Originally published Nov 19, 2026.
How Can Microsoft 365 Copilot Chat Use Power BI Data?
Key takeaway: Power BI no longer has to be the user interface for Power BI data. Microsoft 365 Copilot Chat can use Power BI semantic models through Fabric IQ, letting business users ask questions in the tools and experiences they already use.
Power BI’s Next Frontier
For years, the assumption behind business intelligence has been simple:
If you want an answer from Power BI, you go to Power BI.
Business users open a report. Find the right page, apply filters, inspect visuals, and answer a question.
Generative BI changes that assumption in a profound way.
Microsoft 365 Copilot Chat can now answer questions using Power BI reports and semantic models through Fabric IQ.
That means business users can stay in an experience they already use for files, email, meetings, and conversations - and ask questions about trusted Power BI data without opening a Power BI report, dashboard, or scorecard.
That raises an important architectural question:
Can Power BI become the analytical foundation without being the user interface?
Increasingly, the answer is yes.
Can Microsoft 365 Copilot Chat Query Power BI Data?
The short answer is, “Yes!”
Microsoft documents Microsoft 365 Copilot Chat answering user questions with Power BI data as a generally available Fabric feature.
When a user asks a data question, Copilot can search Power BI content the user already has permission to access, identify relevant reports, and use the underlying semantic model to answer.
Conceptually, I think about the path like this:
Business user → Microsoft 365 Copilot Chat → Fabric IQ → Power BI semantic model
This is critical for understanding Power BI’s next frontier - Fabric IQ isn’t another user interface business users must navigate.
Instead, Fabric IQ is conduit that connects Copilot Chat to trusted Power BI data, measures, visuals, etc.
In the end, the user will mostly likely never know which Power BI semantic model ultimately answered the question.
And that’s the whole point.
Does the User Have to Know Which Power BI Report to Use?
No, this isn’t required. A user can simply ask a business question. For example:
“What was gross margin in the Northeast last quarter?”
In response, Copilot can search the Power BI assets the user has permission to access and select related content based on the question. This completely changes the normal Power BI end-user workflow.
Traditionally, the user must know:
Which report contains the answer.
Which page to open.
Which visual to use.
Which filters to apply.
With Copilot Chat, the business question comes first, and the reporting architecture moves into the background.
Microsoft also supports more explicit ways to guide Copilot when needed. A user can:
Name a Power BI report in the prompt.
Paste a Power BI report URL.
Attach a recently viewed Power BI report.
Paste a semantic model URL to ask questions directly against that model.
That last option is especially interesting.
When a semantic model URL is supplied directly, Copilot can answer from the semantic model without using report visuals as grounding.
What Actually Provides the Answer: The Report or the Semantic Model?
While both Power BI reports and semantic models can provide Copilot context, the semantic model remains critical.
Microsoft says Copilot can identify a relevant Power BI report and use its underlying semantic model to answer data questions. This is a subtle but important detail.
The report can help Copilot discover the relevant business context and reporting asset.
The semantic model provides the analytical foundation underneath it:
Measures
Dimensions
Relationships
Calculations
Security
Business definitions.
This is why Generative BI does not make semantic modeling less important - it makes it more important than ever for Power BI teams.
A business user may no longer see your model, your DAX, or even the Power BI interface.
But the quality of those semantics still shapes the quality of Copilot’s answer.
What Happens When Copilot Finds the Wrong Report?
This is where the experience isn't magic yet. Microsoft explicitly says that searching for Power BI content is still improving.
If Copilot doesn’t find the report you want the user to access, the user can make Copilot’s grounding more explicit by naming the report, attaching it, or pasting its URL.
This becomes particularly important in organizations with many overlapping reports. Imagine an organization has reports for:
E-commerce Sales
Distributor Sales
Regional Sales Performance
Product Profitability
All four may contain concepts such as Sales, Product, Geography, Margin, and Date.
Now, consider the scenario where a user simply asks:
“Which products had the worst margin last quarter?”
In response to this prompt, which report should Copilot choose? That’s not just a search problem. It’s a semantic architecture problem.
Report names, model boundaries, descriptions, terminology, and the consistency of business concepts across Power BI content all become part of the end user’s Generative BI experience.
For Power BI teams serious about Generative BI, inventorying, assessing, and rationalizing their library of semantic models is the first, and most critical, step in their journey.
Does Prep Data for AI Still Matter?
Yes - but I would be very precise about what we know at the time of this writing.
Microsoft provides guidance for using Fabric IQ question answering with semantic models. The guidance emphasizes the measures, relationships, and business definitions in the model.
It also recommends using Prep data for AI when organization-specific terminology, analytical defaults, or ambiguity require more context. This reinforces a principle that is central to real-world Generative BI with Power BI:
A technically correct semantic model isn’t automatically an AI-ready semantic model.
