Copilot in Power BI and Fabric: What AI Brings to Business Analytics

At Microsoft Build 2023, an HR analyst asked Power BI to show employee demographics and hiring trends in plain language. Copilot analysed the data model, assembled a report page, and turned the next question — about staff turnover — into a separate piece of analysis.

Back then, Copilot in Power BI was available only in a limited preview. In 2026 it already creates and edits reports from text prompts, helps with DAX, generates summaries and answers questions about data in natural language.

For businesses, the main shift is the way people work with BI: some tasks can now start with a simple question to the data. Let’s look at how Copilot Power BI works, what it can do and where an analyst is still needed.

What Copilot in Power BI and Microsoft Fabric is

Power BI Copilot is a set of generative AI features for working with reports and data in natural language. It takes into account the tables, fields, measures, relationships and metadata of the semantic model, so the quality of the result depends directly on how well that model is prepared.

In Power BI, Copilot’s capabilities can broadly be split into two scenarios:

Copilot in Microsoft Fabric goes wider than Power BI and covers Data Engineering, Data Science, Data Warehouse, SQL Database and other Fabric components. Among other things, you can generate SQL in natural language and, in the code environment, create and explain code.

For companies that are only now building out Power BI for business, one thing is worth keeping in mind: Copilot will not fix chaotic sources, broken relationships or ambiguous KPIs. It works on top of an analytics system that already exists.

What tasks Copilot solves

What Copilot in Power BI can do is clearest in specific tasks: working with DAX, analysing reports and creating visualisations.

Generating DAX formulas and measures

DAX remains one of the most technical parts of Power BI: even a routine year-over-year sales comparison requires handling context, the calendar and measures correctly.

Copilot lets you describe the result you need in words and returns a DAX query — for example, to calculate an annual growth rate or select a particular customer group. The result can be checked in DAX Query View.

In web modelling, Copilot can also generate DAX measures and suggest changes to the semantic model.

This speeds up routine calculations, but formulas for financial KPIs and management reporting still have to be verified. For more on calculation logic, see our article on DAX formulas in Power BI.

Summaries and report explanations in natural language

In the natural language analytics scenario, a user can ask Copilot to explain an existing report, produce a summary or answer a specific question about the data.

Copilot can analyse an entire report, a single page or particular visuals. The answer can also be cross-checked against the visuals it is based on.

For instance, a manager can ask which categories contributed most to sales over a period instead of manually comparing several visuals.

Creating visuals from a text description

Copilot for report creation lets you describe the page you need in text. The system picks the tables, fields, measures and visualisation types, and the result can then be edited manually or through Copilot.

People often search for this as Copilot for dashboards, but in Power BI Copilot actually creates report pages, which can then be used to build a dashboard.

Artificial intelligence in Power BI shortens the path from idea to first draft of a report, although the quality of the result still depends on the data model and the business logic.

Limitations and risks of AI analytics

Copilot’s weakest point is context. If the model has ambiguous measure names, undocumented business terms or a confusing table structure, AI in Power BI may misinterpret the request.

Microsoft warns that Copilot’s answers are non-deterministic: the same prompt may produce slightly different results. That is why the company recommends preparing the semantic model for AI — simplifying it, describing business terms and setting verified answers to critical questions in advance.

In practice, the main limitations look like this:

Generated DAX has to be verified before it is used. Copilot operates within the permissions of the specific user, and according to Microsoft’s documentation, organisational data is not used to train the foundation models.

There is also a technology caveat. In 2026, part of Copilot’s capabilities in Power BI is already generally available, while individual AI features in Data Factory, Data Engineering, Data Science, Data Warehouse and Real-Time Intelligence in Fabric are still in preview. So before rolling out Copilot in Microsoft Fabric, it is worth checking the status of the specific feature the company plans to use.

Is business in Ukraine ready for AI in BI

Interest in AI among Ukrainian companies can hardly be called experimental any more. Among more than 200 Ukrainian companies surveyed, 93% already use artificial intelligence. The most common areas of use include data and market analysis and forecasting.

For AI in business analytics, the main challenge is the quality of corporate data and the ability to use AI-generated conclusions in management decisions. So far, Ukrainian businesses use AI more often to automate routine work than for strategic analytics and decision support.

This is exactly where the line of readiness for generative AI in business intelligence runs. Rolling out Copilot takes more than giving employees access to a new tool. It needs a foundation it can actually work with:

For a company with a mature BI system, AI in Microsoft Fabric can be the next stage of development. If data is still collected manually from Excel files, the analytics system itself has to be put in order first.

Microsoft is also developing tools for preparing models for AI: AI data schemas, business instructions and verified answers. For natural language analytics this is critical: Copilot has to correctly understand not only the data, but the business context as well.

How IWIS implements AI tools in client analytics

We start implementing AI in business analytics with data preparation: analysing sources, resolving inconsistencies, aligning KPIs and building the analytical model.

For example, for Planeta Kino the IWIS team built a data warehouse, an integration with Google Analytics and custom processes for handling and enriching information. A separate research track worked with machine learning models for forecasting and segmentation. This kind of technology foundation can serve as the basis for further use of generative artificial intelligence.

During business analytics implementation on Power BI, AI features are connected to a data model that has already been validated. Depending on the objectives, these may include:

It makes sense to test AI features on one measurable scenario first — for example, speeding up the preparation of management reports or explaining dashboards to executives.

That way the value of artificial intelligence in Power BI can be assessed before scaling it to the whole company.