Preparing Governed, AI-Ready Data & Fabric Data Agents

Instructor: Alaa Hady · Language: English or Arabic · Suggested duration: 2 days · Level: Advanced (Level 300)

Objectives

• Understand the architectural principles required to make enterprise data truly "AI-Ready" (clean semantics, rich metadata, strict security, and verified truth).
• Build, configure, and ground Fabric Data Agents (formerly AI Skills) over Lakehouses, Warehouses, KQL Databases, and Semantic Models.
• Integrate Fabric Data Agents with Microsoft Copilot Studio and Azure AI Foundry for multi-agent enterprise experiences.
• Establish evaluation frameworks to test AI accuracy, prevent hallucinations, and enforce strict RBAC/RLS security boundaries.

Outline and contents

1. Why 80% of enterprise AI projects stall: garbage-in/garbage-out, ambiguous business definitions, and unindexed flat tables
2. The AI-Ready Data Checklist in Fabric: Gold Medallion Star Schemas, business glossaries, column descriptions, and certified semantic models
3. Optimizing Semantic Models for AI grounding: Prep data for AI features (AI instructions, verified answers, simplified AI data schemas)
4. Designing and configuring Fabric Data Agents:
   • Selecting and scoping tables across Lakehouse, Warehouse, KQL, and Semantic Models
   • Writing system instructions, domain context rules, and few-shot SQL/DAX/KQL example pairs
5. Testing, debugging, and evaluating Data Agent responses (inspecting generated SQL/DAX/KQL queries for hallucinated joins or wrong filters)
6. Connecting Fabric Data Agents to Microsoft Copilot Studio to empower conversational bots in Microsoft Teams
7. Security, Privacy & Governance: how RLS/OLS and OneLake permissions pass through to AI queries, preventing prompt-injection data leaks, and Purview AI auditing
8. Hands-on Lab: Tuning an enterprise Semantic Model for AI, building a custom Fabric Data Agent with verified few-shot examples, and stress-testing its accuracy

Target audience

Analytics engineers, AI architects, senior BI developers, data stewards, and Copilot delivery teams.

Expected outcomes

Participants can transform raw corporate databases into AI-grounded semantic layers and deploy accurate, secure Fabric Data Agents that answer natural-language business questions reliably.