IE 11 Not Supported

For optimal browsing, we recommend Chrome, Firefox or Safari browsers.

Can Artificial Intelligence Help With 911 Staff Shortages?

As government call centers grapple with the nationwide staffing shortage and an influx in demand, some are implementing artificial intelligence tools to improve wait times and accessibility for callers.

Large cities in California, Oregon and Texas are looking to artificial intelligence to combat call center staffing shortages and a backlog of resident calls.

The ongoing U.S. staffing crisis has impacted the workforce across many organizations in what has been dubbed the “Great Resignation.” No sector seems immune; 911 centers have been facing severe shortages that are causing both mistakes and delays in some areas.

Some call centers, however, see AI as a way to bridge the gap between callers and the services they seek.


For the Austin Police Department (APD), the staffing shortage started in late 2019, but the issue worsened during the COVID-19 pandemic, according to APD Lt. Kenneth Murphy.

The high stress of the job, coupled with pay that doesn’t necessarily match daily demands, has made recruiting more difficult. To further complicate things, 911 dispatchers throughout Texas are classified as first responders, so they must meet many of the hiring requirements of a police officer.

Murphy said the city receives over 1 million 911 calls and dispatches about 350,000 police services each year. With 50 vacancies currently, he said APD is sitting at only 70 percent of operational staff, requiring the city to change its response protocol to focus on emergencies.

While non-emergency calls — or 311 calls — are managed by Austin Energy, those still sometimes require action from the police department and thus still populate APD’s queue. As of Feb. 10, 2022, there were over 13,000 customer service requests waiting in the queue for callback. This series of delays, he said, is doing a disservice to the community members — those calling emergency and non-emergency lines.

Here’s how AI could help APD: Austin community members can make requests through an app, a website or voice calls. The AI component would come into play for the voice requests, streamlining the process with a “virtual officer” to record some of the basic information a caller provides. For example, if a vehicle was stolen, the AI can ask the caller about the make, model and license plate number of the vehicle.

This type of call would still require a quality assurance check but would otherwise reduce the time staff spend recording primary details about a report.

In addition, AI could allow for automatic transcription and translation of languages other than English. With a large population of Vietnamese, Chinese and Hindi speakers in the city, Murphy said equal language access is a key reason for APD’s consideration of AI.

APD’s final decision hasn’t been made yet due to some logistics of implementation, as setup on the back end to integrate the questions the AI could ask and when it would do so is an involved process. Murphy said if the city adopts the platform being considered, the product could be in place by summer 2022.


The Portland Bureau of Emergency Communications (BOEC) plans to install Versaterm’s Case Service by July 1, with beta testing running through the end of this calendar year, said 911 Director Bob Cozzie. After testing, the city will determine whether to continue using the product.

Call takers in Portland are currently tasked with answering emergency and non-emergency calls, but Cozzie’s vision is to make Case Service the first touch point for non-emergency calls. As the city establishes its 311 program, his hope is that 311 will use the technology for all non-emergency calls, and the AI could directly answer questions or transfer callers as needed.

In addition, the AI is expected to be able to process information in languages other than English and translate the data within the police records management system. Cozzie said this is one of the most exciting features, but it has not yet been tested by BOEC because the city is still in the scoping phase of implementation.

Although BOEC isn’t immune to the staffing shortages facing government agencies — with only 111 of 131 budgeted dispatch positions filled at this time — it was an influx of 911 calls that most heavily impacted BOEC's decision to use AI.

Cozzie cited city data that found a 43 percent increase in 911 calls for a seven-day period this year compared to the number of calls received during the same seven-day period in 2019.

He noted that public response to the AI tool has been positive, as there have been repeated complaints about the time it takes BOEC to answer the phone.

Technology is part of the solution, but that alone is not sustainable, he said, underlining the importance of increased staffing for public safety answering points (PSAPs), both for the bureau and the industry.

“We’re implementing the technology, but what we’re recognizing — and I think a lot of PSAPs nationwide are recognizing — is there is a new normal,” Cozzie said.

Cozzie said their CAD vendor offered the Case Service product for a beta test at no cost, and it will be absorbed within the bureau’s budget starting halfway through next fiscal year.


The city of San Jose is farther along with its use of AI to address call center challenges, with the city’s multiyear effort to improve the 311 experience well underway.

Over the last year, the city has worked to reduce 911 call times by routing non-emergency calls to the Customer Contact Center. However, this adjustment increased the workload for the center.

By adding Google’s Dialogflow AI-based virtual agent to sort incoming calls and route them to the proper channel, the city was able to free up human call takers for more complex requests.

AI now acts as a touch point for high-volume calls and can answer simple questions. For example, if a resident wants to connect to the trash pickup channel or enter a request for junk removal, they can do so without having to talk to someone, freeing up call takers for requests that require more back-and-forth communication.

The key to successful implementation was having a road map of the city’s goals and the support of city leadership, explained German Sedano, project manager for IT products.

A major need of the city was language translation, according to Arti Tangri, a data architect with the city’s IT department who has been working on the 311 project since the beginning of the pandemic. Now, real-time translation is available for requests in two languages, with the potential to add other languages in the future.

Between 16 and 17 percent of the city’s population have limited English proficiency, so public response to the language translation piece has been positive, Sedano said.

Tangri said the process of training the model involved input of over 1,000 translation pairs. The team used localized, customer-specific terms to improve accuracy.

In regard to spreading public awareness about the AI tool, Sedano noted the significance of a marketing campaign that was conducted through channels like community outreach, mailed flyers, radio and more.

“We’re never going to lose our human touch,” said Kia O’Hara, program manager for the SJ 311 Customer Contact Center. “Even though artificial intelligence is great, we do note that it does have its issues, and so these [tools] will never replace our live agents.”

As Sedano noted, meeting the increased demand on call centers is a journey. As the capabilities of available technology evolve, the city will evolve with them, Tangri said.

Editor's Note: An earlier version of this article made reference to the Portland Bureau of Emergency Management (PBEM) rather than the Portland Bureau of Emergency Communications (BOEC). This error has been corrected.
Julia Edinger is a staff writer for Government Technology. She has a bachelor's degree in English from the University of Toledo and has since worked in publishing and media. She's currently located in Southern California.