Since the integration of Copilot into Power BI, the way users interact with data in analytics environments has undergone a major transformation.
With the help of artificial intelligence, it's now possible to create reports, generate DAX measures, and build entire visualizations simply by typing natural language prompts. However, for Copilot to function effectively, it’s not enough to just load data into a report — your semantic data model must be properly prepared and AI-ready.
An optimized semantic model not only increases the accuracy of Copilot's responses, but also unlocks advanced features like AI-generated schema, validated answers, and context-aware AI instructions, which help the system interpret business terminology and user intent more accurately.
In this article, we’ll walk you through how to prepare your data for AI in Power BI, why it’s critical for modern analytics, and the business benefits it delivers — especially in environments where speed, reliability, and self-service capabilities are essential.
Plus, you'll get access to a free downloadable guide for Copilot in Power BI with all the steps you need to adapt your Power BI environment for seamless Copilot integration.
Source: Microsoft Learn
Copilot has introduced a new way to interact with data in Power BI: users can now ask questions in natural language, request visualizations, generate DAX measures or summarize a report without writing a single line of code.
At the same time, Copilot’s capabilities have expanded. With tools like Copilot Agent Builder and Copilot Studio, Power BI users can now build their own custom AI agents tailored to specific business needs.
But for these AI-powered features to deliver accurate, meaningful results, your data model must be structured properly and optimized for AI.
An AI-optimized semantic model is not just a technical enhancement — it’s a critical foundation that enables Copilot to understand your business context and generate relevant, consistent, and secure answers. Key aspects include:
Without these elements, Copilot may produce incomplete, inaccurate, or misleading outputs — especially in response to open-ended or abstract questions.
In contrast, a well-prepared model allows Copilot to deliver:
An AI-ready model doesn’t just benefit Copilot — it also enhances the self-service experience for end users. Non-technical users can interact confidently with data, unlocking greater autonomy, faster insights, and wider adoption of AI tools across the organization.
According to Microsoft's official documentation, having a well-structured model is essential for Copilot to be able to generate useful and accurate answers, and to access functionalities such as the AI data schema, checked answers or AI instructions.
As of April 28, 2025, Fabric Copilot Capacity is now available across all Microsoft Fabric SKUs, including the entry-level F2 tier. This means that, for just over €50/month, organizations can activate Copilot across all Fabric tools, including Power BI — making it more accessible than ever, even for smaller teams or budget-conscious environments.
This important update was announced on the official Power BI blog as part of the May 2025 feature release.
💡 Want to dive deeper into using Copilot in Power BI?
We have prepared a practical guide where you will find everything you need to start working with Copilot step by step: from technical requirements and activation in the environment, to creating reports, visualizations, DAX measures and AI-generated narratives.
The AI data schema is a foundational component that allows Copilot in Power BI to accurately interpret the semantic model behind your Power BI report.
By defining this schema, Copilot can better understand natural language queries and intelligently select the most relevant fields, tables, and measures to generate accurate answers.
The good news? Setting up an AI schema doesn’t require advanced technical knowledge — just some thoughtful planning. Below, we walk you through the essential steps to build a schema that enhances Copilot's performance:
The AI schema lets you highlight the key columns that carry the greatest analytical weight in your model — such as Date, Category, Amount, Margin, Customer, or Region.
📌 Tip: If your tables contain many rarely used or secondary fields, leave them out. This helps reduce noise in Copilot’s responses and improves relevance.
Power BI allows you to classify fields as:
This distinction helps Copilot create higher-quality Power BI visuals and more meaningful analysis when generating content automatically.
Field names are among the first cues Copilot analyzes. Avoid technical jargon or abbreviations like DimCli or ID_VTA, which may confuse the AI.
Instead, rename them to more intuitive labels like:
CustomerSales IDThis simple change can greatly improve Copilot's ability to understand user intent.
A powerful feature of the AI schema is the ability to assign synonyms to fields and tables.
For example:
Product, you can add terms like item, reference, or article.Once your schema is defined — including fields, attributes, and synonyms — you can save it directly in Power BI Desktop.
When the report is published to the Power BI Service, the schema is automatically transferred and becomes available to Copilot for Power BI.
Creating an AI data schema is the first step toward building an AI-ready data model in Power BI. When the schema is properly configured, Copilot can:
By investing a bit of time in the schema setup, you’ll significantly enhance the quality, consistency, and business value of your Power BI reports powered by Copilot.
Source: Microsoft Learn
Verified answers are one of the most valuable features available when working with Copilot in Power BI, especially in business environments where terms and metrics may vary in meaning depending on context.
This functionality allows data modelers and analysts to define official, pre-approved answers within the semantic model, ensuring that Copilot returns accurate and consistent results when users ask questions in natural language.
