DENVER ‒ In one of thelast places where a real live personanswers the phone immediately, AI is now playing an increasingly important role in helping 911 dispatchers speed police, firefighters and paramedics to emergencies.
Themost sophisticated AI systemscan listen and interact with callers, in some cases by handling non-emergency calls while staying alert for something that's actually a crisis. Other AI systems can automatically translate languages to make sure every caller gets the help they need immediately.
Only in rare circumstances are the AI systems directly answering 911 calls, but are instead more often being used to lighten the high-pressure load on trained dispatchers who are often working inshort-staffed operations centers.
AI helping 911 dispatchers send help faster
"When someone calls 911 they don't want to end on hold. But how do you do that in this day and age?" said Stephen Kennedy, the 911 coordinator for Sumpter County in central Florida, which handles about 80,000 calls annually.
The use of AI by emergency services reflects its growing use across society, from analyzing medical images to detecting fraud and navigating driverless vehicles.
911 a victim of its own success?
For generations, 911 operators have been the backbone of the American emergency response network, answering calls for help, and then dispatching and coordinating police and firefighters, SWAT teams, mental-health counselors and animal-control officers. In a sense, 911 is a victim of its own success: It's such a trusted system that everyone knows to call it when they need help.
And call we do: Each year, 911 operators nationwide answer about 240 million calls. But those same operators are often also responsible for simultaneously answering non-emergency calls, which significantly outnumber actual emergencies. In New York City, that's the 311 line. In Los Angeles, it's 877-ASK-LAPD.
While usually not a crisis, those non-emergency calls are nevertheless still important, as people call in to complain about everything from parking problems to barking dogs, fireworks, fender-benders and stolen bikes, missing cats and abandoned cars. They call to request police reports or an extra patrol in their neighborhood. They ask for help installing car seats or changing smoke detector batteries.
Aurelian, one of the companies offering AI assistance to dispatchers, found that 64% of all calls coming into dispatch centers are not actual emergencies, and that 70% of those non-emergency calls can be handled by AI without the need for human participation. In New York City, people called 311 more than 1.3 million times in March alone.
Longtime police radio provider Motorola Solutions is also offering AI-assisted dispatch services, and body-camera company Axon offers AI-powered cameras that can provide on-the-fly translation services. At the Denver 911 center, workers say they're exploring several options but have questions about cost and reliability.
Max Keenan, the CEO co-founder of Aurelian, said emergency dispatchers are expected to handle calls about barking dogs or broken water lines while also standing by to respond to a police shooting or a terrorist attack. Switching between routine calls and emergencies is exhausting, he said.
Burnout is common among dispatchers. A 2023 study by the International Academies of Emergency Dispatch found a 25% vacancy rate in dispatch centers nationwide, and Keenan said 80% of new dispatchers quit within two years. Aurelian's systems are serving about 5 million Americans daily, from Tennessee's Hamilton County to Kalamazoo, Michigan and Grant County, Washington.
"We try to approach this as, 'how can we help them spend more time on those emergency calls?'" Keenan said. "Because you basically train your 911 operators to be Navy SEALS and they spend 70% of their time being mall cops."
Advertisement
Keeping humans in the loop for 911 centers
At theNational Emergency Number Association, staff are keeping a close eye on how dispatch centers are using AI - and how the public is responding, given publicity around the failings of rushed-to-market AI systems blamed for "hallucinating" facts or inexplicably providing incorrect information. Cost is another factor.
April Heinze, NENA's vice president for operations and standards, said it remains important that someone calling 911 to report an emergency hears a human voice on the other end of the line: reassuring, competent, empathetic.
"People to want to talk to another person," she said.
Heinze said some dispatch centers are experimenting with using AI systems to triage the massive volume of calls 911 receives when there's a significant public event, like a major car crash on a busy interstate or a wildfire sending up a towering column of smoke. Under those circumstances, 911 operators need to focus on dispatching emergency services to the scene but often end up answering hundreds of identical 911 calls from good Samaritans.
Related:The US is waging AI-assisted war on Iran. Here's how
Using AI, a dispatch center could temporarily have all 911 calls from that specific area answered by a computer that asks the caller if they're simply reporting that crash or fire, whether they have specific knowledge that first responders need to know, or whether they have a different emergency, Heinze said.
"When people don’t know what to do, they call law enforcement and 911," she said. "These dispatch centers are really resource managers."
Keenan, the CEO of AI company Aurelian, said in a recent power outage at a dispatch center they assist, about 600 people called a non-emergency line within a two-hour period to report that they'd lost power. But buried in that avalanche of calls was a person seeking help because a lifesaving medical device they depended on had stopped working. The AI system recognized that person needed more immediate help and connected them with a 911 operator, he said.
In a recording of a different non-emergency call demonstrated for USA TODAY, the Aurelian system answered a call from a man reporting his neighbor's barking dogs. The system interacted with the man in much the same way a 911 operator would have, collecting information about addresses, how often the barking occurred and whether he wanted a police officer dispatched.
AI 911 systems speeding help to those who need it
In Florida's Sumpter County, Kennedy, the 911 coordinator, said dispatchers have been pleased with the AI system they're using to automatically transcribe and summarize every incoming 911 call. That allows other dispatchers to share that information in near-real-time with first responders racing to the scene of a crash or a shooting, he said.
The Motorola system can also flag keywords like "cardiac arrest" or "shooting," allowing other dispatchers to get first responders moving even while 911 operator is still collecting information, he said.
He said realtime translation services have also been successful: Depending on what foreign language someone is speaking, it can take five or even 10 minutes to track down a human translator. With AI, he said, people can just speak in their own language and the AI automatically translates it.
"You can imagine the anxiety and stress for both the call taker and the caller while you're trying to let them know help is on the way," Kennedy said. "In 10 years from now, you're going to find everyone is using AI to make things more efficient."
This article originally appeared on USA TODAY:AI is helping 911 dispatchers respond to emergencies
