Stilta is building patent litigation software that automates the heavy research and analysis behind intellectual property disputes. The startup raised a $10.5 million seed round led by Andreessen Horowitz, with backing from Y Combinator and operators linked to OpenAI, Legora, and Lovable. Patent litigation still involves large amounts of repetitive review work. That process is often slow, expensive, and difficult for companies to pursue. Founded in 2026 by Oskar Block, Tobias Estreen, Petrus Werner, and Oscar Adamsson, Stilta wants to turn that bottleneck into software.
What is Stilta and how does the patent litigation software work?
Stilta’s patent litigation software starts with a simple input: a patent number and any supporting material the lawyer wants to add. From there, its AI agents break down the claims and search for conflicting patents and other relevant references. They also pull prosecution and court history, then assemble a report with claim charts and source-level support that a legal team can actually use in a case file.
The product is more technical than the usual “legal copilot” pitch. Petrus Werner, Stilta’s co-founder and CTO, described a system that runs in secure, isolated Kubernetes environments and combines base agent tools like file access, terminal work, and code generation with litigation-specific workflows. The agents operate on a data layer that spans more than 170 million patents from 100-plus jurisdictions. It also includes US and European prosecution and litigation history, 250 million scientific publications, and internet archive material reaching roughly 1 trillion web pages.
For attorneys who don’t want a black box, the customer experience looks deliberate. Stilta frames the software as attorney-led, not autonomous. Users can steer the analysis, test different theories, and inspect the evidence trail rather than accept a single generated answer. That matters.
Who founded Stilta and why did they build it?
A dinner-table idea turned into a company
The founding story is unusually direct. Block had already started one company at 18, building machine learning models for sports betting, then moved into consulting work around AI adoption before taking a role at an autonomous trucking company. There, he got a close look at how manual the patent process still was. The real spark came over dinner with Estreen, when Estreen’s father — a patent attorney — described a workday full of repeated document review done the same way for decades. That conversation led Block and Estreen to pull in Werner and Adamsson and launch Stilta.
Why this team has some real market fit
Block’s background matters because he’s not coming at this as a pure legal outsider who just discovered AI agents last month. He’d already spent time on hard data problems, enterprise AI integration, and applied automation inside a complex industrial company. That gives him some credibility when the pitch is less “AI for lawyers” and more “AI for high-stakes, evidence-heavy workflows.”
The broader founding team also fits the category well. Stilta describes the four founders as engineers from McKinsey and QuantumBlack who spent years deploying secure AI systems for demanding enterprise customers. Werner is listed as co-founder and CTO. The company’s early messaging makes clear that the group is trying to sell not just model capability, but enterprise reliability and auditability.
Early traction, status, and the seed round
For a very young company, Stilta is already giving off more than slide-deck energy. Y Combinator lists it as an active W26 startup founded in 2026 with a team size of 4. Stilta already works with some of the largest IP firms in the world and with several Fortune 500 enterprises. It has also been welcomed into Mannheimer Swartling’s innovation lab in Sweden.
The financing is a $10.5 million seed round announced on Tuesday, and Andreessen Horowitz led it. Alongside a16z, the cap table includes Y Combinator and operators from OpenAI, Legora, and Lovable. The company hasn’t laid out a detailed public spending plan, but the open roles across engineering, design, go-to-market, and patent expertise make the likely priorities obvious.
How Stilta stacks up against competitors
Stilta isn’t alone. Solve Intelligence and DeepIP are two clear comparables, but they aren’t identical businesses. Solve Intelligence leans heavily into invention harvesting and patent drafting. It also covers prosecution and prior-art analysis. DeepIP pitches a broader end-to-end patent workflow covering invention capture, drafting, prosecution, risk assessment, agentic search, portfolio intelligence, and landscaping.
Stilta’s angle is narrower and sharper: litigation and dispute work. That focus matters. Instead of trying to be the operating system for every patent task, it’s going after infringement and invalidation research where the output has to be argument-led, traceable, and usable under real pressure. The legacy alternative here isn’t just older software. It’s armies of lawyers and analysts doing manual review, plus commercial search tools that often stop at retrieval instead of building a case theory.
