We make law firms AI-native.
Hi everyone, we're Peyton and Michael Marcotte, brothers and co-founders of Perceptron ML (YC S26). We make law firms AI-native: we build the AI and software a firm needs to scale. Not a product off the shelf, a system built around how your firm already works, where no fact is used until it traces to a real document.
Perceptron ML (YC S26) Launch Video
The problem. In 2023, two New York lawyers filed a brief written by AI that cited six cases that did not exist. Courts have caught it more than a thousand times since. So most firms went back to doing everything by hand, and by hand the math is brutal: lawyers lose ~40% of their day to work that isn't law, and tens of billions in legal work goes unclaimed every year because firms can't move fast enough to take it.
What we do. Every firm's bottleneck is different, and it is never the one on the shelf. So we build what each firm actually needs: intake answered and filed at 2 a.m., demands drafted from the record, records requests chased until they land, deadlines and statutes watched, every hour matched to its matter, clients updated by text before they think to call. Every firm gets something different, but underneath it's all the same platform we built once: our grounding engine, our connectors into the firm's systems, and our agent runtime. That's what lets us build this broadly without being an agency.
Why you can trust it. The AI we build doesn't guess. Every fact traces to a real document, verified against the primary source before a model can use it. Click anything, and the original opens.
Why us. Michael did AI research at NVIDIA (hardware-debugging agent in production). Peyton led two NASA-supported research projects. We come from places where a confident wrong answer is unacceptable, and we started Perceptron ML because law is the biggest industry still run by hand for exactly that reason.
The ask. If you know a managing partner, or the attorney who signs your invoices, we'd love an intro: founders@perceptronml.com
