AI Is Rewriting Police Reports. What Could Go Wrong?

The International Association of Chiefs of Police tech conference in Fort Worth wasn’t open to the press, but the sales pitches happening inside are reshaping American law enforcement. Startups are pushing a familiar narrative: let algorithms handle the busywork so officers can focus on ‘meaningful’ tasks. The problem is, that busywork—filling out reports, reviewing case history—isn’t clerical overhead. It’s the guts of due process. Automating it with tools like ForceMetrics’ Velocity platform or gunshot detection systems isn’t a productivity hack; it’s a fundamental shift in who, or what, is building a case.

Captain Abrem Ayana of Brookhaven, Georgia, isn’t sold. ‘A lot of it is sales gimmicks that don’t actually deliver on what the promise is,’ he said. He’s right to be skeptical. Without federal standards, departments are taking vendors at their word that black-box algorithms work as advertised. This is a dangerous rerun of past failures. CompStat and PredPol were supposed to remove human bias with objective data. They amplified it instead. The new pitch claims that ‘real-time data’ fixes the mistakes of those early experiments, but it’s the same logic wrapped in a faster processor.

The industry jargon is a tell. Jason Truppi, a former FBI agent who co-founded ForceMetrics, talks about ‘antiquated systems of record’ and ‘decision-assist platforms.’ His company is selling what’s known as a real-time crime center, or RTCC—a centralized digital brain processing everything from body camera footage to parole records. It sounds efficient. It also creates a single point of algorithmic failure where a patrol officer’s judgment is replaced by a software recommendation. ‘We don’t use the ‘p word’ at all,’ Truppi said, referring to predictive policing, ‘because it failed.’ A rebrand doesn’t change the core risk: handing the most sensitive decisions to systems that erode transparency precisely when public trust in police is already shot.