There’s also a speed-and-quality claim behind the pitch. In a test on 40 real PTAB institution decisions, Block said Stilta reached 71% petition recall in 20 to 30 minutes, compared with 18% for general-purpose LLMs and roughly twice the median performance of five large commercial patent search tools. That’s still a company-run benchmark. Take it as directional, not gospel.
Why are investors backing this patent litigation software startup?
This seed round matters less because of the dollar amount and more because of what it validates. Patent work is a brutal category for new software. Buyers are cautious, evidence standards are high, and the cost of a wrong answer is a lot higher than in lighter legal workflows. So when a16z leads a seed here — with YC and a network of well-known operators around the deal — it reads like a bet that patent-specific legal AI is ready to move past novelty.
It also gives Stilta a shot at turning early access into actual market share before the category gets crowded. The company is already hiring across engineering, product-adjacent patent expertise, and GTM roles in Stockholm and New York. The immediate job isn’t inventing a new narrative. It’s expanding deployment, tightening product reliability, and proving that specialized patent intelligence software can earn trust inside top-tier firms and enterprise legal teams.
There’s a subtler reason this round matters. If the product works the way Block says it does, the addressable customer isn’t only outside counsel. It includes companies sitting on patents they’ve never enforced, licensed, or properly analyzed because the economics never made sense before.
How big is the market for patent litigation software and legal AI?
The market backdrop is straightforward: there’s more IP to analyze, and legal buyers are finally spending real money on AI. WIPO said innovators filed a record 3.7 million patent applications worldwide in 2024, up 4.9% from 2023 and nearly double the level seen in 2010. More patents in the system means more prior-art work and more portfolio review. It also means more opportunities for disputes.
On the software side, Grand View Research estimated the global legal AI market at $1.445 billion in 2024 and expects it to reach $3.918 billion by 2030, a 17.3% CAGR. The broader global legal technology market was estimated at $28.7 billion in 2025 and is projected to hit $69.7 billion by 2033. Those numbers won’t map cleanly to patent litigation tools alone, but they show why investors are circling legal workflows that were considered too specialized a few years ago.
The adoption curve is moving too. Thomson Reuters said average law firm spending on technology and knowledge management rose 9.7% and 10.5%, respectively, on top of already elevated 2024 growth. It also found firms with a visible AI strategy were 3.9 times as likely to report at least one form of ROI. So the timing here isn’t random. Legal buyers are no longer just experimenting. They’re budgeting.
Final take on patent litigation software
Stilta is still early. Very early.
But this isn’t one more generic legal AI wrapper dressed up in patent language. The company is making a focused bet that patent litigation software should behave less like search and more like a coordinated research team. It should build arguments, surface evidence, and leave the attorney in control. If that holds up outside founder demos and early design partners, the next question is whether Stilta becomes a specialist tool for elite firms — or the default workflow for a much larger chunk of IP litigation.
Read how Ocean raised $28M led by Lightspeed Venture Partners to build agentic email security software that investigates suspicious messages before employees interact with them.
FAQ
– What funding did Stilta raise?
Stilta raised a $10.5 million seed round. Andreessen Horowitz led the round, and the investor list also includes Y Combinator plus operators connected to OpenAI, Legora, and Lovable.
– How does Stilta’s patent litigation software work?
It takes a patent number and supporting materials, then uses AI agents to analyze claims and search for prior art and related evidence. It also pulls filing and court history, then generates reports and claim charts with traceable support. The system is built for interactive attorney review rather than one-click automation.
– Who founded Stilta?
Stilta was founded in 2026 by Oskar Block, Tobias Estreen, Petrus Werner, and Oscar Adamsson. Block is the CEO, Werner is the CTO, and the team’s background includes enterprise AI work at McKinsey and QuantumBlack, plus Block’s earlier startup, consulting, and autonomous trucking experience.
– Is patent litigation software part of a big market?
Yes. It sits inside a fast-growing legal AI and legal tech category, with the legal AI market estimated at $1.445 billion in 2024 and projected to reach $3.918 billion by 2030, while global patent filings hit 3.7 million in 2024. That mix of rising software spend and rising IP volume is why startups like Stilta are getting funded now.
