PDQ solved AI agent hallucinations in production analytics by encoding Standard Operating Procedures as SKILLs—version-controlled, agent-executable contracts that define inputs, logic, validation, and guardrails. Instead of scaling multiple specialized agents with long prompts, they built one agent with a library of SKILLs deployed via Git-Ops to Snowflake Dynamic Tables and indexed through Cortex Search. This approach eliminated inconsistent answers, improved quality through mandatory validation steps, and made agent reasoning auditable like code. The architecture separates SKILL discovery (lightweight semantic search) from execution (loading complete SOPs) to preserve context windows while ensuring deterministic analytical workflows.
Table of contents
The Stakes: Why “Close Enough” Doesn’t Work in Checkout AnalyticsThe Problem: Agents Are Fluent, Analytics Is StrictWhat Is an SOP, and Why Does It Matter for AI Agents?The SKILL Paradigm: SOPs for AI AgentsWhat a SKILL Actually Looks LikeSKILLs in ProductionGet Oshri M’s stories in your inboxThe Architecture: SKILLs on Snowflake IntelligenceBuild Time: Zero-ETL DeploymentRun Time: Dual-Access StrategyEnforcement: SKILL-First Model PolicyThe Impact: One Agent, Many DomainsLessons LearnedWhat’s NextResources1 Comment
Sort: