The promise of AI with Microsoft Copilot is transforming analytics, but for developers using Power BI Embedded, the reality is complex. While the potential is huge, the actual features you can implement are strictly defined by your embedding architecture—a critical distinction often lost in the hype. This guide cuts through the noise to provide the definitive, no-nonsense analysis for developers and ISVs. We’ll break down exactly what’s possible, what isn’t, and why the ‘User Owns Data’ vs. ‘App Owns Data’ decision is the single most important factor for your Copilot strategy. GigXP.com | The Definitive Guide to Power BI Embedded Copilot

The Definitive Guide to Copilot in Power BI Embedded

What ISVs & Developers Actually Get. An In-Depth Analysis.

Published: August 5, 2025

The integration of generative AI like Microsoft's Copilot into Power BI Embedded is a game-changer, but its features are strictly defined by your embedding architecture. This guide cuts through the hype to give you the ground truth.

The key takeaway: Copilot's availability hinges on one question: are you using `User Owns Data` or `App Owns Data`? The answer determines everything.

The Two Worlds of Embedding: A Tale of Two Architectures

Before we talk about AI, we must understand the two fundamental ways to embed Power BI. Your choice here is the single most important factor for Copilot support.

User Owns Data

Also known as "Embed for your organization."

Who: For internal apps (e.g., company portals).
Authentication: Each user signs in with their own Power BI/Entra ID credentials.
Security: Power BI knows who the user is and applies security rules (like RLS) directly.
Copilot Support: Limited Support.

App Owns Data

Also known as "Embed for your customers."

Who: For ISVs and SaaS products selling to external customers.
Authentication: Your application authenticates using a single identity (a Service Principal).
Security: Your app is responsible for managing users and security. Power BI doesn't know the end-user.
Copilot Support: Not Supported.

This identity difference is the technical reason for the support gap. Copilot is built on a foundation of user-level security and context, which is absent in the `App Owns Data` model.

The Full Spectrum of Copilot Experiences

To understand what's missing from the embedded experience, it's helpful to see the full range of Copilot features available in the native Power BI Service. This context is key to managing expectations.

The Interactive Copilot Pane

The iconic "chat with your report" feature. A side panel where users ask natural language questions about the report's data. (Not available in embedded).

AI-Powered Authoring

For developers and analysts, Copilot can create entire report pages from a prompt, or write and explain complex DAX measures. (Not available in embedded).

The Narrative Visual

An author-configured visual that provides a text summary of a report page. (This is the ONLY feature available in `User Owns Data` embedded scenarios).

The Copilot Feature You Can (And Can't) Use

In the supported `User Owns Data` model, you don't get the full, interactive Copilot experience. Instead, you get one specific feature: the Narrative with Copilot visual.

What You GET: The Narrative Visual

An author-configured visual that provides an AI-generated text summary of the data on a report page. The user interaction is passive—they read the summary, they don't chat with it.

What You DON'T Get: Interactive Chat

The dynamic, conversational Copilot pane that appears in the Power BI Service is not available in any embedded scenario. Users cannot ask ad-hoc questions.

How the Narrative Visual Works

A report author adds the visual to a report. They can guide the AI with a prompt like, "Summarize sales trends in bullet points." This prompt is saved with the report. When an end-user views the embedded report, Copilot generates the summary based on that saved prompt and the data on the page. The end-user cannot change the prompt.

// Conceptual Embed Config (JavaScript)


let embedConfig = {
    type: 'report',
    id: reportId,
    embedUrl: embedUrl,
    accessToken: accessToken,
    tokenType: models.TokenType.Aad,
    settings: {
        // This is the key part for enabling the visual
        // in a programmatic embed scenario.
        copilot: {
            enabled: true 
        }
    }
};

The 'Why': Technical Hurdles for ISV & SaaS Scenarios

The lack of support for `App Owns Data` isn't an oversight; it's a fundamental architectural challenge. Copilot's security model is in direct conflict with how this embedding method works.

The Identity Conflict

Copilot Needs to Know the User

Copilot's security requires knowing who is asking the question to apply permissions and Row-Level Security correctly.

'App Owns Data' Hides the User

This model uses a single Service Principal. Power BI has no idea which of your thousands of customers is viewing the report.

Result: A fundamental conflict. Without a known user context, Copilot cannot run securely and is therefore disabled.

