The productivity gains from AI in patent practice are real. Drafting time is down. Prior art searches that once took days take hours. Office action responses that required senior-level review from the start are being turned around faster—and, in many cases, at a higher-quality baseline than before.

So why are patent professionals still debating about who actually benefits?

That question was at the center of a recent webinar hosted by IPWatchdog and their President & CEO, Gene Quinn, in partnership with DeepIP, bringing together five practitioners to discuss the economics of AI in patent practice. What they found was less a clean answer than a more important question underneath it: 

“What are you actually trying to accomplish?”

‍

‍

The Framing Matters More Than the Savings

When the conversation opened, François-Xavier Leduc, CEO and Co-Founder of DeepIP, set a tone the rest of the panel largely built on: "I don't believe in winners and losers. It's a fundamental revolution that impacts everything—organizations, operating models, margin structure, billing models. Like every revolution, there is an amazing opportunity for those who embrace it."

That optimism was broadly shared. Bryan McWhorter, Partner at Knobbe Martens, said, "I am an AI optimist. I think AI will supplement, not replace, lawyers. I think it will reduce costs. I think it will increase profits. I think it will grow the pie for everyone."

What the panel pushed back on wasn't AI's value, but the assumption that value flows automatically, without intention.

Eli Mazour, Of Counsel at Foley & Lardner and Creator of the IP podcast Clause 8, put it plainly: "Figure out why you are obtaining these patents in the first place. And that's when you should have an honest discussion with your outside counsel about how they're actually using AI."

The answer to that question—whether the goal is cost reduction, portfolio quality, speed to file, or litigation resilience—shapes everything about how AI should be deployed and how its gains should be measured.

‍

Quality Is Where AI Delivers Most

When the webinar audience was polled mid-session on what matters most—cost savings, time efficiency, or quality—quality won decisively. 

It's also not surprising to practitioners who've been using the AI tools built for patent work. McWhorter offered a concrete example: "I can go to a tool like DeepIP and have it scan the entire document for minor typos, for reference: numerals being off. I can have it do a comparison of the disclosure documents and really, fundamentally, make sure that I've covered everything."

‍

‍

That kind of systematic review—checking every reference numeral in a 200-page application, confirming disclosure completeness against every claim—was previously too time-consuming to do routinely. 

McWhorter explained why this matters economically: patent prosecution costs have been largely flat for roughly 15 years. That pressure has quietly eroded the time available for thoroughness. "Most of the time, people that I work with are not proofreading every word of a 200-page draft application." While also accelerating the work, AI simultaneously restores quality that economics had squeezed out.

The implication for pricing conversations is significant. Clients seeking cost reductions may already be receiving a quality upgrade they haven't accounted for.

‍

The Practitioner's Expertise Is What Makes AI Work

Across the panel, a consistent theme emerged: the quality of AI output in patent work is inseparable from the expertise of the person directing it.

Ryan Phelan, Partner at Marshall, Gerstein & Borun, noted his firm has demoed more than 40 tools. He was specific about what that looks like in practice. AI built specifically for patents can account for the legal standards an application needs to survive, but that judgment should be owned by the attorney.

"You want to say certain things in your specifications so you can meet the three-part Federal Circuit test," he said. "As a practitioner, I have built instructions into my AI tool to make sure that each one of my clients' patents gets that vital §101 treatment. Because that could be the thing that saves your patent from being rejected constantly under §101, gets you a faster allowance, and makes it a litigation-quality patent."

The value, in other words, isn't in the tool alone, it's in what an experienced attorney brings to it. Knowing what the Federal Circuit needs, understanding how to build §101 defensibility into a spec from the start, keeping the end game in view throughout patent prosecution: those are the inputs that turn AI output into litigation-quality work.

‍

The Disclosure Bottleneck, and Why It's Actually an Opportunity

One dynamic the panel flagged that tends to get overlooked: AI isn't just changing how attorneys work. It's changing what lands on their desks.

"In the past, we got very little invention disclosures. Engineers and inventors are very busy, and they don't have a lot of time to prepare robust invention disclosures," Phelan said. "But now, with AI tools, we get incredibly robust invention disclosures, and it's created a bottleneck in the form of trying to find a needle in a haystack with respect to what the true invention is."

That bottleneck is real. But Leduc framed it as a structural problem with a structural solution, one that points toward where AI delivers its deepest long-term value.

