Jeffrey Mason, founder of Fast Patent Partner, joins Sunny Ray to discuss how agentic AI is transforming patent litigation. After more than 25 years litigating pharma patents, including a case that settled for 1.4 billion dollars, Mason built tools that use AI to hunt for patent infringers and process the mountains of public data patent cases require. He explains how his company's dashboards let agentic AI deep mine websites, pull product numbers, and build claim charts, tasks that used to require expensive associates and paralegals working manually. Mason describes using adversarial AI agents, including Claude, to check each other's work and catch hallucinations before anything reaches a court. He also discusses Fast PTO Docs, a low cost tool for bulk pulling patents and file histories, and the technical challenges of scraping PTO records at scale, from CAPTCHAs to unstable APIs. Mason argues law firms are paid for inefficiency, so selling speed requires proof, which is why his dashboards surface reasoning so attorneys can trust the output. He closes by weighing what agentic AI means for the future of junior associates.
A patent litigator explains how agentic AI now does associate level work, like hunting infringers and building claim charts, hundreds of hours a day.
You spent 25 plus years litigating pharma patents, including a case that settled for 1.4 billion dollars. What made you start writing code on the side?
My first love was actually code, I was a coder in high school, but this was pre internet, pre AI, so it didn't excite me the way it does now. I moved into biochemistry because that was the hot area for intellectual property, then went to law school. Grad students at Harvard were racing each other by hand to find things, and I knew I wanted to support science without doing that kind of labor myself.
Your tagline is 'your best associate working hundreds of hours a day.' When did you realize the associates' work could actually be automated?
AI at first wasn't up to it, but I worked with a Fable model and had it and Opus working adversarially on a legal brief, checking facts and citations. No mistakes, and no human team does that. At my old firm, Finnegan Henderson, the rule was nothing goes out the door without an attorney reviewing it. Now an AI can check your work just as capably. That was a huge shift for me.
Can you explain Fast PTO Docs to a patent attorney in about 90 seconds? What does it do and what does it remove from their bill?
Fast PTO Docs bulk pulls patents and file histories, work you'd normally hand to a secretary who'd take hours over lunch. Our website does it at a much lower price than commercial services, and it helps people who don't have that data on hand start experimenting with AI. It matters for billing too, since waiting on someone else to gather files means you can't start work, or you end up doing tedious data entry yourself at an expensive rate.
You said most people in your space have no idea it's possible to use agentic AI to find patent infringers. How does that actually work?
We use a framework that first figures out what industry a patent sits in, since any patent could apply anywhere. From there we identify the players and pull product and model numbers from press releases and websites. Clients often help too, since they know their industry and can point us in the right direction. Once the AI has good direction it does excellent research, but we still stage it, test it, and audit everything through dashboards.
Patent law runs on precision and AI runs on probability. How do you keep hallucinations out of work product a court will see?
Adversarial agents are key. I use Claude a lot, and one agent can play the other side while another acts almost like a judge or neutral evaluator, then they all fight it out. Once you run things through a robust system like that, I haven't seen hallucinations like you get from just talking to a raw chatbot with no back and forth. There's still important judgment for lawyers, but this pushes accuracy well beyond what a human could achieve alone.
You built a proprietary system harvesting PTO file histories at scale. What was hard about that in ways nobody would guess?
At first the biggest challenge was CAPTCHAs. We built a central dashboard that fed CAPTCHAs back to us so a large array of agents could keep harvesting while we solved them manually. Later the PTO added an API, but it kept changing and was unstable, so we had to keep developing against a moving target. It was also custom software I had to pay a developer to build the old fashioned way, which cost a lot back then.
Law firms bill by the hour and your product sells speed. How do you sell efficiency to an industry that's paid for inefficiency?
A lot of hunting for infringement is work no one would ever pay a law firm to do that broadly, it costs too much, so in house teams keep the scope narrow. AI lets you widen that scope, but people need confidence it's real, which is what our dashboards are for. When someone sees an actual claim chart and the reasoning behind a rejection, they see it's really finding things, and that trust is what actually sells people.
What happens to the patent litigation associate over the next five years?
I think the top firms and the top associates are still going to be fine, it's other parts of the profession that will look different. It has actually always been true that early associates aren't considered worth their money yet, so that part of the dynamic isn't entirely new, AI is just accelerating it.
Building something daring? Sunny talks to founders like this every day. Fifteen minutes to see if your story belongs on the stage.
Claim your pre-interview