{"resourceId":"kansas-city-counties-granicus-gxa-bedrock-assistants-axios-20261021","versions":[{"version":"external-9d1ff6a63aa052abf2c930cb4689106611581537d7bf80a6c7089537a0c658c2","resource":{"id":"kansas-city-counties-granicus-gxa-bedrock-assistants-axios-20261021","title":"Axios: Wyandotte and Jackson counties add Granicus's Bedrock-based Government Experience Agent; vendor speed claims stand, but the UG did not answer questions","organization":"Unified Government of Wyandotte County/Kansas City, KS; Jackson County, MO; Granicus","sector":"Local government (resident services)","geography":"Kansas City metro, Kansas and Missouri, USA","publishedAt":"2026-10-21","publicationDate":"2026-10-21","eventDate":null,"sourceName":"Axios Kansas City","sourceLabel":"Axios Kansas City — independent local reporting","sourceUrl":"https://www.axios.com/local/kansas-city/2026/10/21/wyandotte-jackson-counties-adopt-ai-assistants-agent-granicus","evidenceClass":"independent-reporting","outcomeClass":"mixed","topics":["knowledge-work","governance-procurement","accessibility-workforce","data-security"],"finding":"Axios reports that the Unified Government (UG) of KCK/Wyandotte County added Granicus's AI chatbot to its Treasury page, which receives about 70,000 calls a year. Jackson County uses the same Government Experience Agent (GXA) for Assessment questions, and Olathe uses it on its home page. The speed and timing figures come from Granicus: chats take about 1.3 minutes versus about 2.5 minutes for typical UG 311 calls, and about 36% of questions arrive between 6pm and 9am. The reporter found it unclear where feedback goes, the UG did not respond, and both the UG and Granicus store some chat data. AWS's February 2026 partner post says GXA's conversational intelligence is powered by Amazon Bedrock with RAG over agency-approved content. It reports Jackson County basic inquiries fell about 10% within months and Wyandotte targets a 30% Treasury call reduction.","sledRelevance":"A live, multi-jurisdiction example of an AWS Bedrock-powered resident assistant from a major govtech vendor. Cross-tagged to local-government and citizen-engagement because the deployments are county resident-service channels.","evidence":"Independent reporting, but the quantitative figures are attributed to Granicus company data, not county measurement. The AWS link and the 10%/30% figures come from vendor sources (AWS partner blog, Feb 2026; Granicus customer story), both inspected. Granicus's own story notes that testing found an incorrect phone number pulled from an old press release. No independent accuracy audit is reported, and the outlet's question about the UG's results went unanswered.","architectureImplications":"SaaS assistant (Granicus) on Amazon Bedrock with RAG over scraped agency web content and admin-controlled refreshes. Answer quality depends on website content hygiene, as the stale-press-release error shows.","governanceImplications":"Feedback paths and data retention were not transparent to a reporter, which signals a need for published AI-use notices, feedback routes and data handling statements for resident-facing assistants.","securityPrivacyImplications":"Granicus says personal information is redacted and conversations are not used for model training or monetized. Axios notes chat and cookie data are stored and used by both the UG and Granicus.","caveats":"AWS involvement is established through vendor sources, not the Axios article, which does not mention AWS. Call and time figures are vendor-supplied aggregates 'across many government agencies'. The 70,000 Treasury calls (Axios) and 310,000 total calls (AWS blog) describe different scopes. Outcomes for the UG are a goal, not a result.","streamIds":["aws","local-government","citizen-engagement"],"roles":{"sales":"Sales framing: county leaders will hear about this peer deployment, so be ready to separate what is measured from what is marketed. Jackson County's roughly 10% drop in basic inquiries is a modest, vendor-reported result. The speed figures are Granicus aggregates, and the UG's 30% target is not yet an outcome. Use it to discuss content readiness, feedback handling and the metrics a county would want to verify itself. Do not claim call-deflection rates or accuracy beyond these sourced figures.","engineering":"Engineering framing: this is a SaaS RAG assistant on Amazon Bedrock that answers from scraped agency content, so content quality is the main accuracy control. Jackson County's testing caught a wrong phone number from an outdated press release. Integration work centers on source selection, refresh schedules, redaction and analytics export. Ask the vendor for retention settings, the feedback data flow, the model-change process and evaluation results, and confirm the redaction approach for PII entered by residents.","delivery":"Delivery framing: success depends on departments owning their web content and on a visible feedback loop, which Axios could not find. Before launch, run staff side-by-side testing on real resident questions, as Wyandotte describes. Publish an AI notice that explains data use and how to reach a human. Track call volume, chat containment and correction requests against a pre-launch baseline, and set a review date to compare results with the 30%-style goals."},"roleDetails":{"sales":{"customerProblem":"Counties face high volumes of repetitive resident questions on taxes and services, small call-center staffs, and multilingual populations.","stakeholders":["County administrator","Treasury/assessment director","Communications/web manager","County CIO","Language access coordinator"],"discoveryQuestions":["What share of calls are repeat questions already answered online?","How current and consistent is your website content?","What baseline call metrics do you have today?","How will residents give feedback or reach a person?"],"valueHypothesis":"A grounded assistant could absorb some routine questions and after-hours demand, if web content is maintained and results are measured against a baseline.","potentialEngagement":"Content-readiness assessment and a measured pilot on one department's pages with baseline and post-launch call metrics.","unsupportedClaims":["That counties will achieve 30% call reduction","That chat answers are accurate in all cases","That the 1.3-minute figure applies to a specific county","That resident data is not retained"]},"engineering":{"fit":"Fits counties with decent web content and a SaaS-first approach. Weak where content is outdated or scattered.","architectureIntegration":"Vendor-hosted RAG on Amazon Bedrock over agency web pages and documents, with admin-console refresh, analytics, and translation.","prerequisites":["Content inventory and cleanup","Defined escalation to human channels","Data retention and redaction requirements"],"constraints":["Accuracy bounded by source content quality","Vendor controls the model and infrastructure","Limited visibility into model changes"],"security":"Require PII redaction, no training on resident data, defined retention, and access controls on chat logs.","proofOfValue":["Test set of real resident questions graded by staff before launch","Check answers cite current pages","Verify feedback reaches an owner","Compare call volume to baseline after 90 days"]},"delivery":{"work":"Clean content, configure sources, run staff testing, publish an AI notice and feedback route, and monitor metrics.","dependencies":["Department content owners","Vendor configuration support","Call-center baseline data"],"ownership":"Department head for the content domain, with the county communications or digital services team owning the assistant.","skillsAdoption":"Train content owners to update pages and review flagged answers. Train call-center staff on escalations.","governanceCheckpoints":["Pre-launch accuracy review","AI notice and privacy statement published","90-day metrics review","Annual vendor data-handling review"],"acceptanceCriteria":["Proposed: at least 95% of graded test answers correct before launch","Proposed: measurable reduction in routine calls versus baseline at 90 days","Proposed: every feedback item acknowledged within a set timeframe"],"risks":["Stale content producing wrong answers","Opaque feedback eroding trust","Vendor metrics substituting for local measurement"]}},"retrievedAt":"2026-10-30T22:10:30-05:00","enrichedAt":"2026-10-30T22:17:58-05:00","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Translation into 30+ languages and after-hours access are the main accessibility benefits claimed. Staff time freed from routine calls is the workforce rationale, and the UG had reassigned staff to answer phones.","procurementImplications":"Counties are buying AI through existing govtech SaaS vendors rather than directly from AWS. Contracts should specify data retention, feedback handling, accuracy testing and reporting of call-deflection metrics.","operatingModelImplications":"Content owners in each department must keep web pages current, because the agent answers from them. Someone must own and act on resident feedback.","sourceVerification":{"openedUrl":"https://www.axios.com/local/kansas-city/2026/10/21/wyandotte-jackson-counties-adopt-ai-assistants-agent-granicus","referenceExcerpt":"chats with the Granicus bot take about 1.3 minutes across many government agencies, per company data","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}