Beyond SQL Generation: How to Teach Agents What Your Database Means
summary
Kris Jenkins (Snowflake) examines why code-writing agents produce syntactically correct SQL but still give wrong answers, arguing the issue is lack of domain context rather than SQL itself. The talk covers teaching agents the meaning of a database—table relationships, column semantics, business intent and tacit analyst knowledge—so they can produce accurate, not just well-formatted, results.
AI-written summary · about
chapters (15)
- 0:00Introduction and setup
- 1:05AI's impact on programming
- 3:13AI and SQL generation gap
- 5:20Folk knowledge in databases
- 8:28Encoding folk knowledge in YAML
- 10:57Open Semantic Interchange overview
- 11:59Level one: Basic schema description
- 14:47Level two: Synonyms and examples
- 18:16Predicting semantic drift in 2027
- 21:39Level three: Canonical metrics and queries
- 23:45Live demo with Cursor database
- 32:29Building a semantic view
- 34:15Validating semantic views
- 35:39Bigger picture and future predictions
- 38:50Closing and call to action
related talks
15marchive – Manohar Sai Jasti
02h 42mSQL (The Return): Demystifying Preconceptions and Using the DB Effectively
24mBuilding Your Intelligent Data Agent
16mGame Data for All: How AI Is Reinventing Game Analytics
12mEnterprise Agents Have a Structure Problem
32mBecause SQL is everywhere but you don't want to use it
40mData Strategy for Generative AI Applications — Why Your Data is the Differentiator
25mFrom SQL to Python: Building Data Context for Agents and people