In an August 8 response, the Orleans Parish Communication District clarified that its AI-assisted 911 triage handles only duplicate reports of known vehicle crashes when human call takers are busy, limiting its use rather than broadly replacing them. The system checks whether a caller is within 200 meters of an already reported crash, while calls outside that use case default to a human.
The Orleans Parish Communication District (OPCD) has clarified that its AI-assisted 911 system is limited to duplicate reports of already-known motor vehicle crashes, rebutting broad claims that New Orleans is using AI to replace human emergency call takers. In an August 8 response, the agency stated that the tool is activated only when all call takers are busy and an incoming caller is within a 200-meter radius of an already-reported crash.
The clarification addresses an August 6 Shreveport Times article, which OPCD described as inaccurate because it created the impression that AI was being used across all emergency-call categories. OPCD said the article did not contact the agency and provided no attribution or sourcing for its claims.
A constrained duplicate-call workflow
According to OPCD's statement, the system is not used for non-crash calls. The agency stated that its default behavior is to route a call to a human when the request does not meet the narrowly defined crash-related use case.
OPCD also stated that it has used the tool for the past two years and has not applied it to other categories of 911 calls. The agency's response states it will not extend use beyond the current scenario unless it can ensure the system's functionality.
Fox 8 reported on July 16 that OPCD began working with Israeli emergency-communications company Carbyne in 2022 amid call-volume surges and staffing shortages. Executive Director Karl Fasold told Fox 8 that calls are answered in the order received, meaning dispatchers cannot know whether an incoming request concerns a life-threatening emergency before it is answered.
The reported workflow is designed around the repeated calls generated by a single traffic accident. The AI voice asks whether the caller is reporting a specified crash location, then provides immediate acknowledgement in applicable cases, according to Fox 8. Fasold told the station that the system allows human call takers to handle more urgent calls rather than repeatedly collect introductory details on a known collision.
Vendor-reported results and remaining evaluation questions
Carbyne's published New Orleans case study reports that OPCD handles more than 1,000 emergency calls daily and had roughly one-third of its authorized and budgeted intake positions unfilled, citing Fasold. The company reported that its AI-V Emergency Call Triage processed more than 3,500 events over 90 days, or more than 40 per day, and reduced redundant calls by more than 30% among calls it triaged.
Fox 8 separately reported Fasold's statement that staff listened to 100% of calls handled during the first three months and found no false positives or false negatives. That is an agency-reported result, not an independent safety evaluation; the report does not provide the reviewed-call denominator, error definitions, or validation method, and the retrieved materials do not publish performance after that initial period.
The distinction between duplicate-incident triage and open-ended emergency-call handling is consequential. In high-stakes dispatch systems, a tool that confirms a known crash within a geofenced radius has a much narrower decision surface than one attempting to interpret arbitrary medical, fire, crime, or public-safety emergencies. For practitioners evaluating similar systems, measurable escalation behavior, location-match precision, audit trails, and human fallback performance are central safeguards, particularly when staffing constraints create pressure to automate queue management.
Key Points #
- 1OPCD limits AI triage to duplicate crash reports near known incidents when human call takers are unavailable, rather than all emergency calls.
- 2Carbyne reports more than 3,500 triaged events and a 30% reduction in redundant calls, while OPCD's initial no-error claim lacks published independent validation.
- 3Comparable emergency automation requires rigorous measurement of escalation errors, geolocation precision, auditability, and human fallback performance before broader deployment.
Scoring Rationale #
This is a notable, safety-critical deployment of AI-assisted triage in a large urban emergency communications system. Its constrained use case limits its immediate breadth, but the reported workflow and evaluation gaps are relevant to practitioners building human-in-the-loop systems for high-consequence operations.
Sources #
Primary source and supporting public references used for this report.
Practice with real Telecom & ISP data
90 SQL & Python problems · 15 industry datasets
[Active Residential CustomersEasy](/problems/sql/active-residential-customers)
[Unlimited Fiber Plans 500Mbps+Medium](/problems/sql/unlimited-fiber-plans-above-500mbps)
[Customer Churn Risk AssessmentHard](/problems/sql/customer-churn-risk-assessment)
250 free problems · No credit card
See all Telecom & ISP problems