The Charleston County Consolidated Emergency Communications Center in South Carolina is using Amazon Connect, an omnichannel cloud contact center by Amazon Web Services, to answer non-emergency calls (Photo courtesy of Jim Lake).

In the middle of a storm, 911 call centers often find themselves inundated with reports of fallen trees, flooded roads and panicked residents. Every call matters, but with multiple reports of the same incident pouring in, the pressure on emergency services can become overwhelming.

Amid the chaos, a technological ally has emerged: artificial intelligence. In the United States, AI is quietly transforming how non-emergency calls are handled in dispatch centers. An AI-powered system can triage and coordinate the flood of reports, promptly alerting the relevant agencies.

For now, AI-powered systems only manage non-emergency calls, which typically come from a non-911 phone number but are answered in the same centers, allowing human dispatchers to focus on emergencies.

The integration of AI technology into 911 centers is partly a response to an acute staffing crisis and the pressing need to address the mental health challenges that emergency responders face. While AI-powered systems in 911 centers offer potential benefits, such as managing call surges and reducing dispatcher workloads, concerns linger among experts about the possibility that these systems may overprescribe police response or make mistakes due to biases.

So far, fewer than a dozen localities in seven states across the country are using or testing artificial intelligence in their 911 centers. But, as in other industries, leaders are wondering how AI can transform workplaces.

“For me, I think that the use of AI for non-emergency calls is a fantastic idea,” said Ty Wooten, the director of government affairs for the International Academies of Emergency Dispatch, an organization that helps set standards for emergency dispatch centers. “I see the huge benefit of being able to alleviate those calls out of the 911 center queue so that the 911 call takers can really focus … on the ones that really matter.”

Emergency call centers are struggling to find workers. Between 2019 and 2022, 1 in 4 jobs at 911 centers were vacant, according to a report published in June of this year from the International Academies of Emergency Dispatch and the National Association of State 911 Administrators. As emergency call centers continue to grapple with understaffing issues, some 911 calls may go unanswered or get stuck in lengthy queues.

“That subsequent loss of staff makes everyone have to work more, which then burns people out and creates more turnover,” Wooten said in an interview. “It’s this vicious cycle.”

For now, there’s little regulation on how artificial intelligence can help. Only a few states have set AI regulatory frameworks. And the definition of AI itself remains uncertain in many states.

Public safety agencies often approach new technologies, including artificial intelligence, with caution because of concerns about service disruptions, said Brandon Abley, the director of technology for the National Emergency Number Association, a nonprofit professional group.

“[Emergency call centers] are not really stumbling over themselves to try and implement AI in their operations because generally, they don’t want huge disruptions to their operations unless they’re very, very certain,” Abley said in an interview with Stateline.

And there could be disadvantages, he added. For example, dispatchers could face heightened mental health challenges if they have to manage more emergency calls because an AI system is taking the bulk of administrative or non-emergency calls.

“We think it looks promising,” Abley said, “but we’re also cautious.”

Boosting efficiency and reducing workload

The testing or implementation of AI systems for call-taking in 911 centers already has begun in municipalities in Colorado, Maryland, Missouri, Oregon, South Carolina, Texas and Virginia.

Among the driving forces is the dual role that call center personnel play. In most public safety centers, the same people answer both emergency and non-emergency calls. With a shrinking workforce, some governments see AI as a solution to alleviate part of the workload.

Among the tech companies offering products to 911 centers is Amazon Web Services, a subsidiary of Amazon that provides cloud computing services, including Amazon Connect, a cloud-based contact center designed to provide verbal assistance. Carbyne is another software company focused on emergency communications services that uses AI for live two-way translation and triaging calls.

In South Carolina, for example, Amazon Connect is used for non-emergency calls in Charleston County’s Consolidated Emergency Communication Center. When a caller dials the county’s non-emergency line, Amazon Connect will answer and ask the caller what they need help with. The system will redirect the caller to appropriate resources, allowing human dispatchers to focus on emergency call-taking. If the system cannot understand the caller, it will send the call to a human dispatcher.

