Here is a fact about American policing that does not appear in any recruiting brochure: a large share of the job is typing. An officer responds to a call, does the thing officers do, and then goes back to a desk - or a laptop bolted to a cruiser - and turns the whole event into a narrative report. By most accounts this eats somewhere between two and four hours a day. That is not a rounding error. That is, on the higher end, half a shift spent describing work rather than doing it.
KARDA, Inc. looked at that number and did the thing that founders sometimes do, which is to refuse to accept that it was fixed. The company, founded in San Francisco in 2023 by Armando Torres, makes software that reads an officer's body-worn camera footage - the video and the audio - and produces a drafted police report. The officer reviews it, edits it, signs it. The company's tagline borrows from police radio code: "10-8 in 10 minutes." (10-8 means back in service, available for duty.) The claim is that a report which used to swallow an hour can be back in front of you, in draft, in roughly ten minutes.
Now, whenever someone tells you an AI will write the document for you, the correct first question is: and then what? Because a police report is not a blog post. It can become evidence. It gets read by prosecutors, defense attorneys, and occasionally juries. A hallucinated detail is not a quirky failure mode; it is a problem with consequences. KARDA's answer is that the AI drafts and the human decides - the system extracts facts, identifies key details, and folds in agency policy and context, but the officer remains the author of record. The pitch is not "trust the robot." The pitch is "stop transcribing, start reviewing."
"No officer spends more than one hour on documentation per day instead of four - dramatically increasing patrolling by three hours per officer, per day."
KARDA, stated mission
There is a second, quieter argument buried in that mission statement, and it is the one that makes the business interesting. The usual answer to "how do we improve public safety" is "more officers." More officers is expensive, slow, and politically fraught. KARDA is selling a different lever: take the officers you already have and hand back the hours they lose to forms. Three hours a day, multiplied across a department, is not a software feature. It is patrol capacity that appeared out of nowhere, without a single new hire.
The company leans on a research-flavored claim to close the loop - that positive community engagement can reduce crime as much as, or more than, simply putting more officers on patrol. Take that as a directional argument rather than a settled fact; the point KARDA is making is that time not spent typing is time available for the parts of policing that build trust. Whether or not the causality holds, the arithmetic of the workday is real, and it is the arithmetic that departments are buying.
Selling anything to law enforcement, though, means clearing a bar that has nothing to do with how clever your model is. It means CJIS - the FBI's Criminal Justice Information Services security policy - and it means encryption, and it means being the kind of vendor that a records custodian can defend in a deposition. KARDA describes itself as FBI CJIS compliant with end-to-end encryption. In this market, those are not nice-to-haves bolted on at the end. They are the price of getting through the front door, and the company appears to have understood that early.
The founder is a useful tell about the company's temperament. Armando Torres studied chemical engineering at Stanford - not law enforcement, not even, strictly, computer science - and arrived at the problem as an outsider. He is a Gates Scholar and a Milken Scholar, credentials that signal the kind of person institutions bet on early. Outsiders have a specific advantage in a field like this: they never learned that four hours of paperwork was normal, so they never stopped finding it strange.
KARDA is small - roughly fourteen people - and early. It raised a $200,000 seed round, reported with the startup generator Antler as a backer, and updated its branding in late 2024. It has reported early engagement with departments including the Boston Police Department, and it sponsors law-enforcement associations like the Virginia Association of Chiefs of Police. This is the unglamorous, relationship-heavy, procurement-gated work of selling to government, and there is no shortcut around it.
The company has also built a second door into the same engine. The KARDA Kiosk is a multilingual, self-service station that lets a member of the public file a non-emergency crime report without tying up an officer at all. Same core technology, different audience: the report engine serves the department, the kiosk serves the community. It is the kind of second act that suggests KARDA sees itself as a documentation platform, not a single feature.
None of this is a sure thing. The competitive field is filling in fast - Axon, the body-camera incumbent, has its own AI report-drafting product, and it already owns the cameras and the relationships. KARDA's counter is integration without replacement: plug into the Records Management Systems departments already run, rather than asking them to rip anything out. In a market that moves at the speed of a city procurement cycle, being the vendor that is easy to say yes to may matter more than being the one with the flashiest demo.
Strip away the AI vocabulary and KARDA is doing something old-fashioned and specific. It picked one boring, expensive, universally hated task - the narrative report - and pointed a stopwatch at it. If it works at scale, the product nobody will talk about is the one that quietly gives a beat cop back the last three hours of the shift.
