Patlytics co-founder and CEO Paul Lee joined Greg Lambert, CIO of Jackson Walker, on The Geek in Review podcast to discuss what happens when an entire legal field gets left behind by the first wave of legal AI, and what it took to build something specifically for that field instead.

The conversation covers a lot of ground: the 100-plus conversations that shaped Patlytics before it existed, why purpose-built beats general-purpose in patent work, and a candid read on who actually captures the value when AI makes legal work faster.

Key Takeaways

• Patlytics works with roughly 55% of the Am Law 100 and hundreds of corporations across biotech, pharma, and technology.
• Paul's research phase involved conversations with more than 100 patent attorneys before building anything.
• General-purpose legal AI gets patent work about 80 to 85% of the way there. Paul argues the bar in patent work is 99%.
• A top-five Am Law firm that rarely used flat-rate arrangements now offers them on certain matters where Patlytics is in use, with more predictable margins for the firm and more predictable bills for clients.
• Attorney sentiment on AI has flipped from roughly 20% optimistic a year ago to 95% optimistic today, in Paul's estimation.

Where the Idea Came From

Paul didn't come to patents through a law degree. He came to it through venture capital, then a series of unrelated moments that kept pointing the same direction. He first noticed the scale of patent disputes as a student at Waterloo, watching the Apple-Samsung litigation play out in the news. Years later, as a VC, he kept seeing portfolio companies burn serious money on IP disputes and couldn't square why the costs ran so high. His co-founder, Arthur Jen, was living the same problem from another angle, handling patent litigation while building his prior company, Magic. The idea took off after Bob Steinberg, then chair of the IP litigation practice at Latham & Watkins and now a Patlytics advisor, invited Paul over to talk through where he saw the opportunity. Paul is careful not to call it a single lightbulb moment. The same problem kept surfacing from three directions until it was hard to ignore.

From there, Paul did what a VC background trains you to do: he talked to people. More than 100 patent attorneys, looking for patterns. He found a practice buried in inefficiency that nobody outside it seemed to notice. Attorneys were spending over 10 hours drafting a single patent specification. Claim charts for litigation took heavy effort, layers of scrutiny, and constant back-and-forth. Bring drafting, prosecution, portfolio work, and litigation prep under one platform, Paul realized, and the value adds up across the entire patent lifecycle.

Purpose-Built for IP vs. General-Purpose Legal AI

Greg pushes Paul on a real tension in legal tech right now: broad, general-purpose legal AI platforms versus a platform built specifically for IP. Paul doesn't dismiss the generalists. He credits them with much of the transformation happening across legal right now. But his argument is that patent work has a ceiling those platforms consistently hit. Even with heavy customization, general-purpose legal AI tends to land around 80 to 85% of the way to what a patent matter actually needs.

Paul's frame for why the gap matters: the patent ecosystem is an F1 race, and the job is getting to 99%. When Patlytics started, he thought 85 or 90% would be enough. That's no longer the benchmark, and the benchmark moves every year. His one-word answer to what earns a permanent place in a firm's stack is relevance: is the experience 10 times better than the way the work was done before? That's the bar he holds the platform to, workflow by workflow.

Where Human Judgment Is Needed

Paul is direct on this point: full automation, where you type something in and a claim chart or application comes out the other side ready to file, has never worked out, and he points to recent evidence of people being pulled into malpractice matters over exactly that kind of shortcut. His position is that Patlytics is built around human-computer symbiosis. AI handles the volume and the first-pass work, but the practitioner still reviews, signs off, and owns the outcome, with checkpoints built into every workflow from prosecution to portfolio management to litigation.

That discipline runs into a new pressure, though. Clients now expect work as good as it would have been after many hours of manual effort, delivered in a fraction of the time, ideally at even higher quality. That squeeze lands hardest on IP attorneys, where, as Paul puts it, each word matters so much. He says the work has given him a new respect for what patent attorneys do.

Who Actually Captures the Value

Greg asks the question a lot of legal tech conversations avoid: when AI turns what used to cost thousands of dollars in attorney time into something dramatically faster, who keeps the saved value, the client or the firm? Paul's answer is that the market hasn't settled it yet. But he points to a concrete example: a top-five Am Law firm that historically rarely used flat-rate arrangements now offers them on certain matters where Patlytics is in use. The firm gets more predictable margins and cycles through matters faster. The client gets more predictable, lower bills. Paul calls it a win-win and expects to see more of it.

Prosecution is where the math is starkest. Costs have risen for a decade while pricing has stayed flat or drifted lower, leaving thin margins. Taking hours out of the work changes that calculation, and Paul says partners have told him they had a record year because of Patlytics and the tooling they put in place.

See how firms are turning hours saved into margin. Read the case studies

On the in-house side, Paul describes teams finally getting bandwidth for work that was always valuable but never fit the calendar: infringement analysis before filing a continuation, faster portfolio review ahead of an M&A decision, earlier visibility into litigation posture before a Markman hearing even comes into view.

The Sentiment Shift, and What's Next

Asked what's changed in the last year, Paul doesn't hesitate. A year ago, he estimated IP attorney sentiment toward AI ran roughly 20% optimistic, 80% skeptical. Today, he puts that at 95% optimistic. Firm leaders who told him they would not look at this for five or 10 years, if ever, now call back with a line he quotes: "Scratch what I said. Could we start something tomorrow?"

Paul's forward-looking concern is economics, not adoption. As free experimentation gives way to consumption-based pricing, he expects a correction: budgets are budgets, and firms and clients alike will need to show that a given AI-assisted workflow pays for itself rather than assuming the value is obvious. Token-maxing, as he puts it, is probably not going to be a thing anymore.

Listen to the Full Conversation

Paul and Greg also get further into the Patlytics origin story and Paul's predictions for the next few years. Listen to the full episode or read the transcript here. And visit the Geek in Review’s page for more of TGIR’s podcasts. The full YouTube video is also available below:

See the platform run on your own patents. Book a Demo