Power BI currently provides AI data schemas, AI instructions, and Verified Answers to help prepare semantic models for natural-language experiences via Copilot.
However, don’t assume that every Prep data for AI feature behaves the same in every Copilot experience - unless Microsoft explicitly documents it.
In other words, different Copilot experiences use different grounding paths.
The larger point in all of this: if Microsoft 365 Copilot is going to use your semantic model to answer business questions, the meaning expressed by that model matters.
A lot.
What Security Does Copilot Chat Respect?
Your business users don’t gain access to Power BI data merely because Copilot can search for it.
Microsoft documents that Copilot uses the user's existing Power BI permissions and respects row-level security and object-level security on the underlying semantic model. That’s critical for enterprise adoption:
The experience changes.
The security model doesn’t disappear.
Microsoft 365 Copilot also honors sensitivity labels associated with the Power BI content used in the conversation, and Microsoft Purview data loss prevention policies can affect whether that content may be processed.
So the architectural goal is not to bypass Power BI controls.
It’s to make trusted Power BI data accessible through an AI-powered, natural language experience.
Can Copilot Combine Power BI Data With Microsoft 365 Context?
This may be one of the most important aspects of the Generative BI architecture - the answer from Power BI doesn’t have to live in isolation.
Microsoft documents that Copilot can reconcile Power BI data answers with broader Microsoft 365 context, including files, chats, and emails. Imagine a business user asking:
“Why did revenue decline for this customer?”
Power BI might provide the revenue trend. Microsoft 365 may contain customer correspondence, meeting notes, or other contextual information.
Bringing those two worlds together is much closer to how business decisions actually happen:
The analytical data tells you what happened.
The surrounding business context may help explain what was happening around it.
That doesn’t mean Copilot has proved causation.
But it creates a much richer decision-support experience than simply opening another dashboard.
Does a Power BI App Become the Copilot Chat Scope?
In short - no.
Microsoft currently supports individual Power BI reports in Microsoft 365 Copilot Chat, including reports distributed through workspace apps. Reports in organizational apps can also appear in the attachment experience, although some linking behavior differs.
But top-level Power BI apps themselves are not currently supported as the grounding object. So this is different from App-scoped Copilot inside Power BI.
App-scoped Copilot uses the curated boundary of a Power BI workspace app.
Microsoft 365 Copilot Chat can use Power BI reports and semantic models, but the Power BI app itself doesn’t become the conversational boundary.
That means these are two different architectural experiences.
Can Microsoft 365 Copilot Chat Use a Fabric IQ Ontology Directly?
Not at the time of this writing.
Microsoft currently says Fabric data agents and ontologies can’t answer questions directly in Copilot Chat unless they are exposed through an explicitly published Microsoft 365 agent. So the paths differ.
For supported Power BI content:
User → Microsoft 365 Copilot Chat → Fabric IQ → Power BI semantic model
For an ontology-grounded experience:
User → Microsoft 365 Copilot / Teams → Published agent → Fabric IQ ontology
That’s an important difference.
It also helps explain why agents become increasingly important as Generative BI expands beyond analytical questions contained within Power BI semantic models.
What Does This Mean for Power BI Teams?
I think it changes how we should think about Power BI's role. Historically, Power BI has been both:
The analytics layer.
The user interface.
With Generative BI, those responsibilities can now be separated where it makes sense for your business users.
Power BI semantic models can remain the trusted analytical foundation while Microsoft 365 Copilot Chat becomes the self-service conversational interface.
The next frontier of Power BI is exciting, and generates a different set of design questions.
It’s no longer a matter of simply asking if a Power BI report is good in the traditional sense. Power BI teams also need to ask:
Can AI reliably understand the semantic model behind it?
Can Copilot find the right reporting content?
Are overlapping business concepts defined consistently across models?
Can business users ask the questions they actually care about without knowing our BI architecture?
Those are Generative BI questions that cutting-edge Power BI teams need to tackle ASAP.
Power BI May Become More Important by Becoming Less Visible
I don’t think Microsoft 365 Copilot Chat reduces Power BI's importance. I think it changes where Power BI sits in the experience.
Increasingly, business users will spend less time navigating reports and dashboards.
But the semantic models behind those reports become even more important because they provide trusted analytical meaning to AI. That’s a significant shift.
The future of Power BI may not require every business user to become a better Power BI user.
Instead, Power BI teams may increasingly build the semantic foundation that lets business users ask natural-language questions from the tools where they already work, right when they need the answers.
And that may be one of the most important changes Generative BI brings to business intelligence.
Microsoft Documentation Referenced
Greetings! I’m Dave Langer
Author | Microsoft MVP | LinkedIn Top Voice
I help Power BI teams move past dashboards to Generative BI using Microsoft Fabric and Fabric IQ.
If your organization is serious about enterprise-class conversational analytics, we should talk.