{"resourceId":"kait8-craighead-county-lezle-ai-nonemergency-56pct-call-load-30-days-20261029","versions":[{"version":"external-aaf72d37d0ec6cbbf02462cd108fbf361daaa032fb3d2eed3e7015cef84f4b12","resource":{"id":"kait8-craighead-county-lezle-ai-nonemergency-56pct-call-load-30-days-20261029","title":"Craighead County, Ark., AI non-emergency call-taker 'Lezle' answered nearly 5,000 calls in 30 days and cut dispatcher call load 56%, with early misroutes","organization":"Craighead County / Jonesboro E-911 (Arkansas)","sector":"911 emergency communications and dispatch","geography":"Craighead County (Jonesboro), Arkansas, United States","publishedAt":"2026-10-29","publicationDate":"2026-10-29","eventDate":null,"sourceName":"KAIT (K8 News, Jonesboro)","sourceLabel":"K8 News: AI dispatcher Lezle: what's working, and what is it still learning?","sourceUrl":"https://www.kait8.com/2026/10/30/ai-dispatcher-lezle-whats-working-what-is-it-still-learning/","evidenceClass":"independent-reporting","outcomeClass":"mixed","topics":["operating-model","accessibility-workforce","knowledge-work"],"finding":"K8 News reports that 'Lezle', an AI agent answering Craighead County non-emergency calls for nearly two months, answered nearly 5,000 calls from Sept. 29 to Oct. 28 and cut dispatchers' call load by 56% over the past 30 days. Dispatchers say it resolves more than half of calls itself, mainly clerical matters such as records or warrants. E-911 dispatcher Terra Summers describes early glitches, including Lezle telling a caller to file a car accident report online when an officer should be sent. Errors are reported to administrators or the programmer. Trigger words such as 'emergency', 'breathing' or 'fire' transfer callers to a live dispatcher.","sledRelevance":"A rare U.S. follow-up with a month of operating numbers and a named failure mode for a non-emergency AI call-taker, after the archive's September 2 report that Jonesboro was only testing it.","evidence":"Local TV report quoting a dispatcher and a resident. The 56% and nearly 5,000 figures are operator-reported without method, baseline or error count; the '70 hours' quote is not explained. The accident-report misroute is described anecdotally with no count of similar errors.","architectureImplications":"Keyword-triggered transfer to a live dispatcher sits in front of the AI; corrections go through an administrator-or-programmer loop rather than a formal change process.","governanceImplications":"The misroute shows non-emergency AI can give unsafe instructions on incidents that need a response; agencies need an incident log, review cadence and a rule for what the AI may never deflect.","securityPrivacyImplications":"Records and warrant questions involve personal data; the source does not describe retention, vendor access or call recording terms.","caveats":"Single local-news follow-up; vendor not named; figures are self-reported for one 30-day window; 'more than half resolved' and '56% call load' may use different denominators. Follow-up to an archived Sept. 2, 2026 KAIT item (different URL) on the testing phase; this record adds the first operating results.","streamIds":["emergency-services"],"roles":{"sales":"For sales, the framing point is that Craighead County now has a month of operating numbers: nearly 5,000 calls and a 56% cut in dispatcher call load, plus an honest admission of early misroutes. Directors will ask both how much load moves and what goes wrong. A bounded engagement is an independent review of containment and misroute data for an existing non-emergency AI line. The report cannot support claims of accuracy, safety or cost savings.","engineering":"For pre-sales engineering, the design is a non-emergency AI agent with keyword-triggered transfer and transfer to the right person for clerical requests. The accident-report misroute shows intent classification can fail on incidents needing a response. Prerequisites are call logging, transfer routing and a defect workflow. Constraints include keyword triggers missing indirect descriptions. Proof of value should audit a sample of AI-resolved calls for wrong deflections, not just count containment.","delivery":"For delivery, the work is running a defect loop: dispatchers flag errors, administrators fix call flows, and harder issues go to the vendor's programmer. Ownership sits with the E-911 center administration. Acceptance should track containment with a stated denominator, misroute rate and time to fix. Risks include unsafe deflection of incidents that need officers, untracked fixes and figures reported without method."},"roleDetails":{"sales":{"customerProblem":"Non-emergency calls, many clerical, consume dispatcher time on a small E-911 floor.","stakeholders":["E-911 director","Dispatchers","County or city administration","AI vendor","Residents"],"discoveryQuestions":["How do you count calls the AI resolves versus transfers?","How many misroutes have dispatchers reported and how fast were they fixed?","Which call types must never be deflected online?"],"valueHypothesis":"Moving clerical calls to AI may reduce dispatcher call load substantially, as Craighead reports 56%, if misroutes are caught and fixed (interpretation).","potentialEngagement":"An independent 30-day audit of AI-resolved and transferred calls with a misroute log review.","unsupportedClaims":["Accuracy or misroute rate","Hours or cost saved","Caller satisfaction","Safety on emergency calls"]},"engineering":{"fit":"Fits PSAPs with high clerical non-emergency traffic (records, warrants).","architectureIntegration":"AI agent on the non-emergency line with keyword-triggered live transfer and routing to the correct office.","prerequisites":["Call logging with outcome codes","Transfer routing to dispatchers and offices","Defect reporting workflow"],"constraints":["Keyword triggers may miss indirect emergencies","Vendor and model not named","Single-county figures"],"security":"Define retention and vendor access for transcripts that include records and warrant questions.","proofOfValue":["Audit sample of AI-resolved calls for wrong deflections","Containment rate with stated denominator","Time from defect report to fix"]},"delivery":{"work":"Operate and tune call flows through a dispatcher-reported defect loop.","dependencies":["Vendor programming support","Administrator access to call-flow configuration","Dispatcher reporting time"],"ownership":"E-911 center administration, with vendor support for model changes.","skillsAdoption":"Dispatchers learn to spot and report AI errors; administrators learn to edit call flows.","governanceCheckpoints":["Written list of incident types the AI must always transfer","Monthly misroute review","Public reporting of containment method"],"acceptanceCriteria":["Proposed: misroute rate below an agreed threshold in audited sample","Proposed: defects fixed within an agreed number of days"],"risks":["Incidents needing a response deflected online","Figures reported without method","Dependence on one vendor programmer"]}},"retrievedAt":"2026-10-29T22:06:52-05:00","enrichedAt":"2026-10-29T22:10:40-05:00","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Dispatchers describe relief on the floor and stay in the loop to report errors; a wary resident accepts AI only if it sticks to non-emergency calls.","procurementImplications":"Buyers should require vendors to report containment with a defined denominator and to fix misroutes within a set time.","operatingModelImplications":"AI handles clerical non-emergency calls while dispatchers act as the error-reporting and escalation layer.","sourceVerification":{"responseId":"curl-lezle-2026-10-29","openedUrl":"https://www.kait8.com/2026/10/30/ai-dispatcher-lezle-whats-working-what-is-it-still-learning/","referenceExcerpt":"From Sept. 29 to Oct. 28, Lezle answered nearly 5,000 calls.","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}