Where Automation Ends and Expertise Begins: Agentic AI in Patent Search

About This Webinar

Agentic AI is already streamlining workflows across the patent lifecycle, but patent search is where general AI has the most limitations. Patents, especially scientific patents, do not use standard language, and references are often obscured sources that are not text-based. Defensibility requires consideration of the entire landscape, not just the five closest matches. Join Kathy Van der Herten, Director of Product Management, and Brian Habersberger, Lead GenAI Engineer at CAS, as they show where off-the-shelf AI breaks down and discuss what it takes to make AI a trustworthy accelerant rather than a liability.

In this webinar, we examine the following key areas:

• Why patents defeat plain-language AI — how non-standard terminology, genus/Markush claims, and sequence data put critical prior art outside the reach of a text-only model.
• The importance of searching for the whole landscape — why freedom-to-operate, invalidity, and landscape work depend on comprehensiveness, and why naive similarity search is built to fail that test.
• What makes AI trustworthy for IP — the case for indexed, expert-curated data as the channel that carries real domain depth into a generalist model, illustrated with CAS Newton for CAS IP Finder®, powered by STN™.