Our Take
A16Z is betting on agentic AI for IP enforcement, but the real test is whether Stilta's agents actually surface evidence that legacy systems miss—no independent benchmark exists yet.
Why it matters
Patent enforcement remains manual and expensive; if Stilta's agentic reasoning across 180M patents and 250M publications can move the needle, IP teams may finally have a reason to abandon spreadsheets. The founder exit from McKinsey's AI practice and customer names (three of five largest IP firms) suggest serious domain credibility.
Do this week
IP counsel: request a Stilta pilot before Q2 budget lock if you currently use manual search or legacy patent tools, so you can measure evidence density and speed gains on a live case.
Stilta Closes $10.5M to Deploy AI Agents in Patent Litigation
Stilta, an AI patent litigation software company, announced $10.5 million in Series A funding led by Andreessen Horowitz. The round includes participation from Y Combinator and founders from leading AI companies including Sana, Legora, OpenAI, Lovable, and Listen Labs (per the company announcement).
The startup was founded by four engineers who departed McKinsey's AI practice. Since launching in February 2026, Stilta has signed paying customers and active pilots with Roche, Alfa Laval, and Maersk, as well as three of the five largest IP law firms globally. The company will deploy the new capital to hire its first employees across Stockholm and New York: engineers, go-to-market specialists, and patent experts.
Stilta's core product is agentic AI software designed to reason across 180 million patents, 250 million scientific publications, and over a trillion archived web pages. The stated goal is to surface evidence of prior art, infringement, or patent validity that legacy manual review and software tools consistently miss.
Patent Enforcement Has Always Been a Manual Bottleneck
The vast majority of patents are never enforced, licensed, or monetized, according to Stilta's pitch. The barrier is not evidence scarcity but analysis cost and complexity. Patent counsel and in-house IP teams spend weeks or months on discovery and prior-art searches that remain incomplete.
If Stilta's agents can compress that timeline and improve evidence yield, adoption among IP firms and enterprise legal teams becomes plausible. The customer list (Roche, Maersk, three mega-law firms) suggests the company has already convinced tier-one players that the problem is real and the approach worth piloting.
The A16Z backing and founder pedigree (McKinsey AI practice, OpenAI ecosystem) indicate conviction that agentic reasoning is the right architectural fit for patent reasoning. Whether it actually outperforms human search remains an open question; no independent benchmark or case study has been published.
How IP Teams Should Evaluate This
If you manage patent portfolio enforcement or defense, treat Stilta as a pilot candidate, not a proven solution. The company is 4 months old with four founders and no published validation. Request a live case study or anonymized evidence audit from a customer before committing budget.
The pitch is specific: agents reason across three massive datasets to surface missed evidence. Ask for a concrete comparison—how many prior-art references does Stilta find that your current workflow (manual or legacy software) missed on a real case? Measure time-to-evidence and false-positive rate. If the pilot moves the needle on either metric, expand. If not, no harm done.
Large IP firms should also watch whether Stilta becomes a white-label integration partner or a replacement vendor. The customer list suggests Stilta is going direct to enterprises; friction with law firms may emerge.