When Phoenix, Ariz., residents call the police department’s Crime Stop nonemergency line, they’re now greeted with a new kind of voice — one powered by AI. The conversational system is being used to ease long hold times, sort routine requests and keep emergency dispatchers free for urgent calls.

The technology behind the initiative is Versaterm’s CallTriage tool, a public safety platform customized in partnership with the city to fit Phoenix’s specific needs. Beyond simply answering calls, the system can respond in 36 different languages, including Arabic, Vietnamese, Spanish, Punjabi and Greek. Once connected, callers are guided through a series of questions before being directed to the right resource — whether that’s the department’s Online Reporting portal, Silent Witness, the city Office of Homeless Solutions; or, when necessary, a live communications operator.

Tailoring the tool for Phoenix meant making sure it aligned with existing processes, Allie Edwards, the department’s Communications Bureau administrator, said: “We wanted to integrate our policies and procedures to tailor the system to Phoenix PD’s needs.”

That kind of customization didn’t happen without support. Rolling out the AI system required both financial backing and political will. The $643,000 project was funded through the police communications budget, but Edwards emphasized that one city leader played a key role in pushing it forward. She credited Vice Mayor Ann O’Brien for being a “big advocate of the program” and helping secure the money needed to purchase the product.

Along with funding, accessibility was a central part of the pitch for introducing AI into Phoenix’s call system. The department wanted residents to know that while the experience may sound different, the technology was built to be intuitive and responsive.

“It is conversational AI and will adapt based on the answers,” Edwards said, describing how the technology adjusts in real time. And even with the new system, she emphasized that privacy protections remain intact: “All call records are subject to a public records request, but redactions occur in the same fashion as all other documents.”

Before the system went public, Edwards said it underwent both controlled trials and limited public launches, allowing staff to observe how it performed in real-world conditions.

“The results were positive both times we turned it on — the call takers saw an improvement with calls holding,” she said.

The system has been live since mid-August, but officials are treating it as ongoing, rather than a finished project. Weekly reviews are underway, and a 30-, 60-, 90-day evaluation plan is in place to measure everything from call volume to caller satisfaction. Edwards said residents will also have opportunities to weigh in directly — and officials will monitor wait times and call volumes through satisfaction surveys.

For Edwards, the larger goals behind the project come down to efficiency and making sure emergency services remain available when they’re needed most.

“While the AI currently supports nonemergency calls, it helps by sorting and directing calls before they reach a live operator,” she said. “This means callers are connected to the right department faster, and emergency lines stay open for those who truly need urgent help. It’s all about getting the right help, right away.”

Ashley Silver

Ashley Silver is a staff writer for Government Technology. She holds an undergraduate degree in journalism from the University of Montevallo and a graduate degree in public relations from Kent State University. Silver is also a published author with a wide range of experience in editing, communications and public relations.

See More Stories by Ashley Silver