The Product Podcast Snowflake VP of AI on Why Enterprises Hide Behind Governance to Avoid Real AI Transformation | Baris Gultekin | E296
May 13, 2026
Baris Gultekin, VP of AI at Snowflake and former co-creator of Google Assistant, leads product for enterprise AI. He discusses why governance often masks stalled AI plans. He explains the semantic layer, running models next to data for safer governance, and why context plus scoped use cases unlock practical AI at scale.
AI Snips
Chapters
Transcript
Episode notes
AI Data Cloud Runs Models Next To Data
- Snowflake is an AI data cloud that centralizes disparate enterprise data for governance and large-scale analysis.
- It runs AI next to the data so models respect access controls and operate inside a single secure environment tied to source data.
Start Small With A Scoped Use Case
- Start small: pick one high-ROI use case and build a semantic model just for it before scaling governance.
- Scope the project to define business semantics, map where data lives, and apply access controls for that project first.
Semantic Layer Makes AI Accurate On Structured Data
- Structured enterprise data needs a semantic layer so AI can write correct SQL and return accurate answers.
- The semantic layer defines metrics (e.g., revenue), column meanings, and joins so LLMs can compute one correct answer from many tables.