Imagine asking your data a question or requesting a report, and having it all happen automatically without writing any code. This is the potential of Copilot for Power BI, and Microsoft’s latest feature updates makes this experience even more powerful. New capabilities like AI data schemas, Verified Answers, and AI Instructions enhance how Copilot understands and interacts with your semantic model, delivering insights that are more accurate, consistent, and tailored to your business context.
These additions build on the foundation of having an AI ready semantic model. If you’re just beginning with Copilot for Power BI, I recommend starting with the basics in this guide: Copilot for Power BI: Best Practices for AI Ready Semantic Models. Once your model is ready, these new tools can take your analytics to the next level.
In this blog, I will explore what each of these new features offers and share how you can begin using them to enhance your own Copilot experience in Power BI.
Feature Overview
Each new feature addresses a specific aspect of the AI analytics journey and contributes to making semantic models more intelligent and impactful within Power BI.
AI Data Schemas
AI data schemas allow authors to define a focused subset of fields within the semantic model that Copilot should give priority to when interpreting and responding to queries. By narrowing the scope to the most relevant fields, Copilot can reduce ambiguity and more confidently identify the data needed to answer a question accurately.
After configuring the AI data schema, Copilot emphasizes this refined set of fields during analysis. This improves the likelihood of producing responses that are aligned with business needs, leading to a smoother, more effective experience
Verified Answers
Verified Answers are curated visual responses stored in the semantic model and triggered by specific user prompts or similar phrasing. These predefined answers ensure users receive consistent and reliable insights for common or critical business questions.
Each verified answer includes trigger phrases, an associated visual, and optional filters. When a user query matches one of these phrases, either exactly or semantically, Copilot returns the verified answer rather than generating a new one. Because they are tied to the semantic model rather than a specific report, verified answers can be reused across all reports using that model.
This feature improves both the consistency and trustworthiness of Copilot responses. It also contributes to long-term improvements by helping Copilot learn from how users interact with verified answers, enhancing its understanding of phrasing and intent over time.
AI Instructions
AI Instructions give model authors a way to embed helpful guidance into their semantic model, providing Copilot with clarity around how to interpret questions and return relevant answers. These instructions allow you to describe key business terms, define preferred analysis methods, or highlight important metrics.
By supplying Copilot with this additional context, you reduce the chance of misinterpretation and ensure that responses better reflect your organization’s way of thinking about and using data. The more tailored the guidance, the more effective Copilot becomes at delivering meaningful insights.
How to Implement These Features
Implementing these features is simple and intuitive. It all starts in Power BI Desktop. First open a semantic model file in Power BI Desktop. Then navigate to the top ribbon and click “Prep data for AI”.
To define the AI data schema, click on “Simplify the data schema”. Then deselect any fields that Copilot does not need to analyze. Once completed, click “Apply”.
To define AI instructions, click on “Add AI instructions”. Then add instructions for Copilot in the provided text box. Once completed, click “Apply”.
To define verified answers, return to the report view. Then select a visual you would like to define as a verified answer and click on the three-dot menu. Then click on “Set up a verified answer”.
Add phrases that should return the selected visual as an answer.
Once all phrases have been added, click “Apply”.
With these steps complete, your semantic model is now enhanced and optimized for Copilot-powered experiences.
Conclusion
As Copilot becomes more deeply integrated into the Power BI ecosystem, preparing your data for AI is not just a best practice. It is a necessary step to unlocking more accurate and relevant insights. AI data schemas help streamline Copilot’s understanding of your data. Verified Answers provide reliable, curated responses to recurring questions. AI Instructions offer the context Copilot needs to align its outputs with your business logic. Try these new features today and take your semantic models to the next level.
If your model is not quite ready yet, don’t worry. You can get started with my article on building AI ready semantic models: Copilot for Power BI: Best Practices for AI Ready Semantic Models.