Phoenix police to use AI to answer non-emergency calls. Is that good?

Starting today, the Phoenix Police Department will begin using artificial intelligence to answer non-emergency calls — joining departments in several other states across the country in adopting the technology for call centers.

In a press release, Vice Mayor Ann O’Brien said she led the effort to “modernize” the city’s call centers. Phoenix also previously adopted a text system for people who call 911.

“When I learned our 911 dispatchers were responsible for handling both emergency and non-emergency calls — and that many residents making non-emergency calls were giving up due to long wait times — I committed to finding a solution,” O’Brien said in a written statement. “After researching technology options, I requested City staff and the Police Department explore voice-activated AI-powered systems to handle non-emergency calls.”

Last year, Phoenix police handled more than 605,000 non-emergency calls — about 1,650 a day. While a Phoenix police department spokesperson has not yet provided Phoenix New Times with data on average non-emergency call wait times, any Phoenix resident who has called the line before knows the wait to speak with an operator is infamously long.

The AI system, called CallTriage, was created by the Canadian firm Versaterm. The system will be used to identify the intent of a non-emergency call after asking callers a few questions before transferring the call to the appropriate city division.

“Whether someone’s reporting a noise complaint, parking problem or another issue that does not require an officer’s immediate response, this new system ensures faster, more targeted help — and it helps our dispatchers stay focused where they’re needed most — handling emergency calls,” O’Brien said.

The city’s non-emergency number is 602-262-6151. O’Brien’s office said the system speaks up to 36 languages.

But with watchdogs flagging potential misuse of AI as it spreads into government, criminal justice and policing, should Phoenix’s program be a cause for concern?

Cristopher Moore, a professor of computer science and physics at the Santa Fe Institute who also studies AI in the criminal justice system, told Phoenix New Times that he isn’t worried about this particular use. More than anything, he’s curious to see how effective it will be.

“As a switchboard, it sounds like an interesting experiment. It’ll be just like when you call an airline or when you call a bank,” Moore told New Times. “And if it connects lots of people to the wrong person, they’ll give the city valuable feedback about whether or not it’s working. It’ll either save people time or frustrate them and waste time.”

Moore noted that while this use of AI is likely benign, sometimes the technology can create problems. For instance, large language models like ChatGPT often make things up, which happened in New York City when chatbots gave citizens incorrect information about city code.

“In general, I think it’s fine to use AI when you can check its work,” Moore said. “Where I get worried is when people become reliant on it as a replacement for humans or when it’s used to make consequential decisions.”

The vice mayor said that she got the idea to adopt AI for this purpose from Portland’s use of AI for non-emergency calls. But what O’Brien didn’t mention in the press release — and what KJZZ first reported — is that Portland discontinued using the service. KJZZ reported that Portland dropped the service due to “hardware issues” that conflicted with its outdated phone lines, an issue that Phoenix officials told the station they do not anticipate encountering.

O’Brien’s office mentioned that “residents will always have the option of speaking to a live dispatcher if they prefer.” Even though he sees the utility in the AI system, Moore might be the type to gravitate towards that option.

“Like a lot of people, the first thing I do when I hit an automated system is I press zero, I press pound, I press star, I say ‘agent’ or ‘representative,’” he said.