Artificial intelligence is moving into a place where most Americans probably never expected to encounter it: the 911 system.
New Orleans is introducing AI into its emergency communications operation, using the technology to identify duplicate reports and determine which callers need to be connected with a human dispatcher.
The Orleans Parish Communication District, which handles roughly 1,000 calls per day, says the objective isn’t to replace emergency dispatchers with chatbots. Instead, the technology is supposed to keep dispatchers from repeatedly handling calls about incidents they already know about.
New Orleans began using AI on its non-emergency 311 service this spring. Now the concept is being extended to emergency communications.
Carbyne’s AI Emergency Call Triage system, for example, can identify what it believes are multiple calls concerning the same incident. A caller reporting something already known to authorities can receive an automated response, while someone reporting a new incident is routed to a person.
OPCD Executive Director Karl Fasold says the problem becomes particularly obvious when call volumes suddenly spike.
“If somebody calls about a relative having a heart attack, when that call is coming in, we don’t know what it is,” Fasold told WVUE-TV. “Calls get answered in the order that they’re received.”
When every dispatcher is occupied, the next caller waits.
That can produce an unpleasant situation for everyone involved. The person experiencing an emergency is stuck in a queue while dispatchers move from one call immediately into another.
And many of those calls may concern exactly the same event.
Fasold said a single automobile accident can generate 20 calls during rush hour. Hiring enough dispatchers to handle every temporary surge isn’t realistic, he argued.
“You can’t really staff for those occurrences of surges,” Fasold said. “It’s impossible, both fiscally and practically.”
The AI system attempts to solve that problem partly through location data.
“What we have the AI doing is looking to see if you’re calling within 200 meters of somewhere we know that there’s an accident already,” Fasold explained.
If the system determines that the caller is likely reporting that known accident, AI can respond immediately instead of consuming another dispatcher’s time.
There is a practical appeal to the idea. If 20 people are calling about one crash, there isn’t much value in forcing 20 different conversations with emergency dispatchers while someone calling about a heart attack waits.
The obvious concern is what happens when number 21 isn’t calling about the accident.
Maybe that person happens to be within 200 meters of the wreck but is reporting an unrelated emergency. Maybe the circumstances have changed dramatically since the original crash report. Maybe the algorithm simply gets it wrong.
With 911, those aren’t minor technical glitches. Seconds can matter.
New Orleans isn’t alone in testing where AI belongs in emergency services.
Seattle has been using AI-assisted technology for approximately two years. The Seattle Fire Department’s system can prompt dispatchers to direct certain patients to a nurse-staffed call center in Texas rather than sending them through the conventional emergency-response process.
Seattle officials emphasize that trained dispatchers retain the final decision.
“The dispatcher still has the ultimate authority,” Seattle Fire Department Assistant Chief Chris Lombard said.
Corti, the company providing Seattle’s technology, makes the same point.
“Every determination on a 911 call … is made by a trained SFD dispatcher following SFD’s protocols,” a company representative said. “Corti’s role is to support that work, not replace or override clinical judgment.”
Still, University of Washington law professor Ryan Calo has raised another issue: Does the caller know AI is involved?
“The potential that this company could be a part of the experience that Seattle 911 callers have and they don’t know it, that raises serious concerns,” Calo said. “I’m troubled on a number of levels.”
That transparency question will become increasingly important as these systems spread.
There is a substantial difference between using AI behind the scenes to organize information for a dispatcher and allowing an automated system to influence whether a caller reaches a dispatcher at all. New Orleans says its system is designed to recognize duplicate incidents rather than independently manage emergencies, but even that limited role requires the technology to make consequential judgments about incoming calls.
The potential benefit is easy to see. Nobody wants a person suffering a medical emergency waiting in line behind the 19th report of the same traffic accident.
The risk is equally easy to see.
An algorithm doesn’t need to replace the 911 operator to become an important part of an emergency response. It only needs to make the wrong routing decision at exactly the wrong moment.
New Orleans is betting that AI can clear away enough duplicate calls to get human dispatchers to genuine emergencies faster.
For the person dialing 911, however, the most important feature of the system will remain decidedly old-fashioned: When something truly is wrong, a human being needs to be reachable.
