Don't give your AI agent the whole warehouse. Give it 20 tables.

Teams that start with full access get ~5% reliability and a dead demo.

The production recipe from Nao Labs: pick ~20 tables and ~20 questions that must be answered with 100% accuracy. That narrow scope starts around 50% reliability - and gives you a ladder to 85%.

It mirrors what works in my GA4/BigQuery projects: we never start with "analyze everything", we start with the five queries the founder actually asks every Monday.

Scope isn't a limitation. It's the first feature.

Want a stronger data analyst role or a raise? Grab the FREE Product Analyst Playbook and get the exact roadmap to your next offer.