Editor’s Note: This article is part of the Police1 Innovation Report, which examines how emerging technologies are reshaping law enforcement operations and leadership decision-making. This series explores how AI is transforming emergency communications, including call processing and information sharing, telecommunicator roles and training, and connections with RTCCs, cameras, drones and other response technologies.
Dispatchers are specially trained to answer emergency calls and get the right resources to the right place efficiently and safely.
But in San Diego County, California, as in many jurisdictions, the work is stressful, maintaining full staffing is challenging and an increasing number of non-emergency calls are coming into an already busy 911 call center.
Why non-emergency calls were straining the communications center
To help address staffing challenges and better serve the 1,200 to 1,800 daily calls to its non-emergency line, the San Diego County Sheriff’s Office (SDSO) is establishing a remote communications unit (RCU). Once fully trained, members of the 10-person RCU will answer non-emergency calls from their home offices using SDSO hardware and software rather than working at the communications center.
While equipping the RCU and training the team, Capt. Nathan Rowley, the Communications Center division commander, began researching AI tools that could provide additional assistance with non-emergency calls, such as noise complaints or property theft. The agency ultimately selected Hyper.
“We onboarded Hyper as part of a larger project for our remote communications unit,” Capt. Rowley said. “In doing so, we found a lot of ways that we can integrate [Hyper] in reducing work on our emergency services dispatchers.”
In the video below, learn how the San Diego County Sheriff’s Office uses an AI-powered system to streamline non-emergency calls, reduce dispatcher workload and connect callers with the right resources more quickly.
| RELATED: The future of the telecommunicator: How AI is changing dispatcher training
The AI-powered call-answering service, launched in March 2026, can answer every non-emergency call simultaneously, eliminating hold times. It can respond to callers’ questions and route them to the person, information or resource they are seeking, including transferring callers to the 911 operator queue.
“Because our 911 operators are the same people who answer our non-emergency line, the less traffic we have coming in over that non-emergency line, the more available they are to answer 911 immediately,” Capt. Rowley said. “As well as reduce some of that mental workload in their shift.”
For example, SDSO regularly receives non-emergency calls about complaints within the city of San Diego or services provided by the city. Instead of involving a 911 operator, Hyper can recognize that the call is intended for the city, provide the caller with the correct phone number and even transfer the caller to the city’s communications center.
“We’re able to give them the information, meaning as soon as they call in, AI is telling them this isn’t the correct number to call, but here is the correct number to call, and also offer to text that number to them if they want,” Capt. Rowley said. “In many cases, [Hyper] can connect them to San Diego PD, sheriff’s licensing or the courts — without them having to do anything additional.”
Not every non-emergency call will be resolved by AI, and Hyper’s goal is not to replace a person. The goal is to quickly and accurately connect callers with the person, information or resources they need.
| RELATED: Calif. sheriff’s department to answer non-emergency calls with AI
What Hyper handles and what it does not
Capt. Rowley needed an AI call-taking solution that could:
- Understand the nature and location of the call
- Recognize when a call should be transferred to the emergency line queue
- Connect callers with the appropriate resource more quickly when SDSO is not the correct agency, rather than making them wait in the queue
After offering callers the option to speak in Spanish, Hyper listens to the reason for the call and asks follow-up questions as needed. It can then transfer the call, provide an answer or text information and website links to the non-emergency caller. Callers can also ask to speak with a 911 operator. If Hyper determines that the call involves an emergency, such as a crime in progress, it can transfer the caller to the emergency line.
Although SDSO is still collecting data, Capt. Rowley reported that the wait time for callers transferred from the non-emergency line to the emergency queue has already been cut in half.
More 911 operators are available to answer emergency calls, reducing 911 answer times. For those working in the communications center, it “just seems like a lighter day,” he said.
| RELATED: How AI is helping a Nevada police department reduce non-emergency dispatch calls
Early results for callers and dispatchers
Capt. Rowley explained that Hyper is delivering both expected and unexpected benefits. One of the primary expected benefits is that callers — San Diego County residents and visitors — receive an immediate answer when they call the non-emergency line.
Before Hyper, non-emergency callers remained on hold until a 911 operator was available. To callers’ frequent frustration, they might wait 20 minutes to speak with a 911 dispatcher, only to learn that they needed to call a different number or could find the information on a website.
Now, imagine the impact on a 911 operator who answers a call from someone who has just spent 20 minutes waiting.
“It’s mentally exhausting for our call takers because now they’re dealing with an irate person who called the sheriff’s office to be told they have to dial another number,” Capt. Rowley said.
These negative and stressful calls can take a toll on people already working in a stressful and emotional role. Hyper is reducing the frequency with which frustrated callers direct their frustration toward 911 dispatchers. This unexpected benefit is also easing operators’ workloads.
The improved work environment and reduced workload make Capt. Rowley hopeful that Hyper will positively affect retention. Adding staff, including the remote communications unit, while reducing call volume may ease stress and further improve the work environment.
| RELATED: Chatbots: The next technology revolution in 911 dispatch centers
Six months of testing before launch
SDSO and its partners in the Data Services Division spent six months testing and training Hyper, a process that required substantial time from one full-time staff member. Hyper was given access to tens of thousands of recordings from the non-emergency line to learn common call types, local dialects, geographic terms and regional jargon, helping the system understand the service area and callers’ common needs. During testing, Capt. Rowley’s team tried to confuse Hyper so they could identify and correct inaccuracies before the system went live.
“If we found inaccuracies, we sent it to the company to update their AI, essentially to teach it. When the caller says ‘this,’ this is what they mean,” he said.
After the initial training period, staff fed recordings from recent days into Hyper to identify and correct additional inaccuracies. When corrections became few and far between, Hyper began answering non-emergency calls for five minutes at a time, then 10 minutes and eventually a few hours. This testing took place under the watchful eye of communications center staff, who could intervene or document the need for additional training.
Once Hyper performed well during the 12 busiest hours of the day, the agency enabled the system around the clock. With nearly 2,000 calls coming in each day, SDSO continues to monitor and fine-tune the system as needed. If Hyper is unsure how to assist a caller, it transfers the call to the 911 operator queue. The software can then be updated to better assist callers with similar needs in the future.
While training Hyper, SDSO explained the system to the public and provided examples of what callers could expect. However, Rowley acknowledged that the agency could have done a better job explaining Hyper to deputies and dispatchers. Deputies received questions from the public that they lacked the training and information to answer correctly, while some dispatchers were concerned that they needed the AI system’s permission to send a deputy to an incident.
Capt. Rowley encouraged other leaders to plan transparent education and training for both external and internal audiences — a key lesson learned during SDSO’s Hyper deployment.
AI as a tool, not a dispatcher
Although Hyper is answering up to 20% of non-emergency calls, the department’s goal is not to replace people with AI.
“We are not trying to make Hyper a dispatcher,” Capt. Rowley said. “We are trying to help people get to the right resource as quick as possible.”
He also stressed that Hyper is not used to answer 911 emergency calls. It is used only to better assist non-emergency callers.
“Being able to direct [callers] to the correct resource immediately is not only from a customer service perspective very nice to have, it’s also huge for public safety,” Capt. Rowley said. “Now we’re getting people to the right resources immediately so that if somebody does need to be dispatched, there’s no wait, because [they aren’t] in this queue of people waiting to talk to somebody.”
Capt. Rowley is hopeful that a less stressful work environment will eventually improve retention.
“If we’re able to decrease that stress and make the work environment better, that’s a win for everybody,” he said.
| NEXT: Smart dispatching: How artificial intelligence is reshaping emergency response
