Equity Snowflake’s transition from storing data to shipping with it
44 snips
Apr 8, 2026 
Sridhar Ramaswamy, CEO of Snowflake and former Google ads exec, describes a shift from chatbots to agentic AI that acts on governed enterprise data. He discusses new products like Cortex-Code and Project Snowwork. Short takes cover who benefits, governance and reliability, why many pilots fail, and how Snowflake is reorganizing to ship with data.
AI Snips
Chapters
Transcript
Episode notes
Cortex-Code Showed Nontechnical Users Could Build Apps
- Cortex-Code began as a coding agent to automate Snowflake setup tasks and reduce human error.
- Sridhar recounts making Cortex-Code available internally, then non-technical sales execs started writing applications, revealing it as an abstraction agent.
Evaluate Data Agents With Software Engineering Rigor
- Validate data agents through an evaluation process and standard software engineering checks before deployment.
- Ramaswamy recommends asking targeted questions and regression checks every time you add agentic features to ensure correctness and governance.
Autonomy Grows From Detection To Low-Risk Actions
- Autonomy is incremental: start with detection and low-risk actions, then expand as reliability increases.
- Sridhar describes SREs using Cortex-Code to speed detection, then moving to selective autonomous actions for routine problems.