📰 Key Takeaways

Blue Voice is a Boston-based AI startup founded by Harvard Law dropout David Lawrence, along with co-founders Amit Patankar (Harvard MBA, former Google engineer) and Michael Gropman (retired Boston PD deputy superintendent), building real-time policy guidance for frontline police officers. Lawrence got the idea after witnessing an officer-involved shooting on campus and realizing most law enforcement mistakes come down to officers not being able to look up department policy in the moment — so he decided to go build an AI solution for it.

The company recently came out of stealth, announcing a $6 million round led by SignalFire and Las Olas VC. Officers across 225 county-level police departments in 25 states are already using the tool daily to make sure their actions stay within department policy, and Lawrence is positioning it as the Harvey of law enforcement, or the OpenEvidence of policing.

Before this, if an officer forgot a specific procedure — say, step seven of processing a crime scene — their only options were digging through a manual that can run up to 15,000 pages, calling a supervisor in the middle of the night, or turning to general-purpose tools like Google or ChatGPT. But those consumer AI models get things wrong up to 30% of the time in this context. Blue Voice trains on each department’s specific regulations, local ordinances, and operating procedures — content that generic AI tools simply can’t pull from the open web. The platform also gives officers detailed campus maps on their phones for active shooter situations.

Blue Voice currently answers roughly one question per minute, and it’s built by design to not issue direct commands — instead it points officers to the actual regulatory text and lets them combine that with their own judgment on the ground, which is how the tool earns officer trust. Over the past year, the company’s customer count has grown 11x, and departments are reporting concrete results like lower crime rates and fewer incidents involving controversy. In one recent case, a rookie officer used the tool to confirm in real time that a suspicious situation met the legal elements of “enticement of a minor,” helping stop what looked like an attempted kidnapping.

💬 JudyAI Lab Take

What stands out to us about Blue Voice is how simple the underlying problem is: you can’t flip through a 15,000-page manual in the field, calling a supervisor at 2am isn’t realistic, and general-purpose AI tools get it wrong up to 30% of the time in a domain this specialized.

This is a design pattern worth noting for anyone building AI products: instead of chasing a model that “knows everything,” go deep on training against one domain’s specific regulatory documents, and let the AI only answer what it’s actually confident about. Just as important is the interaction design choice — the system doesn’t issue commands, it points officers back to the source regulation and lets a human combine that with real-world judgment. That “assist, don’t replace” positioning is a big part of why Blue Voice’s customer base grew 11x in a year, and it’s the core of how they built user trust.

If you’re building an AI tool, it’s worth asking: is your product’s job to make the decision for the user, or to help them find what they need to make the decision faster?

📅 Source Info

🔗 Further Reading