover-20261006","versions":[{"version":"external-d82e7b173a4e8428c81a1f8aec3252460927ceb6894f3c77784546c9b953cbae","resource":{"id":"texas-treo-sam-aws-govcloud-dir-discover-20261006","title":"Texas Regulatory Efficiency Office presents SAM, a citizen-facing regulatory assistant hosted as an AI service in AWS GovCloud","organization":"Texas Regulatory Efficiency Office (Office of the Governor); presented at Texas DIR Discover","sector":"State government","geography":"Texas, United States","publishedAt":"October 6, 2026 (DIR Discover presentation)","publicationDate":"2026-10-06","eventDate":"2026-10-06","sourceName":"Texas Department of Information Resources (DIR)","sourceLabel":"DIR Discover presentation deck (PDF)","sourceUrl":"https://dir.texas.gov/sites/default/files/2026-10/Making%20Government%20Navigable%20A%20Texas%20Use%20Case%20for%20Citizen-Facing%20AI%20-%20Vulcan.pdf","evidenceClass":"vendor-claim","outcomeClass":"emerging","topics":["knowledge-work","infrastructure","governance-procurement","data-security"],"finding":"The October 6, 2026 DIR Discover deck describes SAM on efficiency.texas.gov as a public, no-login assistant that helps Texans find which agency, rule and form applies, citing a corpus of 44,000 regulations in the Texas Administrative Code, plus statutes, agency forms and guides. The security slide states the AI service runs in AWS GovCloud on a texas.gov site with FedRAMP noted; guardrails say SAM does not give legal advice or decide cases, and answers point users to the linked agency source. This is the first explicit public statement located tying SAM's hosting to AWS GovCloud; the May 2026 launch is already archived (texas-treo-sam-regulatory-chatbot-statescoop-20260515), where the director described orchestration with multiple commercial models.","sledRelevance":"State-government exemplar of a citizen-facing, citation-grounded regulatory navigation assistant on AWS GovCloud; directly comparable to permitting and licensing assistants in other states and large cities.","evidence":"Presentation slides only: architecture claims (GovCloud, FedRAMP, no accounts), scope (44,000 regulations) and guardrail lists. No usage counts, accuracy rates, evaluation method, cost or incident data are presented in the deck.","architectureImplications":"Pattern: public website front end on texas.gov, retrieval over statutes, administrative code and forms, model service in AWS GovCloud, answers point to linked agency sources. The deck does not state which foundation models are used or whether non-GovCloud model calls occur; the archived StateScoop launch report quoted the director describing partnering with Gemini, Claude and ChatGPT, which should be reconciled against the GovCloud framing.","governanceImplications":"Explicit scope limits (no legal advice, no case decisions, no unrelated tasks) and the deck's criteria for AI opportunities (repetitive rule-bound work, existing records, recoverable mistakes, a person stays on the decision) give a reusable governance checklist for state agencies.","securityPrivacyImplications":"No-login design minimizes stored identity data; users are told to keep sensitive information out of the chat. FedRAMP and GovCloud are noted, but the authorization boundary and model-provider data flows are not documented in the deck.","caveats":"Operator/vendor-presented material (file name references the implementing firm Vulcan); not an independent evaluation. Slide text is terse and partly image-based. No outcome metrics. Model provenance unclear relative to earlier statements.","streamIds":["aws","state-government"],"roles":{"sales":"Sales framing: Texas gives a credible state reference for a citizen-facing regulatory navigator on AWS GovCloud, useful when agencies ask whether anyone has put generative AI in front of the public safely. Position the conversation around scoped guardrails and cited answers rather than efficiency claims, because the deck reports no accuracy or usage metrics. Use it to open discovery about licensing and permitting confusion, and avoid implying the state endorsed a specific model or vendor stack beyond what was presented.","engineering":"Engineering framing: the reference pattern is retrieval over a large statutory and administrative-code corpus with citations, served from a public texas.gov page with the AI service in AWS GovCloud. Before reusing it, confirm which models run inside GovCloud versus any external model calls the director previously described, how the 44,000-rule corpus is chunked and refreshed, and how refusals are enforced. Treat the FedRAMP mention as a prompt to verify the actual authorization boundary, not as proof of it.","delivery":"Delivery framing: the hard part is not the chat front end but keeping a multi-agency rules corpus current and keeping answers inside stated limits. A delivery plan modeled on SAM needs named corpus owners at each agency, a rule-change sync cadence, an accuracy and citation-fidelity test set, and a refusal regression suite. Because the deck provides no measured outcomes, acceptance criteria must be defined locally and measured from launch rather than borrowed from Texas."