The news
In early August, New Orleans officially confirmed what a viral post had brought to light: some 911 calls are now answered by an artificial intelligence. The confirmation came on August 6, after a post by the More Perfect Union account on X drew more than 32,000 views and reignited the debate over AI's growing role in emergency response. According to TechTimes' reconstruction, the system has been in use for about three years without callers being notified.
The story, however, is more nuanced than it was told. The AI is not replacing all 911 operators. The system — called AI Emergency Call Triage and developed with emergency communications company Carbyne — handles a very specific category of calls: redundant, low-priority ones, received when several people report the same incident.
How it works
The mechanism is simple. When a call arrives, the platform retrieves the caller's GPS coordinates through NG9-1-1 location standards. If the caller is within 200 meters of a traffic accident already logged in the dispatch system, the call is routed to the AI voice agent. The agent asks whether the caller intends to report that incident: if so, it assures them the situation is already being handled and asks them to stay on the line only if involved or if they have new information.
Why they do it
The stated goal is to take repetitive work off human operators, leaving them more time for emergencies that truly require human judgment. The Orleans Parish Communication District (OPCD) handles more than 1,000 calls a day — over a million a year. During rush hour, a crash on Interstate 10 can generate dozens, sometimes hundreds of identical calls: a cardiac emergency can sit in queue behind a burst of reports about the same fender bender.
There is also a staffing issue. The average vacancy rate at US 911 centers hovers around 25%, and more than half of centers report a genuine staffing emergency. As OPCD executive director Karl Fasold explained, hiring enough people to cover call surges is impossible, both fiscally and practically. Before reaching 911, the system had been tested on the 311 line — dedicated to non-urgent city services — starting in April.
The numbers behind the triage
According to Carbyne's case study, over a 90-day measurement period the system triaged more than 3,500 events, over 40 per day, cutting redundant calls by 30% and saving up to 16 minutes of dispatcher capacity. OPCD data tell the same story: for every 20 calls about a single accident, the AI handles between 15 and 18. The agent was trained for three months on real 911 recordings, learning local accents and tricky street names like Tchoupitoulas.
Understanding a call is not understanding a situation
An algorithm can understand a call without necessarily understanding a situation. A person in panic may speak incoherently, fail to explain what is happening, convey decisive information through tone of voice, hesitation, context. This is why the question is not simply "AI yes or AI no": the real question is how much work we can delegate to machines before the human factor becomes indispensable.
Concerns are not lacking. Tulane University professors Aron Culotta and Nick Mattei point out that no city or national policy specifically governs AI in emergency response; according to Mattei, the unease also comes from the fact that it is the government deploying it, not a service bought for one's home. Colleague Rob Lalka is blunter: if the system gets it wrong just once, someone's life is on the line. Research also shows AI voice systems remain vulnerable: some studies report successful manipulation rates between 79% and 96%.
Not man against machine, but man plus machine
New Orleans is not an isolated case. Seattle has been using AI on 911 medical calls since December 2023, Atlanta has launched similar dispatch experiments, and no federal law requires cities to disclose AI use in emergencies. The Louisiana city is acting as a laboratory for a transformation that could touch many other public services: not necessarily man against machine, but man plus machine. And this distinction could become one of the most important of the next decade.
Sources