"Patent work is incremental. And today, there are huge silos. The engineer is inventing, preparing a disclosure, handing it to the IP counsel, who hands it to outside counsel—and everyone has to restart from scratch." The context built at each step evaporates at the handoff.

The answer, in Leduc’s view, isn't faster processing at individual steps. It's continuity across the entire lifecycle. "We believe in a tool where R&D can work and build a context where in-house IP teams can select and build incremental context to this initial disclosure, which will be handed over to the outside counsel with the rest of the context." From first disclosure through prosecution to portfolio pruning, the value compounds when nothing is lost between stages.

Mazour saw an adjacent opportunity in this same dynamic. "With AI, you actually have a team working on whatever project you have." Where patent prosecution has historically been solitary work—drafting alone, responding alone—AI creates a dialogue. "You can discuss what the possible continuations are that we can file for this application. You can have that kind of dialogue with AI as you're working through it, and you don't have to worry about bothering somebody 10 times."

That shift is a real change in how the work feels, and in the quality it produces.

‍

Adoption at Scale Requires More Than a Good Tool

Individual practitioners adopting AI is one thing. Rolling it out across a law firm of 30, 50, or several hundred attorneys is another problem entirely.

McWhorter was clear about this: "You don't buy AI, you build institutional capability." What that means in practice is integration—into document management systems, into existing workflows, into interfaces attorneys already use.

"Things like incorporating into Word, which may incorporate into a document management system, finding ways that we can take these tools and expand them to teams and make them intuitive—that's a really big part of successful AI adoption," he said. Tools that require attorneys to change how they work won't move an organization.

Phelan added that adoption also varies by technical area. Computer science-trained practitioners tend to pick these tools up quickly. Those working in biotech or life science may be slower to engage—not because the tools aren't useful to them, but because the interface needs to meet them where they already are.

‍

What Should Billing Look Like?

This question generated a passionate debate in the chat amongst attendees. The panel didn't land on a single answer—and that was probably the honest outcome. But a few principles emerged:

‍

Tiered Offerings Make Sense 

Phelan described developing structured options at his firm: different configurations of AI assistance, attorney review, and application depth depending on what the client actually needs. A one-off filing for a low-priority invention is a different project than a patent anchoring a core product.

‍

Expectations Need to be Set Explicitly 

McWhorter's closing call was direct: "I would encourage anyone who's in-house to talk to their outside counsel about use of these tools. Outside counsel, talk to in-house counsel, make sure expectations become aligned—because until we have a level of understanding that's shared among us, we really can't reach those efficiencies that I do think will come."

‍

Start with the Tools Yourself 

Mazour's practical recommendation: don't rely on secondhand accounts. Get hands-on. His suggested entry point was continuations. "You already have a defined application. And if you know what you have, an example of a lot of claims…AI or LLM platforms have something to work with." This way, you can evaluate what the tool actually produces before negotiating around it.

‍

The Inventorship Question Isn't Going Away

An issue the panel raised cuts through all of this: as AI becomes more embedded in claim drafting, the question of inventorship is moving from theoretical to practical.

Phelan anticipated that deposition transcripts in future litigation will routinely ask whether AI was used to generate claims—and that inventors will need to demonstrate, under penalty of perjury, that they contributed to the conception of the claimed invention, which remains the legal standard for inventorship.

"I think it's very dangerous to have an inventor wholly generate a claim set from nothing and then submit that as the invention," Phelan said.

This isn't a reason to pull back from AI in prosecution. It's a reason to use it with a practitioner driving the process, one who understands where human judgment is legally required, not optional.

‍

The Cost of Standing Still

Leduc opened the session by rejecting the winners-and-losers framing—and he closed it the same way, this time with a sharper edge: there are no permanent winners or losers here, only those who engage deeply with the transformation and those who wait.

"The cost of inaction—the cost of not getting your hands dirty and deeply understanding what it does, what it doesn't, and how it might reconfigure your relationship with your counsel or with your clients—is huge," he said. "Don't underestimate how deeply it will reconfigure everything."

The economics of AI in patent practice won't resolve in a single billing conversation. They'll resolve through a shared understanding—between firms and clients—of what quality actually requires, what AI genuinely makes possible, and where the expertise of an experienced practitioner remains the difference.

That understanding starts with the conversation. And by all accounts, the profession is just getting started.

‍