Getting AI-ready in Power BI isn’t complex, but it requires intentional schema design, carefully created verified answers, and thoughtful AI instructions for next-generation analytics
AI Data Schema Design
1
Access Prep Data for AI
Navigate to Power BI Desktop and select the “Prep data for AI” button on the Home ribbon. In Power BI Service, select your semantic model and click “Prep data for AI” from the ribbon. This opens the unified experience for all AI preparation features.
Desktop & Service
Required: Q&A Enabled
2
Define Key Entities
Identify and classify your most important fields as entities (Customer, Product, Region) and their attributes (Type, Segment, Category). Focus on fields with the greatest analytical weight like Date, Amount, Margin, and Customer data.
Entities
Attributes
Analytics Focus
3
Use Clear Field Names
Rename cryptic field names like “DimCli” or “ID_VTA” to human-readable labels such as “Customer” and “Sales ID”. This helps Copilot understand user intent and provides more accurate responses to natural language queries.
Human-Readable
No Abbreviations
4
Add Synonyms
Assign synonyms to fields and tables to improve natural language flexibility. For example, add “item”, “reference”, or “article” as synonyms for “Product”. This allows Copilot to interpret a wider range of user expressions.
Natural Language
Multiple Terms
Pro Tip
Focus on the 20% of fields that drive 80% of business questions for better Copilot response relevance and accuracy.
Setting Up Verified Answers
1
Select Target Visual
Choose a visual representing a key business insight or frequently asked question. This visual will serve as the verified answer when users ask related questions.
Visual Selection
Business Critical
2
Access Verified Answer Setup
Click the “…” menu on your visual and choose “Set verified answer” (Desktop) or “Set up a verified answer” (Service). Ensure you have authoring permissions in a Copilot-enabled workspace.
Edit Mode Required
Copilot Workspace
3
Define Trigger Phrases
Create natural language trigger phrases users might ask, e.g. “What are our top sales regions?” or “Show me quarterly revenue trends.” Use varied phrasing to capture different user expressions.
Natural Language
Multiple Variations
4
Save to Semantic Model
Verified answers are saved to the semantic model and are available across reports using the same model for consistent AI responses.
Model-Level
Cross-Report
Important Requirements
Must be in a Copilot-enabled workspace with authoring permissions, edit mode, and Q&A enabled.
Creating AI Instructions
1
Navigate to AI Instructions
In the “Prep data for AI” dialog, select “Add AI instructions” tab to provide context, business logic, and guidance for Copilot on your business terminology.
Context Setting
Business Logic
2
Define Business Context
Provide specific business instructions, e.g. “Busy season is October to February”, or “When users mention ‘ABCD’, it refers to total invoice field.”
Business Terms
Industry Context
3
Set Analysis Rules
Guide Copilot on analysis approach, e.g. “Always analyze sales quarterly”, “Prioritize customer segmentation for retail insights”.
Analysis Rules
Data Priority
4
Apply and Test
Click “Apply” to save and test immediately via Copilot pane. Refine instructions based on responses.
Iterative Testing
Continuous Improvement
Example AI Instructions:
# Customer Identification
– `accountid` refers to customers in Revenue table
– `earningsid` refers to Customers in Partners table
– Define “top customers” by revenue table, highest order values
# Product Metrics
– Filter by State=Washington OR State=California unless specified
– For Total Active Partners, use measure `Monthly Active Partner Count_ID`
– Food products always show store from Store table via store_id
# Customer Identification
– `accountid` refers to customers in Revenue table
– `earningsid` refers to Customers in Partners table
– Define “top customers” by revenue table, highest order values
# Product Metrics
– Filter by State=Washington OR State=California unless specified
– For Total Active Partners, use measure `Monthly Active Partner Count_ID`
– Food products always show store from Store table via store_id
Understanding Copilot Features
| Feature | Analyze Report | Analyze Data |
|---|---|---|
| Primary Function | Interprets and answers questions about existing visuals within a report | Uses semantic model to respond to natural language questions and create new analysis |
| Data Source | Analyzes current report visuals and their underlying data | Accesses full semantic model data beyond current report |
| Use Cases | Explain chart trends, summarize visual insights, interpret existing analysis | Ad-hoc queries, new visualizations, custom calculations, data exploration |
| Output Type | Narrative explanations, insights about visible data | New visuals, charts, tables, DAX measures, comprehensive analysis |
| Availability | Report Copilot Pane – Read and Edit Modes | Standalone Copilot, Report Pane, Desktop and Service |
Analyze Report Best Practices
Use “Analyze report” for explaining existing insights and summarizing dashboard performance. Ideal for executive briefings and report consumption.
Analyze Data Best Practices
Leverage “Analyze data” for exploratory analysis, ad-hoc questions, and new visualizations. Ideal for investigative analytics.
User Experience Design
Design AI-ready models considering casual users needing explanations and power users requiring analysis flexibility.
Testing & Validation
1
Use Skill Picker
Test capabilities like “Answer questions about data”, “Analyze report visuals”, and “Create new pages” in the Power BI Desktop Copilot pane.
Environment Simulation
Capability Testing
2
Iterative Testing
Refresh Copilot after updates to instructions, test with sample questions, analyze responses, and refine iteratively.
Iterative
Systematic Testing
3
Use Diagnostic Tools
Use “How Copilot Arrived At This” (HCAAT) and diagnostics download for transparency into AI processing.
HCAAT Analysis
Diagnostic Downloads
4
Mark Model as AI-Ready
Mark your semantic model “Prepped for AI” in Power BI Service to unlock full Copilot capabilities and remove warnings.
Production Ready
Full Capabilities
Processing Time
Changes may take an hour to 24 hours to propagate fully; saving minor report updates can speed up readiness.
Best Practices for AI-Ready Models
Security & Compliance
Use row-level security to ensure AI responses respect user permissions. Test Copilot with various security contexts.
Performance Optimization
Optimize models for speed and efficient calculations to support large data volumes and fast AI response.
Comprehensive Coverage
Model all relevant business processes and KPIs logically to allow AI to provide meaningful insights.
Flexibility & Adaptability
Design models with generic field naming and adaptable structures to evolve with business needs.
Minimal Model Strategy
Maintain minimal semantic models, leverage reuse across reports to reduce maintenance and improve consistency.
Continuous Learning
Regularly analyze Copilot usage and refine AI instructions, verified answers, and schema based on feedback and diagnostics.
Ready for Next-Generation Analytics
Getting AI-ready in Power BI isn’t complex, but it requires intentional schema design, carefully created verified answers, and thoughtful AI instructions. Once set up, you can test Copilot confidently using “Analyze report” and “Analyze data” features, ensuring the AI understands your business needs and delivers precise, actionable insights.
Empower Your Organization with AI-Ready Analytics
By following these detailed steps, you empower your organization with next-generation analytics—fast, accurate, and designed for modern business challenges. It’s not just about using AI; it’s about preparing your data and environment to harness AI’s full potential.
Start Your AI JourneyAi in Power Bi : Create Next Level Reports in Seconds Using the Latest Update