},"roleDetails":{"sales":{"customerProblem":"Residents and businesses cannot easily determine which agency, rule, license or form applies to them across a large administrative code.","stakeholders":["State CIO or CDO","Agency licensing and permitting directors","Governor's or efficiency office policy staff","Agency general counsel","Public information officers"],"discoveryQuestions":["Which licensing or permitting questions generate the most calls or emails today?","Who owns the authoritative rule and form content for each agency?","Is GovCloud or FedRAMP hosting a requirement for public-facing AI in your state?","How will you measure answer accuracy and citation fidelity before and after launch?"],"valueHypothesis":"A scoped, citation-grounded navigator may reduce time residents spend locating the right agency and form, to be confirmed with local baseline call and search data.","potentialEngagement":"A bounded discovery and prototype on one agency's licensing rules, with an evaluation harness for accuracy, citation fidelity and refusals.","unsupportedClaims":["Any measured reduction in calls, time or cost (none reported)","Accuracy rates for SAM","That all model inference runs inside GovCloud","Guaranteed savings"]},"engineering":{"fit":"Good fit for RAG over statutes, rules and forms on Amazon Bedrock in AWS GovCloud where public, no-login access is acceptable.","architectureIntegration":"Public web front end, retrieval index over administrative code and agency documents, model inference in GovCloud, citation rendering, and links out to agency application systems.","prerequisites":["Machine-readable rules and forms corpus","Agency content owners and refresh process","GovCloud account and model access approvals","Evaluation question set with expected citations"],"constraints":["Model availability differs in GovCloud","Unclear whether external commercial models are also called","Legal-advice disclaimers required","Public records exposure of chat content"],"security":"Avoid collecting PII, log prompts under records-retention rules, and document the FedRAMP boundary and every model endpoint used.","proofOfValue":["Measure citation fidelity on a fixed question set","Run refusal tests for legal-advice and off-topic prompts","Verify all inference endpoints sit inside the stated boundary","Track rule-change-to-index latency"]},"delivery":{"work":"Stand up a scoped regulatory navigator for one agency domain, with corpus pipeline, evaluation suite and public AI disclosure notice.","dependencies":["Agency rule and form owners","State security office approval","Hosting account in GovCloud","Accessibility and language review"],"ownership":"State efficiency or digital-services office, with each agency owning its content.","skillsAdoption":"Train agency content owners on corpus updates and staff on interpreting cited answers.","governanceCheckpoints":["Legal review of disclaimers","Security authorization of hosting boundary","Pre-launch accuracy and refusal evaluation","Quarterly content-currency review"],"acceptanceCriteria":["Proposed: at least 95% of test answers cite the correct governing rule","Proposed: zero legal-advice responses in the refusal test set","Proposed: rule changes indexed within an agreed number of business days"],"risks":["Stale rules producing wrong guidance","Over-reliance on answers as legal advice","Undisclosed model routing outside the authorized boundary"]}},"retrievedAt":"2026-10-29T22:03:04-05:00","enrichedAt":"2026-10-29T22:16:30-05:00","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"The deck positions SAM for both residents and public servants navigating policies, records and requirements, implying staff self-service use as well as public use; no accessibility testing is described.","procurementImplications":"Buyers should request the hosting boundary, model list, FedRAMP authorization scope and evaluation results before treating this as a reference architecture; the deck alone does not establish contract terms or cost.","operatingModelImplications":"Content currency depends on keeping the regulation corpus synced with rule changes; ownership of corpus maintenance across agencies is not described.","sourceVerification":{"openedUrl":"https://dir.texas.gov/sites/default/files/2026-10/Making%20Government%20Navigable%20A%20Texas%20Use%20Case%20for%20Citizen-Facing%20AI%20-%20Vulcan.pdf","referenceExcerpt":"Texans can ask a question and find out what to do next and which agency to contact.","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}