The Robot That Answers 911 Is a Symptom
New Orleans is putting a machine on the 911 line. The headline says the city “will use AI to answer 911 calls instead of a human.” Read that and you picture an algorithm fielding a cardiac arrest while a human plays Solitaire. The real story is weirder — and more human — than that.
The Orleans Parish Communication District is the first agency in the country running Carbyne’s “Call Triage” software. Call 911 within 100 meters of an incident already in the dispatch system — a three-car pileup, say — and a bot asks if you’re calling about that crash and whether you’re involved. A witness reporting what dispatch already knows gets told to hang up. Involved, or something new? You get a human, same as always.
That’s the whole machine: a spam filter for redundant calls. It isn’t diagnosing chest pain or deciding who gets an ambulance. It answers so a human doesn’t have to say “yes, ma’am, we know about the accident” forty times an hour.
The reason it exists should make you angrier than any robot. The center is short 18 call takers out of 140 — a third of budgeted positions unfilled — and a survey found 82% of centers report chronic staffing shortages. New Orleans is supposed to answer 90% of calls within 15 seconds. It’s hanging at 70-80%.
The stat that tells you everything: they once hit that 90% standard. For four months. By forcing mandatory overtime. Then people quit. The city chose burnout over hiring, and when that stopped working, it bought software.
Now the counterarguments, because they’re real.
“The system is already failing. A busy signal saves nobody.” True. A bot that answers in two seconds beats a human who answers in forty-five. The director estimates triage does the work of two full-time call takers. When you can’t fill the seats, a filter beats a queue.
“It’s just a filter — humans still take the real calls.” Today. It handles only motor-vehicle accidents, one at a time. The director says he’s ready to expand to multiple incidents, then to other nonviolent call types. That’s the playbook: every scope expansion sounds reasonable at the time, and each one moves the line for what “needs a human” means. The question was never whether a bot can triage car wrecks. It’s where the line gets drawn, and who draws it when the budget’s due.
“What happens when it gets it wrong?” Its failure mode is telling someone “you can hang up” when they shouldn’t. A human who makes that mistake has a name and a supervisor. A machine has a vendor and an SLA. When the machine is wrong in a way that costs a life, there’s no one to fire — which is precisely why it’s attractive to the people signing the checks.
Here’s where I land. The triage bot isn’t the villain; it’s a symptom. The villain is a funding model that would rather automate the answering than fund the answerers — that treats burnout as a scheduling problem and software as a cheaper substitute for a living wage. Duct tape is fine on a fence. It’s a different story when it’s holding up a load-bearing wall.
I’d change my mind if the bot shrinks headcount instead of filling the gap, or if it creeps from “we already know about that crash” into deciding who gets a human response at all. The day a machine decides you don’t get help is the day it should have to answer for it. Machines don’t answer for things. That’s the problem.
The robot that answers 911 was never the story. The story is that a city looked at the humans it couldn’t afford and bought a machine instead. The machine will do fine. The humans were already gone.
Sources: StateScoop, Shreveport Times, GovTech, Verite News