The center spends about $2,800 per month on its Amazon Connect subscription, which Jim Lake, the center’s director, said is cheaper than hiring staff solely for answering non-emergency calls. The system has reduced the volume of calls to the administrative line by 36% since March, Lake told Stateline.

“Those are calls that our 911 public safety telecommunicators don’t want to take. They are not emergencies. So we’re showing them that we’re making their jobs more efficient and giving them the opportunity to do more on those emergency calls,” Lake said.

Several other call centers — including the Arlington County Emergency Communications Center in Virginia, the St. Louis County Police Department in Missouri and the Jefferson County Communications Center Authority in Colorado — also are adopting the Amazon Connect system or similar technologies.

Since Jefferson County began using Amazon Connect’s program last December, AI has processed about 40% of the emergency center’s administrative calls.

“We’re processing just under a million calls a year, so for us to handle it through technology — freeing up personnel to handle more acuity-style calls — works much better for us,” said Jeff Streeter, the center’s executive director.

While there are concerns about AI displacing dispatchers’ jobs, many leaders of 911 call centers emphasize that their goal is to make existing roles more manageable.

“I cannot stress enough that it does not take away jobs, especially in the 911 industry. It’s there to help them enhance their job,” said Jacob Saur, the emergency communications center administrator for Arlington County Public Safety Communications and Emergency Management. “I just cannot see in any way, shape or form an automated bot answering a 911 call.”

Brian Battles, the communications administrative specialist for the St. Louis County Police Department’s Bureau of Communications, which oversees the county’s 911 operations, echoed this perspective.

“It has been very beneficial to the call takers, who are already overworked,” Battles said. “Anything we can do to relieve that stress while actually providing a more efficient service to the citizens is a no-brainer on our part.”

Addressing bias and funding

Like other new criminal justice technology, concerns about bias loom large with AI systems.

“All AI models are only as good as their developers,” Daniela Gilbert, the director of the Vera Institute of Justice’s Redefining Public Safety initiative, wrote in an email. The potential is there, she wrote, for AI to replicate human biases on a large scale.

“If these systems are [designed] to take calls, rather than assisting call takers, it would remove a human empathy that is so often essential in crisis situations,” Gilbert wrote. “Imagine being in a time of stress and great need and having to negotiate with a bot.”

If, for example, developers have a particular bias that favors police response, AI systems may overprescribe police involvement when alternative resources might be more suitable, Gilbert wrote.

Martha Buyer, a telecommunications law attorney and 911 expert, emphasized that AI systems are prone to errors, which could lead to liability issues. The systems must be capable of accommodating a diverse range of callers, including those who speak languages other than English or have specific needs related to their abilities, Buyer added.

“To have an AI system answer a 911 call — that’s so fraught with liability I don’t even want to think about it,” she said. “Timing is critical.”

Artificial intelligence systems aren’t available everywhere in part because many dispatch call centers find themselves stuck in a technological time warp, relying on old systems that struggle to keep pace with rapid tech advancements.

“The reality is the system of 911 as it is today across the country is still kind of operated off technology that was developed in the 1930s,” said Wooten, of the International Academies of Emergency Dispatch. “That technology has to be upgraded, and we have to get that to a point where we understand and it is more equitable.”

Even as cellphones have become ubiquitous, for example, some outdated systems grapple to accurately pinpoint a mobile phone caller’s location. Instead of obtaining precise GPS coordinates, these centers might only get the location of the nearby cell tower, hampering response efforts.

“Nobody ever plans on needing to call 911, so from a government perspective, it’s often pushed to the side in terms of funding,” Buyer said.

Wooten said that despite AI’s potential, many centers need basic tech improvements before getting involved with artificial intelligence.

“We really have to get the infrastructure in place and taken care of first before we will ever be able to see the benefits and understanding of what other future technologies, whether that be AI or any other future technology.”

Stateline is part of States Newsroom, a nonprofit news network supported by grants and a coalition of donors as a 501c(3) public charity. Stateline maintains editorial independence. Contact Editor Scott S. Greenberger for questions: info@stateline.org. Follow Stateline on Facebook and Twitter.