Patent Attorneys by Day, Coders by Night: Sterne Kessler Launches ‘Homemade’ AI Assistant

The IP specialist bypassed vendors and turned to its own ‘tech geeks’ to create an in-house AI solution for drafting patent applications. WIPR meets the attorney coders who did it for themselves.

When US patent law firm Sterne Kessler needed an AI solution, they didn’t have to look far for talent.

With an office full of self-described “tech geeks” who code for fun, developing their own proprietary AI system was a natural next step.

The firm’s new Patent Assist AI tool wasn’t outsourced to vendors or consultants—it was created by the same attorneys who are just as at home writing software as they are drafting claims.

“We hire technical specialists—really nerdy, geeky folks—and they’re thrilled to death that we’re doing this,” says Michael Specht, director and chair of Sterne Kessler’s Electronics Practice Group.

Daniel Block, a director in the firm’s Electronics Practice Group and one of the tool’s primary developers, sees the project as more pleasure than work.

“It doesn’t really feel like I’m spending time on this because I’m energised by it. It’s sort of my de-stressor from my day-to-day legal work,” he explains.

Launched two weeks ago, the early version is already being rolled out to clients, with positive initial feedback, says the Washington, DC firm.

Vendor tools ‘created problems’

The journey began about a year ago at a firm retreat, where partners Block and Joseph Mutschelknaus, director in the firm’s Electronics Practice Group, recognised AI’s growing impact on legal practice.

“Most folks immediately understood that it was going to have a significant impact on patent drafting. Generative AI, we thought, might be particularly well-suited for this, given the right guardrails and considerations,” explains Block.

The firm evaluated existing tools through a market study, testing eight vendors using identical inputs in a blind comparison.

They even benchmarked against public LLMs like ChatGPT and Claude.

The results were striking—their internal control ranked second in quality, prompting them to consider developing their own AI solution.

A small team at Sterne Kessler developed a proof-of-concept in just one weekend, showing enough promise to get management’s approval for an in-house development initiative.

After vendor interviews, “we identified several issues with how they were approaching the problem,” explains Block.

“The biggest concern, from a patent attorney’s perspective, was AI being named as an inventor.”

Many tools were automatically generating patent claims, raising concerns about whether AI could be considered an inventor under existing legal frameworks.

As a result, legal compliance became a priority for Sterne Kessler’s approach.

The firm wanted to integrate AI into its established patent drafting workflow, rather than replace anyone.

“Our goal has never been to replace patent attorneys. Some vendors seem to have that objective, but our approach is to assist attorneys by allowing them to focus on the aspects where human expertise adds the most value,” notes Block.

What challenges did you face?

Developing an AI tool in-house comes with its own set of challenges, particularly tackling legal, ethical, and technical concerns.

So that’s exactly why Sterne’s team did it themselves.

“By developing our own tool, we gained greater control over these issues, including data confidentiality, ensuring alignment with patent office rules, and more,” explains Specht.

“That has really been a parallel effort—the technical development of the tool alongside what I’ll call the legal and ethical development—to ensure we remain compliant with the rules.”

Another major challenge was addressing scepticism around AI, particularly over confidentiality and privacy concerns.

To address this, Sterne Kessler prioritised transparency and security.

By developing in-house, it was able to reassure clients that their data would remain confidential and would not be used for training AI models.

“We can confidently say that because we built it,” adds Block, who says this helped the firm get clients on board sooner than if it had gone the vendor route.

Convincing the clients

Some clients, particularly smaller ones, jumped in once they understood how it worked and were reassured about key concerns such as confidentiality.

For larger clients, it wasn’t as straightforward—they had more questions, and multiple meetings were needed to get comfortable.

These larger clients still had to convince legal departments and IT teams. Some were based in other jurisdictions where there are different privacy issues they have to deal with, making the process longer.

Specht attributes part of their success in winning over clients to the initial research the team conducted, showing the market comparisons and demonstrating how the firm had addressed key challenges like internal control and quality.

The biggest hurdle Specht faces when talking with clients is that many of them aren’t AI-ready.

“They haven’t developed their own internal policies and procedures for how to deal with AI, and so they don’t know what tests to apply,” he explains.

They’re working through that process while the tool is introduced to them, which has actually “opened up additional client opportunities”, says Specht, whereby they advise on AI-related policies and regulations beyond Sterne’s own tool.

Not just a time-saver

For Block, the real win is that the patent attorneys at the firm can allocate their time to add value where humans are best to do so.

“What I mean by that is developing claim sets, making them more robust, covering perhaps more embodiments, and having more time to think about that. Then, also broadening the disclosure by adding more details,” he explains.

Attorneys can now tackle tougher legal issues related to their field and get a stronger patent as a result.

He also notes the efficiency benefits of the tool help overcome “writer’s block”, making it easier to pick up where attorneys left off.

But it’s not just about saving time.

“It allows us to do aspects of the application more efficiently, it gives us the ability to spend more time on those areas where we really can add value,” says Specht.

With clients feeling price pressures on applications, the tool helps the firm “address those price pressures while maintaining a good quality application or even improving the quality.”

An unexpected bonus has been improving the firm’s ability to support clients in the AI, machine learning, and data science sectors.

“By developing this tool, we’ve become far more knowledgeable about the various issues that our clients are confronting,” Specht notes.

“We can now talk the talk and walk the walk with our clients in this area.”

This article first appeared on Life Sciences IP Review.