What’s actually changing about patent prosecution

The conversation about AI and patents right now is mostly about drafting. Are AI tools good enough to draft a patent? Will attorneys still be involved? These are reasonable questions, but they fail to capture broader structural shifts poised to transform the IP industry. The bigger question is what AI does to the business model underneath patent prosecution: how the work gets structured, who delivers it, and what the buyer is actually paying for.

Emergence Capital calls it “AI-native services.” Instead of buying AI tools or working with traditional services vendors, companies partner with a new type of vendor that has built their service delivery method from the ground up using AI. The line between technology companies and service providers gets blurred as almost everything becomes outcomes based. What changes is what gets billed and how.

Patent prosecution is one of the harder categories to map onto this shift. Drafting claims on a frontier technology requires judgment. It depends on a senior attorney’s understanding of the field, the company’s strategy, and the market. None of that will be automated entirely, and any version of this story that pretends it will be is wrong.

But that’s not all of patent work. A meaningful portion of prosecution is precision work — figure consistency, terminology harmonization, mapping specification elements to drawings, USPTO formalities. That work has been priced as if it were judgment for a long time. It isn’t.

Also worth noting as it often gets lost in conversations about AI replacing lawyers – part of what a company actually buys when it hires a patent attorney isn’t drafting capacity. It’s the transfer of risk to a person whose name goes on the filing and who is accountable for what’s in it. Any new model that ignores this is going to fail, and clients are right to be skeptical of anything that does. What’s changing is everything around the attorney. The attorney isn’t going anywhere.

The two models companies use today

When I talk to IP leaders at growth-stage companies, they’re choosing between two models, and neither is fully enabling them to do the work they need at the pace they require.

The first model is in-house. Hire patent attorneys, bring drafting under your own roof, control your own pace. This works at small scale. It runs into trouble fast as filing volume grows. The IP leader either keeps adding headcount, which most CFOs don’t love, or they hit a ceiling. And the ceiling is lower than it looks, because IP leaders rarely get to spend all their time on IP. Commercial contracts pull them in. Board materials, NDAs, trademark issues, the occasional employment matter. The practical bandwidth is always a fraction of the headcount.

The newer twist on in-house is purpose-built AI drafting tools. The pitch is that a small team can produce more applications by using AI to generate first drafts. In my experience these tools are useful, but they don’t actually solve the bandwidth problem. They shift it. Time that used to go into drafting now goes into reviewing AI output. The bottleneck moves. The bottleneck doesn’t go away.

The second model is the law firm. Most companies pick this one. The market has been pushing on cost for at least a decade, and most firms responded by moving to fixed-fee patent prosecution. That genuinely lowered per-application pricing, but it didn’t change how the work gets done.

The model is still associate-heavy drafting with partner review. To make the fixed fee work, firms compressed the senior time. For many firms, this meant 40 hours of associate work and 1 hour of partner review. The application gets filed. The quality follows the math. Office actions get longer. Prosecution histories get messier. And five years later, when somebody actually needs the patent in a litigation or a licensing conversation, the cleanup is expensive. We’ve written about where that budget goes.

Both models suffer from a version of the same underlying problem. Neither has changed what a patent actually costs in real economic terms. (In-house cost is hidden in salaries; firm fees are still high.) And neither has changed the experience of running an IP function that’s trying to grow at the speed of the business.

The model that’s starting to work

This is where the AI-Native Services model becomes particularly interesting for IP work.

Here’s how it actually works.

A senior patent attorney sets the claims strategy based on what they know about the technology, the company, and the competitive landscape. This is the irreducible judgment work, and nothing about the new model changes it.

AI then drafts the claims, figures, and specification. The AI tools are curated specifically for the attorney, with outputs accompanied by confidence scores, notes on where to focus review, etc. The attorney reviews the draft and adjusts. Optimizing claim breadth. Catching where the model misunderstood the technology. Making the calls only an attorney can make.

The attorneys, importantly, are trained to use the tools. They know where the model is strong and where it’s weak. They don’t rubber-stamp the AI output, and they don’t reconstruct the application from scratch either.

What you end up with: roughly 5 hours of senior attorney time produces a better application than the legacy fixed fee law firm model. In other words, when the strategic work gets concentrated in senior attorney time and the production work gets concentrated in tools that don’t make consistency errors, you get a better result. This has been our experience on the matters we’ve taken through this model, and the customers I’ve worked with consistently report the same thing.

There’s a second piece of the new model that isn’t discussed as often, but I think is the more durable part over time.

AI-native services providers are also software companies. They have to be. The IP management system that runs the filing workflow is the same system the in-house team uses every day for everything else — tracking the portfolio, managing deadlines, capturing strategic notes, running board reports. That means the team’s data lives in infrastructure the team controls, not in a law firm’s docketing system that the team can read by request and export with friction.

For a growing IP function, this is the thing in-house teams have actually wanted from their outside counsel for two decades and never had. The ability to scale the function without locking the company into a single outside provider. The ability to switch service providers without losing institutional memory. The ability to actually run the IP function as a function, rather than as a project that lives mostly inside someone else’s matter management system.

What changes when filing isn’t rationed by cost

It stands to reason that a 10X drop in patenting cost will lead to a comparable increase in the volume of patents filed. The reality might surprise you. Yes – a lower cost model will improve access to IP protection for earlier stage companies and may even lead to some bad actors flooding the market with patents. But the truth is – companies in IP-intensive industries already find a way to pay for the patents they absolutely have to file. The protection that’s critical to the business gets filed. The constraint isn’t the must-file applications. It’s everything else.

Here’s the version of this conversation I’ve had a dozen times. A growth-stage hardware company wants to file 30 patents this year. At roughly $20K per application through a fixed-fee firm, that’s $600K. The IP leader knows the company has more than 30 patentable ideas — it’s closer to 70 if they really sat down and inventoried. So they triage. They file 30. The other 40 live in a notes doc that nobody opens again. The defensive filings get cut. The continuations get dropped because the team is already at capacity on higher-priority work. This is the decision IP leaders make every quarter, and it has very little to do with the quality of their judgment about what’s worth protecting.

When costs drop, the actual change is how budget is allocated. We recently surveyed IP attorneys at high-growth companies and asked a version of the question: if patent costs dropped by 10x, what would you do? 60% told us they would file roughly the same number of patents and spend the savings on higher-value strategic work. Freedom-to-operate analysis. Portfolio strategy. Competitive landscape work. Licensing readiness. Enforcement readiness. The work that requires senior judgment and actually drives commercial outcomes.

That redistribution is the real shift. The companies moving first on this aren’t doing it to lower the IP line on the budget. They’re doing it to free up capital for the legal work that compounds.

So where does this leave traditional IP Law firms?

The smart firms have wanted out of the prosecution business for years. Margins are bad. The work doesn’t really use the firm’s senior expertise the way upstream and downstream work does. The economics of the new model push in exactly the direction the smart firms already wanted to go: prosecution shifts toward AI-native providers, strategic IP counsel stays with the firms best positioned to deliver it, and both ends of the market get better.

The bull case is straightforward. An IP function that gets to spend its budget on strategy instead of production, with portfolio data the in-house team actually owns. I think that’s where this is going, and I’d rather IP leaders see it coming than have it arrive.