Seattle scrutiny links AI-assisted triage to gaps in downstream ambulance accountability

Bring together the EMS medical director, dispatch manager, contract owner and patient-access representatives to examine where accountability ends after a transfer. Ask who measures transport waits, who sees repeat calls and how a patient returns to urgent handling. Offer a bounded handoff and evidence assessment before discussing expansion of AI assistance. The value hypothesis is visibility into unresolved patient journeys and clearer service responsibility. Do not promise better outcomes from retained human control or attribute reported harm to AI without evidence. Discovery should establish access to records and provider cooperation; the problem may require contract and workflow changes more than new software.

Design a read-only event trail across telephony, AI prompts, dispatch decisions, nurse disposition and transport. Prerequisites include approved patient linkage, synchronized clocks, accessible vendor logs and an adjudicated outcome definition. Protect recordings and minimize identifiers in analytical datasets. Test failed transfers, repeat callers, delayed transports and AI outages while preserving existing emergency protocols. A proof of value should account for every disposition and report missing timestamps, escalations and time to actual care against a prespecified baseline. Evaluate language and speech differences. The article supplies no evaluation protocol, so local clinical oversight must define tests and stopping rules.

Assign a clinical quality owner and contracting lead to map the complete caller journey with dispatch and transport partners. Establish data-sharing dependencies, audit current timestamp coverage and train staff on escalation after a failed diversion. Governance checkpoints should address transparency, data reuse and whether contract terms permit meaningful oversight. Proposed acceptance includes traceable dispositions, a reviewed process for overdue responses, tested return to urgent handling and independently assessed patient-access measures. These are proposed criteria, not observed improvements. Track adoption and justified overrides without pressuring staff to increase diversion. Risks include incomplete records and moving patients beyond the agency's operational visibility.