The call came in around 11 p.m. The caller was a young girl, not a native English speaker. She dared not speak above a whisper – the shooter was still in her home. The recipient at 911 wasn’t usually a call-taker but was filling in. No additional sets of ears were available to help listen.
In those difficult circumstances 30 years ago, the call-taker never discerned the word “shooting.” Officers were dispatched for the more routine reason of an injured child. They arrived to find the caller shot in the face and her three siblings killed. All were between ages 6 and 12.
It was a disturbing call – both the crime and, for Ed Davis, the response. The miscommunication delayed immediate medical aid and could have let officers walk into catastrophe. Davis, at the time superintendent of that responding police department in Lowell, Massachusetts, looked into what could have been done differently.
“It became readily apparent there was problem in the communication of the information,” recalled Davis, who later became police commissioner in Boston, in a recent Police1 webinar, “From the 911 call to the scene: Transforming situational awareness and officer safety with AI.”
“[The call-taker] was listening very carefully but couldn’t make out some of the things [the girl] was saying … They didn’t even kick the door in. The door was ajar – they were able to walk through, and they found the bodies.”
While this tragedy occurred in 1995, many departments today still rely on the same kinds of voice communications that fell short that day in Lowell. The difference is that today, there are tools that can help.
“AI,” Davis noted, “could make all the difference in the world in that type of situation.”
AI bolsters every phase of the call
Ending situations like this is one of the goals of the advanced AI-powered emergency communications platform offered by Prepared by Axon. It’s designed to improve how 911 calls are processed and responded to, strengthening the accuracy, timeliness and usability of the data that comes to call-takers, dispatchers and officers in the field. Prepared, which has served 911 professionals for half a decade, was recently acquired by Axon.
Prepared brings the benefits of AI to four distinct segments of the call, starting with what it calls the pre-call phase: the non-emergency calls that need not necessarily even reach an emergency telecommunicator. With emergency calls it can help call-takers distinguish keywords and situations like the one missed in Lowell. It also brings key advantages to the dispatcher and quality assurance and analytics portions of the 911 experience.
Voice assistant handles nonemergency calls
The number of 911 centers with dedicated non-emergency call-takers is tiny – around 6% in a recent informal survey by Prepared by Axon. In the other 94% of cases, non-emergency calls are answered by the same people who field actual emergency requests. That means a spike in non-emergency calls, even during a normal 911 call volume, may create delays for those in urgent need.
The Prepared platform’s non-emergency triage function handles calls to 10-digit nonemergency lines automatically via an AI voice assistant. This naturally and dynamically talks to callers in English or Spanish to determine their needs and connect them to assistance. If it identifies an emergency, it shifts the call to a human; otherwise, it forwards their information to the appropriate resource (e.g., roads department, animal control, homeless services, etc.). Call-takers gain more time to focus on true emergencies.
In Fairfax County, Virginia, this is helping solve a problem with non-emergency hold times. Of around 3,000 calls a day to that system, around 60% are nonemergency. Diverting many of those to the AI assistant gives staff time to breathe, Director Scott Brillman says.
Help and highlights support emergency response
When true emergencies occur, AI drives assistive call-taking that provides personnel faster and fuller clarity into what’s happening. Key aspects to that are real-time translation, transcription and keyword insights.
The automatic translation of non-English calls covers more than 30 languages and saves the significant time traditionally required to identify a caller’s native tongue and locate and contact an interpreter.
“Most of the language line services in this country are very slow in responding,” noted Monica Million, ENP, past president of the National Emergency Number Association (NENA) and a former 911 telecommunicator, in the Police1 webinar. “There’s a several-minute delay to get a language unless the initial call responder is a Spanish speaker. And oftentimes in our communities that’s not the language that’s needed.”
In Million’s jurisdiction several years back, a distraught man pulled over and called 911 from the side of the highway. He was panicking and screaming, and she couldn’t identify his language. It was Farsi, and the man needed help for his mother, who was having a heart attack in the car alongside him.