The Price of Admission: Prerequisites & Costs

Enabling even the limited Narrative visual has specific requirements. The primary path forward is **Microsoft Fabric capacity**, as Power BI Premium P SKUs are being retired for new commercial customers.

Interactive Chart: Capacity & Copilot Support

Use the filter below to see which capacity SKUs support Copilot. The cost/power is a relative illustration.

The Definitive Prerequisite Checklist

To provide a clear, actionable summary, the following table consolidates all necessary configurations for the `User Owns Data` scenario, reflecting the latest requirements.

Category	Requirement	Details & Notes
Embedding Model	User Owns Data	Must use the "embed for your organization" model.
Capacity Type	F2+ or P1+	Workspace must be on a paid Fabric F2+ or Power BI Premium P1+ capacity. Note: Power BI Premium P SKUs are being retired for new commercial customers in favor of Fabric F SKUs. P SKUs are now a legacy option for existing agreements or specific government cloud tenants.
Excluded Capacities	A, EM, PPU, Trial	Not supported on Embedded A/EM SKUs, PPU, or Trial capacities.
End-User License	Pro or PPU	Each viewer needs a Pro/PPU license (unless a P1+/F64+ capacity is used).
Tenant Admin Setting	Enable Copilot	The master "Copilot...powered by Azure OpenAI" switch must be on.
Data Residency	Allow Cross-Geo	If tenant is outside US/France, must allow cross-region data processing.

Decision Tree Infographic: Is Copilot For You?

Your Path to Embedded Copilot

1. Embedding Model?

Which do you use?

→

2. User Owns Data?

If yes, proceed. If no...

(✗ App Owns Data: Not Supported)

→

3. Capacity?

F2+ or P1+?

(✗ A, EM, PPU: Not Supported)

→

🎉 Success!

You can use the Narrative Visual.

Feature Showdown: Embedded vs. Power BI Service

To manage expectations, it's crucial to see the feature gap side-by-side. The embedded experience is a small subset of what's available in the native Power BI Service.

Copilot Feature Availability Matrix

Feature	Power BI Service	Embedded (User Owns Data)	Embedded (App Owns Data)
Interactive Copilot Chat Pane	Supported	Not Supported	Not Supported
View/Consume Narrative Visual	Supported	Supported	Not Supported
AI-Powered Report Creation	Supported	Not Supported	Not Supported
DAX Query Generation	Supported	N/A	N/A
Summaries in Email Subscriptions	Supported	N/A	N/A

Strategic Recommendations & Future Outlook

So, what should you do? Your strategy depends entirely on your business model.

For Internal Apps (User Owns Data)

✅ Assess the Value: With the F2+ requirement, the cost barrier is much lower. The question is now less about cost and more about the utility of the narrative visual.
✅ Manage Expectations: Be crystal clear with stakeholders. You're delivering an "automated summary visual," not a chatbot.
✅ Prioritize Data Quality: Good report design with clear labels is essential for good AI summaries.

For ISVs / SaaS (App Owns Data)

❌ Do Not Plan for Near-Term Integration: Interactive Copilot is not on the roadmap. Don't promise it to your customers.
❌ Explore Alternatives: Consider building your own "lite" chat experience using Azure OpenAI services directly.
❌ Monitor Official Channels: Keep a close eye on the official Microsoft Fabric blog for any announcements. Don't rely on forums.

Future Outlook: What's Next for Embedded AI?

Microsoft's investment in Copilot is immense, but its focus is strategically targeted on the core Power BI service. For the embedded world, especially ISVs, the path forward is less clear.

Rapid Innovation in the Core Service: Expect monthly improvements to the interactive chat and authoring experiences within PowerBI.com. This is Microsoft's priority.
The Unclear Path for `App Owns Data`: There is no public roadmap for bringing interactive Copilot to ISV scenarios. Solving the identity and security problem is a major engineering challenge that is not trivial.
The Key Strategic Question: The question for ISVs is not *if* Microsoft will address this, but *how* and *when*. Any solution will likely require a significant change to how apps handle authentication with Power BI. Plan accordingly.

Disclaimer: The Questions and Answers provided on https://gigxp.com are for general information purposes only. We make no representations or warranties of any kind, express or implied, about the completeness, accuracy, reliability, suitability or availability with respect to the website or the information, products, services, or related graphics contained on the website for any purpose.