Vendor-reported figures — source: www.contextwindows.ai
Roche's IP legal team needed exhaustive legal research across roughly 50,000 European Patent Office cases to defend the company's intellectual property. Attorneys spent hours on manual keyword searches that often missed relevant precedents, and a later rollout of isolated, disconnected AI tools left them juggling fragmented tooling instead of solving the underlying problem.
Roche built an AI orchestrator on the Dataiku platform that routes attorney queries to specialized sub-agents within a single chat interface, replacing the fragmented toolset with one unified workflow. The legal team evaluated different prompts and compared token costs across multiple LLMs to tune the system for patent research.
The orchestrated system saves Roche an estimated $100K-250K annually in patent research costs. It also cut the build time for new GenAI projects from months to days, letting the legal team iterate faster on research tooling.
Have a similar implementation?
Share your customer's AI results and link it to your vendor profile.
Submit a case study →
AI Roadmap — Associates spending the week on doc review. Contracts taking two weeks to turn. Both have AI fixes. We'll show you which, in what order. 72 hours.