“I had never really been exposed to Arabic languages before, so I didn’t recognize what he was trying to say,” Million recalled. It took the language line service 12 minutes to decipher things, which was too late for the man’s mother, who died by the side of the road.
Schools also don’t teach typing anymore, Million noted, making the real-time transcription another key feature.
“Most of us have lowered our words-per-minute requirements to get people in the door that actually passed the test,” she said. “They’re slower [and] they’re spending time listening instead of typing. It’s a real skill problem. So now during the training session we’re spending valuable time … teaching them to type to shore that up. So transcription to me is a game-changer.”
A major safety gain in the Prepared by Axon platform is the recognition and highlighting of keywords in 911 calls. As those conversations unfold, AI identifies and flags essential data like addresses and plate numbers, as well as important words and phrases – e.g., “burglary,” “gunshot,” “unconscious,” etc. – that may indicate urgency, incident types, threats or context. These flagged terms are surfaced prominently for call-takers, dispatchers and responders so they don’t miss crucial information during rapid, stressed conversations. Linked to Axon Fusus real-time crime centers, they can help trigger actions like camera activations, video requests and drone deployments.
The platform can also ingest next-generation media and text: Call-takers can send callers links that let them provide video or images from the scene.
Dispatch: ‘It’s all right there’
Dispatchers benefit from similar features. For them, assistive dispatch automatically transcribes all radio channels and provides summaries and insights that keep things easy to digest. They are pinged when keywords are detected – ensuring, for instance, that they never miss a mayday. Conversations are time-stamped and searchable.
Sharing the Prepared platform shrinks the figurative distance between call-takers and dispatchers. One major department the company worked with had them on separate floors of its 911 headquarters. If information was missing, dispatchers had to pick up the phone and call the call-takers. Now both groups can scroll both AI-generated call summaries as well as AI insights.
The platform’s dispatch auto-transcription assisted the rescue of four boaters whose craft capsized in Lake Erie in August. They called 911 from seven miles offshore, triggering a multiagency response that included the Coast Guard. While the 911 system’s older location technology placed the caller on shore, Prepared by Axon’s provided actual GPS coordinates for the boat in the lake, and its location breadcrumbs tracked the craft as it drifted two miles before rescuers arrived. With assistive dispatch, telecommunicators tracked every transmission word for word and could refer back as needed. All victims were rescued unharmed.
Postcall review closes the loop
Once a call is complete, all data feeds into automated quality assurance. This can encompass every call, rather than the samples traditionally used. Departments can check calls against their own custom protocols: Did that call-taker/dispatcher say everything they were supposed to? If faced with x, did they do y? This can feed individual scores that allow benchmarking.
Big-picture analytics provide higher-level views into performance. This can inform focused training where it’s needed and compress the timeline for addressing shortfalls.
“The [quality assurance] tool that’s built into Prepared by Axon is phenomenal – I think it can be used great as a training tool and also as a way just to make certain we’re providing the best service to our communities,” said another webinar panelist, former San Diego Police Chief David Nisleit. “One thing to show your community … is, look how quickly now these calls are being answered [and] how the information is being pushed out.
“I think you can show with those metrics how it really is a force multiplier, how it is providing better customer service.”
Everybody benefits
The Prepared by Axon platform uniquely combines the benefits of AI across the spectrum of 911 calls, bringing better situational clarity to officers in the field and faster and more accurate operation to call centers – a rare tool that can benefit both.
Call-takers and dispatchers, noted Million, are “going to have to start leveraging technology to help mitigate some of these gaps … There aren’t enough people to help aid the call-taker [and] dispatcher to have a second ear to listen to a specific incident they already recognize is critical. If the AI tool such as Prepared by Axon is in place, it’s already doing it for you, so you don’t need to leverage another human. I think that’s a huge win for us to get that technology in our centers.”
For officers in the field, added Nisleit, “the more information they can get very quickly while they’re responding to a scene, the better they’re going to be able to perform their tasks, and the better they’re going to be able to keep not only themselves safe, but their communities safe.”
For more information, visit Prepared by Axon.
