At some point soon, a person will phone the nonemergency line for Westcomm Regional Dispatch, and, for the first time, artificial intelligence will answer the call.
Will it be a gun owner looking for information on obtaining a permit to carry? A construction company ordering a police detail to coordinate traffic at a worksite? A resident looking to contact animal control?
Westcomm, which operates emergency dispatch for six communities in Western Massachusetts, is implementing an AI-driven system to handle certain nonemergency queries. It seeks to free up dispatchers to respond to more urgent calls.
The regional dispatch center has been tweaking the system, testing under what situations the AI decides to route a call to a dispatcher. While no exact date is set for the system to go live, officials hope to make the switch in the next two months.
“We really want to make this as close to perfect as possible because you get one shot,” said Anthony Massoia, deputy director of Westcomm. “We want to make sure we do it right because it sets the tone for not only Western Mass., but Massachusetts.”
The center is the first in the region to integrate AI into its emergency dispatch system.
As call volume has climbed, many see artificial intelligence as a tool to help dispatchers better respond to true emergencies.
The regional dispatch provides services to Longmeadow, East Longmeadow, Chicopee, Monson, Ware and the Hampden County Sheriff’s Department. It fielded almost 30,300 calls to 911 this year as of July 1. But nonemergency calls to its dispatchers on its business line dwarf that number at 79,500 calls.
New AI tools in use
Over the last few years, emergency dispatchers on the top floor of its building along Shawinigan Drive in Chicopee have worked with several artificial intelligence tools. These are tools that Westcomm Executive Director Penny Ryan calls “a game changer.” They allow the whole room to gain situational awareness and help dispatchers more fully focus on the people calling 911.
As soon as a dispatcher answers a call (“911, what is your emergency?”) an AI tool begins listening in the background, logging a transcript of the call, flagging key details or keywords. It’s a tool Westcomm has used for about four years.
If the caller speaks a language other than English, the state offers translation services for emergency calls but not for callers reaching out on business lines. Westcomm has an AI tool that allows dispatchers to both communicate in multiple languages by text but also voice by conferencing in a bot
Another tool, which Westcomm started using last month, transcribes radio traffic – a useful tool when law enforcement sends out an alert to be on the lookout for, say, a suspect and the description runs long.
It also uses artificial intelligence to help its human quality assurance team monitor the quality of the calls. Humans review a fraction of calls coming in to see whether dispatchers worded questions well and asked them in the correct order Artificial intelligence provides generic oversight, making sure addresses are verified, for instance
While the price for the technology depends on the number of calls it receives, Westcomm pays about $100,000 for its AI tools, reimbursed by a state 911 grant.
Next generation of 911 dispatch
For the past several years, emergency dispatch systems across the country have been transitioning from decades-old analog infrastructure to “Next Generation 911,” which runs over the internet.
The National Telecommunications and Information Administration, an agency that advises the White House, has said a modern emergency dispatch system will have the ability to integrate AI tools.
“AI is increasingly viewed as a foundational component of (Next Generation 911) modernization, not just a parallel innovation,” the agency wrote in a white paper last September.
An emergency dispatch center in Colorado that deployed AI reduced the number of administrative calls it received by about a third, which in turn increased how quickly dispatchers picked up emergency calls, according to the agency’s paper.
Still, there are barriers to AI adoption, the paper notes. Some communities face inflexible procurement processes. And a national standard for AI in emergency dispatch has yet to be established.
An evolving field
Ryan, the center’s executive director, began her career as a single-seat dispatcher at a time when warrants were printed out on paper and she used a manual typewriter.
“Things were so different back then. … Nobody, nobody called. You had to be dead, literally, before somebody would call 911,” Ryan said.
When she started, she worked as dispatcher in Charlton where she grew up and residents knew her family. Some would begin with small talk before getting to the reason for their emergency call.
A lone dispatcher can get inundated with calls, she said. Passersby could be calling about a structure fire, a person could be in the lobby and prisoners could be yelling for help.
“God forbid another call comes in, you know, ‘my baby’s not breathing’ at the same time,” she said. “It’s happened.”
A push for regionalization of emergency dispatch began in 2012 when the state mandated that emergency dispatchers stay on the line during medical emergencies, talking callers through, say, how to administer CPR
The technology won’t replace dispatchers at Westcomm, Ryan said. Rather, it will help the center triage calls.
“They’re inundated with all these business line calls,” Ryan said. “So being able to take some of that burden off them just a little bit will be so beneficial.”
Massoia said there were times when he was working, helping someone who phoned a business line, and he had to abruptly end the call because “all of a sudden the whole 911 screen fills up with 911 calls.”
Still, most calls to the nonemergency line will be handled by humans, he said.
For instance, if an older adult fell slowly to the ground and only needs some firefighters to get them back on their feet, a live dispatcher will help. That will be true as well for calls from medical alarm companies reporting that one of their clients pressed a medical pendant.
Questions about the technology
Some have questions and concerns about the technology.
The American Civil Liberties Union of Massachusetts worries how the use of the technology may affect people charged with crimes. It would like to see prosecutors disclose if they use 911 transcripts created with AI.
“911 calls often serve as evidence in criminal cases,” said Katy Naples-Mitchell, the ACLU of Massachusetts’ justice policy counsel, “and the use of AI to create transcripts or summaries could introduce errors into the record that may impact defendants.”
Rebecca Ocasio, 911 deputy director for Springfield, said dispatchers there receive almost 131,000 calls to their emergency lines a year. Include calls to the business lines and that means the city fields 400,000 calls a year to the city’s police department and fire stations.
While the city has been approached by companies offering AI tools, there are no plans to integrate it into operations. Not yet.
“I think that AI has a lot of potential in the future once we understand a little bit more about it,” Ocasio said.
The high cost of the technology is also a factor, she said.
She said the city would want to know that an AI answering service would quickly reroute a call if a caller reached out on the business line but the situation turned into an emergency.
“I think we’d want to be really confident that the system would appropriately route that to an emergency call handler and … nothing would ever be missed,” Ocasio said.
She thinks that language translation will soon be more widely adopted by emergency dispatchers. When a 911 call needs a translator, emergency dispatchers in Springfield conference in a translator. But it can take several minutes to figure out what language a person is speaking and bring in a person who can speak that language.
Shaping guardrails
Over the past few months, officials at Westcomm have put the system through its paces, creating and tweaking when the AI system will pass a caller off to a dispatcher. Humans will handle medical calls.
The system is trained to spot when calls escalate.
In one test, Massoia called the system describing a hypothetical noise complaint but then described an escalating a series of events – they’re fighting! – that pulled in a dispatcher.
The system will have a “scram button” that would divert the call to a human dispatcher, he said. And supervisors can watch in real time the interactions that callers have with the AI system.
As for the data the system collects, Massoia said Westcomm and the communities it serves own it. The AI company it is using – Prepared — is part of Axon, a large company that also makes items like body cameras for police officers.
Aliyah Hernandez, shift supervisor at Westcomm, said the AI tools serve as a level of security to help dispatchers catch details of the call. It helps dispatchers confirm details, such as the exact street number – an easy thing to mix up, she said.
The core function of the job remains, however.
Many people who dial 911 want to stay on the line, Hernandez said, finding that it makes them more comfortable.
“Being there for callers, I would say, is the biggest thing for us,” Hernandez said. “At the end of the day, we’re still the ones picking up the phone. We’re still the ones talking to them.”
