Zach Stanford

Your data warehouse holds the answers to your most important business questions. But without an AI-ready data architecture, the LLM can’t access or use them. We solved this problem at Block by building the Block Data MCP, combining a governed semantic layer, a Snowflake client and AI agents (Goose), to map business questions to trusted metrics, layered with generative AI insights. In this session we share how we designed and implemented the underlying architecture, eliminated AI hallucinations through deterministic data flows, and scaled our Block Data MCP to support ~2,000 internal data users and various business use cases.

At Block (Cash App), we’re breaking down technical barriers to empower any employee to engage with data — no SQL required. In this session, we’ll share our three-phase journey to democratize data insights across the company at scale: - Phase 1: Developing Streamlit dashboards that empower teams to independently make informed decisions - Phase 2: Automating our analytics with Agentic AI — from question to insight — cutting analysis time by 95% - Phase 3: Building integrated, scalable solutions that leverage Snowflake MCP, Codename Goose and Streamlit Make data human by letting it tell its own story.