MANCHESTER, NH – The Manchester Police Department (MPD) is experimenting with artificial intelligence (AI) as a way to save officers’ time and improve content of reports they write each day.
Lt. Matthew Barter, at the monthly Community Advisory Board meeting at the police station on Wednesday, gave a talk and slide presentation on artificial intelligence available to police departments nationwide and what is currently in use by Manchester police.
Barter explained that people use artificial intelligence in their daily lives when they use their cell phones and laptops. Simple examples of that are autocorrect and forms being automatically filled with the users’ information.
Now, however, there is generative AI which Barter described as “auto complete but on steroids. Think of the whole internet being turned on.”
Currently, AI used at Manchester’s police department includes ShotSpotter, a gunshot detection system that can, in real time, alert police to gunfire; body worn cameras; drones; TAC-ID, which is used to scan drugs and identify them on scene; fingerprinting; and Police Narrative, to help an officer write reports.
MPD also was involved in a pilot program for the software Axon Draft One, the first department in the country to use it. The software used audio recorded on an officer’s body worn camera to complete a draft report of an incident.
The audio is uploaded to “a secure government cloud” according to Barter and auto-transcribed with the information generating a draft report.
A concern with that software, Barter said, is “Who is writing the report? The officer or AI,” among other issues including data leakage.
A total of 87 officers were in the six-week pilot program. Half had access to the software, the others did not. At the end of the program, the officers felt the software was helpful and that it was quicker than if they wrote it on their own. [Link to 2024 study and 2025 follow-up study.]
“We found no difference,” Barter said of the time it took the officers to write up their reports or obtaining it from Draft One. “It was not what we expected at all.”
The software engineers, he said, were eager to work with the department to find out why that was. “The sales people, not so much,” he said.
Officers are now trying PoliceNarratives.AI which incorporates a checklist that officers must follow in investigating an incident and writing their reports.
Barter said this software looks promising because the officer has more control over the content. With it, after each incident, the officer records the necessary details of the incident – and makes sure everything needed from the checklist is included — and then uploads the information to the cloud.
A draft report is issued which the officer then edits, adding or deleting information.
With that software, Barter said, the officer has more control of their report and is able to add context.
“We like this better with the use of the checklist,” he said. He said a few of the more experienced officers said they didn’t need the checklist but just Wednesday morning, Barter said he rejected three reports.
He said it is not that the officers don’t know what they need to do – they do – but sometimes they are distracted. That could be that a witness gave them information that the officer had to immediately follow up, distracting them from some other detail on the checklist.
The pilot program was funded with a grant from the funded with a grant from the Federal Department of Justice, Bureau of Justice Assistance, SMART Policing Program.
Barter said the software is not inexpensive, costing $30 per month per officer. He doesn’t envision that, however. He said for now it would be assigned to detectives or officers who routinely write more detailed reports, up to about 175 officers.