Estimated read time: 5-6 minutes
WEST JORDAN — Ramon Ramirez is one of many police officers patrolling the streets of West Jordan.
"Eventually, I hope to be a detective in one of our specialties," he said.
For now, his favorite part of the job is interacting with the community.
"It's just being out and about having that human-to-human contact and telling someone in need that I'm going to help them the best way that I can," Ramirez said.
But no job is perfect. And paperwork is definitely among the things police "don't love," he admitted with a smile.
Now, Ramirez and a selection of other employees at the West Jordan Police Department are getting some help with that paperwork from artificial intelligence.
Agencies turning to AI
Through the AI Revolution series, the KSL Investigators have explored the impacts of artificial intelligence across several sectors.
When it comes to law enforcement, we've found there are several ways the emerging technology is changing and potentially improving public safety at a time when departments across the country are struggling to recruit and retain officers.
The KSL Investigators surveyed Utah's 25 largest police agencies, finding South Jordan, West Valley and Provo Police Departments are also piloting the Axon Draft One report-writing software. Five other agencies said they're considering using AI tools for report writing.
More than half a dozen departments told us they already use AI-powered license-plate recognition software. AI can also be used to identify weapons in public spaces, and for facial recognition.
"There's a lot of jobs for AI to do in policing," said Anthony Tassone, CEO of the tech company Truleo.
While reviewing 100% of all body camera footage captured by a department wasn't feasible in the past, Tassone said with AI, every frame can now be efficiently reviewed and translated into data, giving officers credit for positive interactions.
"The main problem that Truleo solves is we improve police morale, we improve recruiting, we improve retention. How do we do that? Because we give them credit for doing a good job," Tassone said. "The problem is that none of the videos are being reviewed unless there's a horrific use of force or civilian complaint."
But he believes, with AI, that will change.
"One hundred percent body camera review will become the standard," he said. "It will get mandated; it will become the law."
He said AI can also put a "virtual" training officer in the field, to give immediate feedback.
"This partner will talk to them based on the body camera videos and say, 'Hey, I think in the future you should give more explanation,' or 'Hey, you forgot to introduce yourself in that last call,'" Tassone explained.
He said departments across the country use Truleo's technology, though none in Utah yet.
Draft One demo
The KSL Investigators asked West Jordan police to show us how Axon's Draft One — the report writing software it is testing — works in practice.
Sgt. Andrew Hercules posed as a civilian reporting a crime to Ramirez. The software then used the audio from Ramirez' body camera to create a first draft of a police report almost instantaneously.
In a matter of minutes, Ramirez had a draft report on his laptop. But Draft One gives him just that: a draft. He's still required to edit and add his own observations, then submit it for another review by a supervisor.
Ramirez said that over time the program has gotten better at mimicking his personal writing style. But he's not worried about AI taking his job one day.
"Not even close," he said with a chuckle.
Built-in safeguards
"An AI-generated report can never be a substitute for police work," said Salt Lake County District Attorney Sim Gill.
By the time an AI-generated police report makes it to his desk, Gill said he knows it's been reviewed and edited by a human. Additionally, he said police reports are not evidence. They're based on actual evidence that can be double-checked, like body camera footage.
For the sake of transparency, Gill said the reports also include a disclaimer, letting both prosecutors and defense attorneys know the AI software was used.
"For us as prosecutors, at this initial stage, we're OK with this as coming in because of those different safeguards that are in there," Gill said.
AI = time savings?
Noah Spitzer-Williams is a product manager for Axon's Draft One. He said he can understand why the public might be skeptical of police using artificial intelligence.
"You've got to make sure that at the end of the day, the human is always in control," he said.
But he said their goal is to save officers time, while also creating higher quality reports. And he said they've built in their own safeguards to avoid errors and AI-generated embellishment.
"Even though we use a lot of the same underlying technology that Chat GPT uses … one of the most effective things we've done is we've effectively turned off the creativity in that software model," Spitzer-Williams explained. "We really instructed it to just stick to the facts of what's been captured on camera."
He said the program is leading to significant time savings for officers.
"They're telling us that we're basically cutting the report writing time in half," he said.
But a recent study of Axon's Draft One by researchers at the University of South Carolina found little time savings.
"The study hasn't been peer-reviewed yet," Spitzer-Williams said. "It also doesn't take into account the actual anecdotal feedback from officers."
In search of more anecdotal feedback, the KSL Investigators asked Ramirez how much time the software is saving him.
"A lot of time," he said. "Like a lot of time."
He estimated Draft One saves him 20 to 30 minutes on each report. That could translate to a couple of hours per officer on a typical shift — precious time they can use to get back to the parts of policing they love the most.
"This tool so far has seemed to be very effective at making report writing time more efficient," Hercules said. "So, our officers are able to get back out with the community."
Hercules said the department is participating in a 90-day trial of the software and is still evaluating whether to adopt the technology going forward.
