Roche: Transforming patent workflows with agentic AI

Roche's patent department did labor-intensive keyword-based legal research across roughly 50,000 European Patent Office cases, leading to incomplete insights and slow response times. Starting in 2021, Roche used Dataiku's platform, beginning with a pilot called AskWhitebook (semantic search over EPO appeal cases) that evolved into Themis PatAI, an agentic AI interface unifying several GenAI projects via Dataiku's LLM Mesh, Knowledge Bank, Prompt Studios and Agent Connect, letting attorneys build workflows without IT bottlenecks while keeping governance over sensitive legal data. Time to build new GenAI projects fell from months to days, saving an estimated $100K-$250K annually in attorney hours and $375K-$475K in avoided consultancy costs; it was deployed to 80 European patent attorneys with plans to expand to 250 globally. Andres Buser of Roche said 'the entry barrier to starting with AI and data analytics tools is remarkably low when using the Dataiku platform.'

Dataiku is a low-code/no-code AI platform whose Prompt Studios and LLM Mesh let teams prototype and A/B-test candidate LLM approaches before committing to a production workflow.

This sits in the Design & Stress-Test stage of the AI use-case lifecycle: Named examples from the tools pharma teams use to scope, prototype and benchmark an AI use case before committing to build it — data-catalog, sandbox and evaluation platforms.