The legal profession's relationship with artificial intelligence has always been uneasy—perhaps nowhere more so than in patent law, where a single error can invalidate years of work and millions in intellectual property.
Enter Stilta, a San Francisco startup barely two months into public existence, which believes it's cracked the code on what patent attorneys actually need from AI: not magic, but transparency.
The company, fresh from Y Combinator's Winter 2026 cohort, launched in late January with software that applies what's known as "agentic AI" to patent research and analysis. Think of it as AI that iterates and refines rather than spitting out a one-time answer—closer to how a junior associate might work through a problem than how ChatGPT responds to a prompt. The founders describe their product as "Cursor for patent practitioners," borrowing the metaphor from a popular AI coding tool.
And the early numbers? They've caught attention. Stilta says it reached $30,000 in monthly recurring revenue within six weeks, according to the company's Y Combinator profile. That's modest in absolute terms but notable velocity for enterprise software in a notoriously conservative field.
The four-person team has signed customers it describes as leading IP law firms and in-house teams at Fortune Global 50 companies, though it hasn't named them publicly. That discretion is typical in legal tech, where client confidentiality often trumps marketing needs.
Why Patent Work Resists Black Boxes
Patent attorneys operate under a different calculus than most professionals experimenting with generative AI. A missed reference in a prior art search, an imprecise claim construction, a sourcing error—any of these can expose clients to litigation or kill an application outright.
Which explains why Stilta's pitch centers less on speed and more on something harder to quantify: trust.
The platform focuses on patent research and what the company calls "interactive, argument-led analysis." Every output includes full source attribution, allowing practitioners to trace the AI's reasoning backward. It's the opposite of a black box—more like a transparent workbench where the attorney remains in control.
According to materials posted during the YC launch, the system is purpose-built for the "manual, high-stakes nature" of prior art searches and patentability assessments. Third-party product listings suggest Stilta has also developed modules for patent drafting, invention disclosure capture, and portfolio monitoring, though the company's own marketing leans heavily on research workflows.
"AI-native software for patent attorneys, built on agentic workflows to amplify how IP work is done end-to-end," reads the company's positioning. The language is dense, maybe a touch too jargon-heavy for comfort, but it signals the audience: this isn't software for generalists.
Who's Building It
Stilta's founders come from the world of high-level consulting and analytics, not law school. CEO Oskar Block logged time at McKinsey and Goldman Sachs before bootstrapping earlier ventures past $1 million in annual recurring revenue. CPO Oscar Adamsson also worked at McKinsey, advising dozens of CEOs, and studied business and AI at Chalmers University in Sweden.
The technical firepower comes from co-founders Tobias Estreen and Petrus Werner, both alumni of QuantumBlack, McKinsey's analytics division, where they built agentic AI systems. Werner, now CTO, also worked at Amazon Web Services.
It's a team built for pattern recognition and systems-building rather than legal domain expertise. Whether that's an advantage or a liability depends on how well they've listened to their customers.
Y Combinator invested $500,000 as part of its standard deal, according to Swedish Tech Weekly, a startup newsletter that covered the launch in January.
A Crowded—and Cautious—Market
Stilta is hardly alone in seeing opportunity at the intersection of AI and patent law.
DeepIP, a New York- and Paris-based competitor, raised a $15 million Series A in March 2025 and claims its Microsoft Word integration can cut drafting time by up to 50 percent. &AI, from YC's Summer 2024 batch, markets AI agents for patent prosecution and litigation, including claim drafting and responses to office actions.
Incumbents have also moved. Questel offers AI copilot tools under the Qthena brand. XLSCOUT introduced a drafting-focused language model. Some major law firms have gone proprietary: Sterne Kessler debuted Patent Assist AI in March 2025, designed exclusively for its own attorneys.
The rush of products reflects both the technical complexity of patent work and the profession's deep-seated wariness of automation. Patent lawyers don't just need tools that work—they need tools they can defend in front of examiners, judges, and their own malpractice carriers.
Stilta's emphasis on traceable outputs seems calibrated to that anxiety. Whether it's enough to win over a skeptical audience remains to be seen.
The Road Ahead
The company launched its public social media presence only in January 2026, leaning on Y Combinator's network for early visibility. (A LinkedIn post from YC congratulated the team, though such posts are standard for the accelerator's portfolio.) Materials from the launch stress a vision of an "agentic workspace" for IP teams—patent research as beachhead, not endgame.
Sustaining early momentum will be the test. At four people and $30,000 in MRR, Stilta is playing in a market where competitors have raised tens of millions and incumbents have decades of client relationships.
The bet is simple but unproven: that patent attorneys will pay a premium for AI that reveals its reasoning, not just AI that claims to do their work faster. In a profession where trust is currency and mistakes are expensive, showing your work might matter more than speed.
Then again, it might not. Six weeks of revenue is a data point, not a trend line. What happens at six months—or when a better-funded rival ships something similar—will tell the real story.
