{"resourceId":"seattle-times-sno911-aurelian-ava-nonemergency-ai-transfers-hold-times-liability-20261031","versions":[{"version":"external-c5991374b135cd4e787b0a7917e74cc9e0f827596b34649cfe624327d1b06c15","resource":{"id":"seattle-times-sno911-aurelian-ava-nonemergency-ai-transfers-hold-times-liability-20261031","title":"Seattle Times: Snohomish County's AI nonemergency agent cut hold times from 10+ minutes to 47 seconds, but missed-emergency counts come only from the vendor and the agency carries liability","organization":"Snohomish County 911 (SNO911); Aurelian (vendor of Ava and Cora); Washington State 911 Advisory Committee RAIN subcommittee","sector":"911 emergency communications / nonemergency call handling","geography":"Washington State, United States (Snohomish County and at least seven other counties)","publishedAt":"2026-10-31","publicationDate":"2026-10-31","eventDate":null,"sourceName":"The Seattle Times","sourceLabel":"Dialing a nonemergency call line in WA? It might be AI that answers","sourceUrl":"https://www.seattletimes.com/business/dialing-a-nonemergency-call-line-in-wa-it-might-be-ai-that-answers/","evidenceClass":"independent-reporting","outcomeClass":"mixed","topics":["operating-model","governance-procurement","accessibility-workforce","data-security"],"finding":"The Seattle Times (Oct. 31) reports that SNO911 has used Aurelian's AI agent Ava on its nonemergency line since November 2024 while 911 calls still go to people; the reporter says the technology is used in 20 or more states and at least eight Washington counties. About 41% of SNO911 calls that begin with Ava are transferred to nonemergency dispatchers and nearly 10% go to the emergency queue; agencies tune a 'resistance' setting that controls transfers. The agency says nonemergency hold times that regularly exceeded 10 minutes fell to an average of 47 seconds over the past year. SNO911 bought Ava without competitive bids and was billed $118,944 for 2026 (about 192,000 projected calls at 61 cents each), with a discount for dispatcher feedback. Under the contract SNO911 bears liability for dispatcher mistakes based on Aurelian recommendations, including the Cora emergency-call co-pilot adopted in January. Aurelian's CEO says Ava mistook an emergency for a nonemergency once (2024) and reports a 5% to 8% word error rate; the article notes those are company-only figures, that no error rates for competing tools (GovWorx NEAL, Versaterm CallTriage, Axon Prepared AI) are public, and that no state or federal law requires reporting missed emergencies. A Washington 911 Advisory Committee subcommittee finished drafting voluntary 'responsible AI in 911' practices in October.","sledRelevance":"It's an unusually detailed public account of one US PSAP's AI nonemergency deployment, covering transfer rates, hold-time change, per-call price, sole-source buying, liability terms and the absence of independent accuracy data. State 911 offices and county boards can use it to set contract and disclosure terms.","evidence":"Independent newspaper investigation that draws on interviews, an invoice and the contract. Hold-time and transfer figures are agency-reported, while missed-emergency and word-error figures come from the vendor. The article reports no independent audit, sample of call outcomes or demographic accuracy data. It cites a 2022 University of Washington study on speech-recognition disparities as context, and that study isn't specific to Ava.","architectureImplications":"The AI agent sits in front of the nonemergency line and uses several models (location cues, attempts to reach a human, emergency detection) to decide when to transfer. A tunable 'resistance' threshold sets how much goes to humans, and a separate co-pilot (Cora) listens to 911 calls and prompts dispatchers.","governanceImplications":"Procurement terms matter. The deal was sole-source, discounts were tied to staff feedback, liability sat with the agency and there were no audit or error-reporting rights. Washington's voluntary RAIN practices and calls for mandatory disclosure of word error rates show governance is still being set.","securityPrivacyImplications":"AI speech recognition processes every caller's voice on the nonemergency line, and the co-pilot listens to emergency calls. The article doesn't describe how recordings are retained or used for vendor training, and it raises accuracy differences across accents and demographics.","caveats":"Single-agency focus. Benefits are agency-reported and error figures are vendor-reported. Aurelian didn't clarify how many Washington counties use Ava. A 47-second average hold time isn't the same as caller outcomes. Interpretation: results may not transfer to centres with different call mixes or staffing.","streamIds":["emergency-services"],"roles":{"sales":"For sales, the framing point is that buyers now have a public reference price and contract terms for AI nonemergency agents: SNO911 paid $118,944 for 2026, bought sole-source and kept the liability. Many PSAP leaders will want help writing better terms, not just a tool. A bounded engagement is a procurement and contract review covering audit rights, error reporting and data use. This article doesn't support claims that AI catches every emergency or saves money overall.","engineering":"For pre-sales engineering, the