Enabling verified answers brings multiple benefits:
You can assign verified answers to several types of elements in your Power BI model:
Customer → “A person or entity with at least one purchase in the last 12 months.”Profit Margin → “The difference between revenue and operating costs, excluding taxes.”Follow these steps in Power BI Desktop to define verified answers for Copilot:
Open the Model Properties Panel
Navigate to your semantic model in Power BI Desktop.
Select the Target Field or Measure
Choose the specific element (column, measure, table) to which you want to assign a verified definition.
Write a Clear, Natural Language Description
Use simple, business-friendly language. Avoid unnecessary technical jargon so that any user can easily understand the meaning.
Mark the Description as a "Verified Answer"
In the Advanced Properties, flag the description as official. This tells Copilot to prioritize this definition when responding to related queries.
Publish the Model to a Premium or Microsoft Fabric Workspace
Verified answers only function in environments that support AI-ready models, such as Fabric or Premium capacities.
When working with Copilot in Power BI, it's not enough to have a well-structured model or clearly defined DAX measures. To get the most accurate and relevant insights, it’s also essential to provide contextual guidance — especially when dealing with organization-specific terminology, internal processes, or non-standard metrics.
That’s where AI instructions —also known as AI prompts— come into play.
What Are AI Instructions in Power BI?
AI instructions are short text statements linked to your semantic model that provide Copilot with additional context about how to interpret and analyze your data.
Unlike field or measure descriptions, which are tied to specific elements, AI instructions are global and contextual. They help guide Copilot’s reasoning by clarifying priorities, definitions, and usage patterns across the model.
They are used, for example, to:
"In this model, the 'Customer Status' field distinguishes between active and inactive based on whether they have made at least one purchase in the last 12 months. Margin and profit measures should only be applied to active customers."
These types of prompts help Copilot avoid misunderstandings and provide answers that are more aligned with each organization's specific business logic.
You can create and manage AI prompts directly in Power BI Desktop:
Open the Model Properties Panel
Launch Power BI Desktop and open your semantic model.
Access the AI Schema Settings
Select the AI data schema icon or access it from the model pane.
Add a New AI Instruction
Write a custom prompt based on your business context — or use one of the templates provided by Power BI.
Review and Save
Ensure your instruction is clear, concise, and written in natural business language. Then save your changes.
Publish and Validate in the Power BI Service
Once your report is published, check in the Power BI Service that the AI instruction is active within the dataset properties.
Alongside the AI data schema and verified answers, AI prompts are a core component of building an AI-ready semantic model in Power BI.
Used correctly, they allow Copilot to:
Ultimately, AI instructions bridge the gap between technical data modeling and real-world user intent, enabling smarter, more meaningful interactions with your data.
According to Microsoft’s official documentation, the following table summarizes the most important criteria to consider when defining your semantic model in Power BI. These criteria ensure that Copilot can generate Power BI reports effectively.
| Element | Consideration | Description | Example |
|---|---|---|---|
| Table Linking | Define Clear Relationships | Ensure that all relationships between tables are clearly defined and logical, indicating which are one-to-many, many-to-one, or many-to-many. | Sales table connected to Date table by DateID field. |
| Measures | Standardized Calculation Logic | Measures should have standardized, clear calculation logic that is easy to explain and understand. | Total Sales calculated as the sum of SaleAmount from the Sales table. |
| Measures | Naming Conventions | Names for measures should clearly reflect their calculation and purpose. | Use Average_Customer_Rating instead of AvgRating. |
| Measures | Predefined Measures | Include a set of predefined measures that users are most likely to request in reports. | Year_To_Date_Sales, Month_Over_Month_Growth, etc. |
| Fact Tables | Clear Delineation | Clearly delineate fact tables, which hold the measurable, quantitative data for analysis. | Transactions, Sales, Visits. |
| Dimension Tables | Supportive Descriptive Data | Create dimension tables that contain the descriptive attributes related to the quantitative measures in fact tables. | Product_Details, Customer_Information. |
| Hierarchies | Logical Groupings | Establish clear hierarchies within the data, especially for dimension tables that could be used to drill down in reports. | A Time hierarchy that breaks down from Year to Quarter to Month to Day. |
| Column Names | Unambiguous Labels | Column names should be unambiguous and self-explanatory, avoiding the use of IDs or codes that require further lookup without context. | Use Product_Name instead of ProdID. |
| Column Data Types | Correct and Consistent | Apply correct and consistent data types for columns across all tables to ensure that measures calculate correctly and to enable proper sorting and filtering. | Ensure numeric columns used in calculations aren't set as text data types. |
| Relationship Types | Clearly Specified | To ensure accurate report generation, clearly specify the nature of relationships (active or inactive) and their cardinality. | Mark whether a relationship is One-to-One, One-to-Many, or Many-to-Many. |
| Data Consistency | Standardized Values | Maintain standardized values within columns to ensure consistency in filters and reporting. | If you have a Status column, consistently use Open, Closed, Pending, etc. |
| Key Performance Indicators (KPIs) | Predefined and Relevant | Establish a set of KPIs that are relevant to the business context and are commonly used in reports. | Return on Investment (ROI), Customer Acquisition Cost (CAC), Lifetime Value (LTV). |
| Refresh Schedules | Transparent and Scheduled | Clearly communicate the refresh schedules of the data to ensure users understand the timeliness of the data they're analyzing. | Indicate if the data is real-time, daily, weekly, etc. |
| Security | Role-Level Definitions | Define security roles for different levels of data access if there are sensitive elements that not all users should see. | Sales team members can see sales data but not HR data. |
| Metadata | Documentation of Structure | Document the structure of the data model, including tables, columns, relationships, and measures, for reference. | A data dictionary or model diagram provided as a reference. |
Once you’ve defined your AI data schema, added AI instructions, and configured verified answers, the final step is to mark your model as AI-ready in Power BI.
While this step is optional, it is strongly recommended to unlock the full potential of Copilot in Power BI.
When a model is marked as AI-ready, Power BI recognizes it as optimized for artificial intelligence features. This designation enables advanced capabilities such as:
Before you can mark a model as AI-ready, you need to meet the following conditions:
Open the Report in Power BI Desktop
Load the semantic model or dataset you wish to optimize.
Access the AI-Ready Setting
Navigate to Model > AI-Ready and select the checkbox:
“Mark this model as AI-Ready.”
Save and Publish to a Supported Workspace
Deploy the file to a Premium or Fabric workspace.
Verify in the Power BI Service
Go to the Power BI Service portal and check the dataset properties to confirm that it’s recognized as AI-ready.
Marking your model as AI-ready helps Copilot:
If you're serious about scaling AI-driven analytics and maximizing the value of Copilot in Power BI, this step is a must.
Ready to harness the power of Copilot in Power BI?
Our step-by-step practical guide for Copilot in Power BI includes everything you need to get started with confidence:
🎯 Whether you're a business user or a data professional, this guide will help you unlock Copilot’s full potential in just a few steps.
Designating your Power BI model as AI-ready brings immediate advantages:
Source: Microsoft Learn
Having an AI-ready model in Power BI not only enhances how users interact with Copilot, but also increases the overall impact of data analytics across the organization.
Below are some of the most relevant benefits of working with an AI-optimized semantic model:
Copilot is able to interpret colloquially written queries , but its accuracy depends largely on how the model is built. By having an AI data schema, proven answers and clear instructions, ambiguity is reduced and the quality of the answers is improved.
A well-structured model allows users to generate visualizations, reports or calculations without having to write complex formulas or navigate through menus. This speeds up data analysis, especially for business profiles that are not experts in DAX or model design.
Thanks to the clarity of the model and the interaction with Copilot, users can work with the data in a self-sufficient way, without constantly depending on the BI team or the technical department. This facilitates access to information and fosters a data-driven, self-service BI culture.
Copilot not only answers questions, but can also generate the most appropriate Power BI visuals for each data set and narrative explanations directly within the report.
With an AI-ready model, these visualizations and text are more relevant and aligned with business objectives.
Microsoft is optimizing and extending the artificial intelligence capabilities in Power BI and Microsoft Fabric.
Having a model marked as AI-ready ensures that your environment will be prepared for future updates, such as new Copilot agents, translational flows, task automation or integration with other natural language-based solutions.
In short, preparing your data model for AI is not just another technical requirement, but an opportunity to improve the accessibility, efficiency and accuracy of data analysis across your organization.
The integration of Copilot into Power BI marks a turning point in how users interact with data. Thanks to the power of artificial intelligence, tasks that once required technical expertise —such as building visualizations, writing DAX measures, or interpreting results— can now be accomplished simply by asking a question in natural language.
However, for Copilot to generate accurate, reliable, and business-relevant insights, having a well-prepared data model is essential. Throughout this article, we’ve explored:
But preparing the model is just the beginning.
To fully leverage everything Copilot has to offer, it’s important to understand how it works, what experiences it supports, and how to use its capabilities strategically across your organization.
At Bismart, we’ve created a complete, updated guide to help you make the most of Copilot — step by step:
✔ How to correctly enable Copilot in your environment
✔ Technical requirements, licenses, and setup instructions
✔ A detailed overview of available (and unavailable) features
✔ How to generate reports, query data, and create DAX with natural language
✔ Best practices to maximize performance and value
Take the first step toward a smarter, more efficient way of working with your data in Power BI. Download the practical guide to Copilot in Power BI below.