Sponsored By: DataHub | Agents That Get it Right: How FIS Grounds Databricks Genie in Enterprise Context

FIS processes 75 billion transactions a year for 20,000+ clients. They're building a client-facing analytics product that lets financial institutions ask plain-English questions about their customers' data, a business capability that simply didn't exist before. The bottleneck wasn't the AI model. It was the context. Fragmented across spreadsheets, Unity Catalog, and Snowflake, context was nearly impossible to iterate on. Every Genie model change meant tearing down the space, restaging it, and resyncing from scratch.

The fix: centralize all business context in DataHub, let Unity handle the technical layer, sync bidirectionally. Because DataHub context is editable and syncable in one place, the team could iterate fast, tuning Genie models without rebuilding every time. With DataHub as the context layer, Genie knows which tables to trust, which metrics are current, and how the organization actually uses its data. Agents stop guessing. Answers become ones analysts can stand behind.

In this talk, Frank Showalter, VP of Data Engineering and Governance at FIS, walks through the architecture and what it took to go from metadata chaos to demoing a net-new analytics product to clients, one that doesn't replace existing analyst workflows but opens an entirely new line of business.