Blue Voice, a Boston company that puts a police department's own rules in front of officers while they are working, has raised $6 million three years after launching in stealth. SignalFire and Las Olas VC led the round. The product is in 225 county departments across 25 states and currently answers about one question a minute. Its founder, Lawrence, left Harvard Law School to build it, after a shooting on the university's campus involving a police officer set off a sharp argument about how officers act and he concluded that many of their mistakes come from not having instant access to their own department's rules.
He built the company with two co-founders: Amit Patankar, a Harvard MBA and a former Google engineer, and Michael Gropman, a retired deputy chief of the Boston police.
The problem Lawrence describes to TechCrunch is a memory problem. An officer making a decision is working from recall and personal judgement against thousands of pages of statute, internal policy, municipal ordinance and state law. If the exact seventh step at a crime scene will not come to mind, the options were a 15,000-page manual, waking a supervisor in the middle of the night, or asking Google or ChatGPT, neither of which is trained on police-specific rules. Traditional search took too long, and consumer AI models, Lawrence says, got these answers wrong as often as 30% of the time.
Blue Voice indexes the material general-purpose tools cannot reach: a given department's governing law, local ordinances, protocols and standing instructions, none of it published on the open internet. During emergencies, and Lawrence's example is an attack on a school, officers can also pull up detailed building maps on their phones.
The design decision that matters is that the system does not answer so much as cite. Officers trust it, Lawrence says, because every response leads back to the source regulatory document rather than generating text the way ChatGPT does. It shows the applicable law and stops there. The decision stays with the officer, who weighs it against his own experience.
Customer count is up 11x over the past year. Lawrence attributes that to the results departments report back: crime down, and fewer disputes over what officers did.
He offers two cases. In the first, he says the product recently helped prevent a kidnapping. A young officer saw a man trying to talk a girl into his car and was not sure he had grounds to step in. He opened Blue Voice on his phone and it confirmed on the spot that the situation met the legal criteria for child enticement, an attempt to lure a minor in order to commit a crime against them, which gave him immediate legal basis to act.
In the second, Blue Voice reminded a department chief that he could not return an officer to active duty after a shooting until the officer had passed an independent mental health evaluation.
The competitor is Lexipol, which is backed by private equity. Blue Voice has also built features for detectives working unsolved cases. All of this lands while police use of AI draws heavier criticism, with Flock Safety's licence plate surveillance systems as the standing example, and Lawrence's stated ambition is to show that AI deployed properly can serve public safety without taking civil rights with it. Before the company he worked in the administration of Connecticut Governor Ned Lamont, and he says he does not regret leaving law school, calling the startup the most meaningful use of what he can do.
The numbers are smaller than they sound. One question a minute is roughly 1,440 a day across the entire customer base, which works out to about six questions per department per day. That is real usage, and it is not yet the rhythm of a shift. Blue Voice is being consulted at the margin, in the cases an officer already senses are unusual. The $6 million round is sized to match: this is money for a company that has proved distribution across half the states and has not yet proved depth inside any of them.
The 30% figure deserves the same scepticism. It comes from the founder, it is about his competitors, and no methodology travels with it. What is missing is the equivalent number for Blue Voice. Nothing in the account states how often it surfaces the wrong regulation, and nothing addresses what happens when a department's own manual is wrong or out of date, which is the one failure mode the architecture cannot design around. Its ground truth is the customer's paperwork.
The claim that the platform never tells anyone what to do is also carrying more weight than it can hold. In the enticement case the app confirmed that the conduct met the criteria for a crime, and the officer moved. That is the tool making the call, whatever the interface says about the decision remaining human. Citation-first is genuinely why officers trust it, and it is also a liability posture: when the product only shows the law, the person who acted on it owns the outcome.
The more interesting of Lawrence's two stories is the second one. A system that tells a chief he cannot put an officer back on the street is the version of police AI that civil-liberties critics should want, and it is the version that is hardest to sell, because departments buy tools that help officers act. Blue Voice's value to everyone who is not a police officer sits almost entirely in the moments it tells one to stop.