New Orleans is among the first major American municipalities to integrate an AI voice agent into its 911 emergency call workflow as a limited triage tool.
Overseen by the Orleans Parish Communication District (OPCD), the system — known as AI Emergency Call Triage and powered by software from Carbyne — has been in use on the 911 line since approximately 2023. This is following development work that began around 2022, the department added.
A separate AI agent for the city’s non-emergency 311 service was planned and began rolling out later, around spring 2026.
OPCD reportedly handles more than 1,000 emergency calls per day. The AI tool was designed to reduce workload pressure on human dispatchers during high-volume periods, particularly by filtering redundant reports of the same incident. It functions strictly as a triage layer rather than a replacement for human operators.
Under current protocol, the AI agent activates primarily when all human call-takers are busy and an incoming call originates within roughly 200 meters of a previously reported incident (most commonly a traffic collision). The system asks whether the caller is reporting that known incident.
If the caller confirms they are, the AI can provide reassurance or updates that help is already on the way. If the caller indicates a different or new emergency, or requests a human, the call is immediately routed to an available dispatcher. All calls remain logged and recorded.
OPCD Executive Director Karl Fasold has since defended the limited deployment, noting that 911 centers nationwide face staffing shortages and budget constraints. Filtering duplicate calls during major incidents frees human staff to focus on life-threatening or novel emergencies, he argued.
While other cities, including Atlanta and Seattle, have similarly adopted AI tools for supporting functions such as precise caller location, real-time translation, or dispatcher prompts, New Orleans stands out for using a conversational AI agent that can directly field a narrow subset of active 911 calls under the specific conditions described above.
However, public-safety advocates have voiced growing concern through the media, questioning how well the AI handles panicking callers, heavy accents, ambient noise, or covert distress calls like domestic violence. Critics also cite cybersecurity vulnerabilities and the risk of hallucinated or inaccurate responses during fast-moving crises.
“I was a 911 dispatcher for 11 years and this is the worst idea you could possibly think of,” warned one online critic and alleged former dispatcher. “You have scenarios where the person can’t speak well [or have strong accents] — or if they’re in a domestic situation and they pretend to call for pizza or something — you have to be able to determine how someone feels or pick up on what they aren’t saying.”
Nonetheless, OPCD officials maintain that the tool operates under strict conditions and oversight, has performed reliably in its limited role, with early monitoring showing no false positives or negatives in audited periods, and is not used for other categories of 911 calls.