Your Agreements Are a Database You Can't Query. We're Fixing That

Agreements power every enterprise business, but mismanaging them has a big cost. Deloitte and Docusign found that inefficient agreement management costs businesses nearly $2 trillion in lost economic value annually, and organizations burn more than 55 billion hours on manual, disconnected agreement workflows. This session walks through what it takes to turn agreement repositories into structured, queryable data that can power search, analytics, and agentic AI workflows using Nvidia's open AI platform as the foundation. We'll cover the full pipeline: Extraction: How Nemotron Parse handles the hardest document understanding problems, including contract tables where pricing tiers, SLAs, and payment schedules live in dense layouts that traditional OCR destroys. Embedding and retrieval: How NeMo Retriever's embedding and reranking models make extracted contract data searchable and surfaceable for RAG-based applications. Agentic workflows: How these components connect to power contract analysis agents that can answer questions, surface risks, and act on obligations at scale. Nvidia will present the platform architecture and model capabilities. Docusign will share what we've learned evaluating this stack against real customer agreements: what worked, what broke, and what production deployment actually requires when serving over a billion users. Attendees will leave with a realistic view of where the Nvidia AI platform excels at document intelligence, where the gaps remain, and how to think about building agentic contract workflows in their own organizations. Speakers: Hiral Shah — Docusign; Sean Sodha — NVIDIA.