key control is the transfer threshold. At SNO911 about 41% of Ava calls go to nonemergency dispatchers and nearly 10% to the emergency queue. Prerequisites are call-type baselines, transfer routing and full logging. Constraints are that missed-emergency counts and the 5–8% word error rate come only from the vendor, and accuracy differs by accent. Proof of value should replay sampled calls, measure misses independently and break accuracy down by caller group.","delivery":"For delivery, the work is a governed rollout. Set the threshold deliberately, review transferred and non-transferred calls, and report results publicly. Ownership sits with the PSAP director under the 911 board. Acceptance should include an independently audited missed-emergency rate, hold-time targets and demographic accuracy reporting. Risks include agency liability for AI-influenced errors, vendor-only metrics, caller distrust and staff over-reliance on the co-pilot."},"roleDetails":{"sales":{"customerProblem":"PSAPs short of staff face long nonemergency hold times, with dispatchers juggling routine and life-threatening calls in the same shift.","stakeholders":["PSAP executive director","911 board or county council","Dispatchers and union","State 911 office","Procurement and legal","Vendor"],"discoveryQuestions":["What are your current nonemergency hold times and call volumes?","Who would bear liability for AI-influenced errors under your contract?","Can you require independent testing, audit rights and missed-emergency reporting?","How would you measure accuracy for callers with different accents?"],"valueHypothesis":"An AI nonemergency agent with a tuned transfer threshold may cut hold times sharply, as SNO911 reports, if missed emergencies are independently monitored (interpretation).","potentialEngagement":"A procurement and contract-terms review plus an independent call-sample audit design for an AI nonemergency line.","unsupportedClaims":["AI never misses emergencies","Net cost savings","Equal accuracy across accents","Results at other agencies"]},"engineering":{"fit":"Fits PSAPs where dispatchers also staff high-volume nonemergency lines.","architectureIntegration":"AI voice agent in front of the nonemergency line with a configurable transfer threshold into nonemergency and emergency queues, plus an optional co-pilot on 911 calls.","prerequisites":["Baseline hold times and call mix","Transfer routing into CAD/phone queues","Full call recording and AI decision logs","Contract audit rights"],"constraints":["Missed-emergency and error data are vendor-reported only","No state or federal accuracy standard","Speech recognition accuracy varies by accent and demographics","Agency bears liability under the contract"],"security":"Define retention, vendor training use and access for nonemergency and co-pilot call audio and transcripts.","proofOfValue":["Independent replay of sampled AI-handled calls for missed emergencies","Word error rate broken down by caller group","Hold-time and transfer-rate tracking at different threshold settings"]},"delivery":{"work":"Deploy with a documented transfer threshold, regular audits of AI-handled calls and public reporting.","dependencies":["Vendor configuration and data access","Dispatcher review time","911 board approval"],"ownership":"PSAP executive director, accountable to the 911 board.","skillsAdoption":"Dispatchers learn to handle AI transfers and to avoid over-relying on co-pilot prompts; supervisors learn to audit AI logs.","governanceCheckpoints":["Contract review of liability, audit and data terms","Threshold change approval and log","Periodic independent missed-emergency audit","Alignment with state RAIN voluntary practices"],"acceptanceCriteria":["Proposed: independently audited missed-emergency rate below an agreed threshold","Proposed: hold-time target met for nonemergency calls","Proposed: published word error rate by caller group"],"risks":["Agency liability for AI-influenced errors","Vendor-only metrics","Missed distress in callers who downplay emergencies","Caller distrust"]}},"retrievedAt":"2026-10-31T22:04:58-05:00","enrichedAt":"2026-10-31T22:09:49-05:00","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Dispatchers describe relief from queue backlogs, and leaders frame Ava as filling vacancies. Some agencies pay for it from unfilled dispatcher budgets. Accent and demographic accuracy gaps are an accessibility risk for callers.","procurementImplications":"The $118,944 annual price at 61 cents per projected call, the sole-source award, the feedback discount and agency-held liability give buyers concrete benchmarks and warnings. Experts urge contract rights to independent testing and audit.","operatingModelImplications":"Dispatchers move from answering routine calls to handling transfers and supervising AI, and agencies own the threshold setting that balances automation against human handling.","sourceVerification":{"responseId":"curl-seattle-2026-10-31","openedUrl":"https://www.seattletimes.com/business/dialing-a-nonemergency-call-line-in-wa-it-might-be-ai-that-answers/","referenceExcerpt":"Currently, about 41% of SNO911’s calls that begin with Ava are transferred to dispatchers on nonemergency lines. Nearly 10% are sent to the emergency queue.","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}