411: Jeremy Burton, GM of Observability at Snowflake: The Data Layer Was Never a Commodity. It’s the Whole Game

Send us Fan Mail Jeremy Burton is the General Manager for Observability Business Unit at Snowflake following the acquisition of Observe, the developer-centric observability platform where he served as CEO. A veteran enterprise technology executive with a thirty-year career spanning Silicon Valley, Jeremy previously served as Executive Vice President of Marketing and Corporate Development at Dell, President of Products at EMC, and CEO of Serena Software, with earlier leadership roles at Symantec, Veritas, and Oracle. He has served on Snowflake’s board of directors since 2015 and sits on the McLaren Racing Advisory Board. In this episode, Jeremy draws on watching customers go from collecting a petabyte of data in a decade to generating one every two days to make the case that without a real data foundation, enterprises will either get the wrong answer from their AI, or burn a fortune getting to the right one. In this conversation, we discuss: Why pointing AI agents at hundreds of terabytes of telemetry burns tokens fast, and what Jeremy says has to change first. How an AI agent facing 10 sources of truth can deliver the wrong answer with total confidence. Why Jeremy calls the LLM “a new animal in the zoo,” and what that means for anyone trying to monitor AI agents. How AI agents are starting to build the knowledge graphs that once took human implementation teams to create. Why evals are becoming essential to running an LLM, and what a single model release can do to existing applications. What Jeremy would tell his younger self stepping off the plane to meet Mark Benioff, and the career superpower he undervalued. Explore the conversation: 00:00 Intro and AI Fun Fact: Pope Leo XIV's Encyclical and Anthropic's Vatican Address 04:08 Introducing Jeremy Burton, GM for Observability at Snowflake and Former CEO of Observe 07:09 The Observe Origin: Pitching Observability as a Unified Data Problem Natively on Snowflake 10:20 Observability as a Data Problem: Metrics, Logs and Traces 13:39 Snowflake Culture and Leadership Eras 16:58 Joining Snowflake: Running Observe After the Acquisition 18:57 Data Foundation vs. Frontier Models: Balancing CPU and GPU Burn with Context 22:23 Snowflake's AI Strategy: Data Foundation and Sources of Truth 28:11 Non-Human AI Telemetry: Using Knowledge Graphs to Structure Machine Data 32:31 Probabilistic AI and LLM Observability: Why Evals and Harnesses Are Essential 36:14 Future of Work: The "LLM as Modern-Day Tractor" and Software Engineering Practices 41:18 Career Superpowers: The Value of Networks and Relationships in Silicon Valley 43:06 Closing: Where to Connect with Jeremy Burton and Explore Snowflake Observability Resources: Subscribe to the AI & The Future of Work Newsletter Connect with Jeremy Burton on LinkedIn AI fun fact article On How Ashu Garg, General Partner at Foundation Capital, finds and invests in the future of AI Other episodes mentioned on the show: Automate the Work, Amplify the Humanity, with Snowflake's CPO Arnnon Geshuri - Learn how Box connects secure enterprise content with AI to power smarter workflows. Visit box.com/ai to learn more.