How to Use Generative AI in Power BI

Building reports in Power BI often means wrestling with Data Analysis Expressions (DAX), a powerful but notoriously tricky formula language. Microsoft is changing that by adding generative AI directly into Power BI, powered by the same technology behind models like GPT. This article will walk you through how to use these new AI features to generate complex calculations instantly and automate report summaries, turning hours of tedious work into a few simple prompts.

A Quick Look at Generative AI in Power BI

For years, creating advanced reports in Power BI felt reserved for dedicated data analysts who had mastered the steep learning curve of DAX and data modeling. Generative AI aims to flatten that curve, making sophisticated analysis more accessible to everyone.

Instead of thinking of it as a single feature, it’s a set of capabilities integrated into the Power BI ecosystem. The two most immediately useful features for most users are:

• DAX Query Suggestions: This allows you to write what you want to calculate in plain English, and Power BI translates your request into a proper DAX formula. It’s like having an expert analyst sitting next to you, ready to write code on command.
• Copilot for Power BI: While still rolling out and requiring specific licensing (a Fabric subscription), Copilot can analyze your entire report page and automatically generate a written summary of the key insights. It can also help you build reports from scratch just by describing what you want to see.

The goal is simple: speed up the reporting process and empower users who aren't DAX wizards to get more out of their data.

Before You Begin: What You'll Need

To follow along with these examples, you’ll need a few things set up. Don’t worry, the setup is straightforward.

• Power BI Desktop: Make sure you’re using a recent version of the Power BI Desktop application, as these features are continuously updated.
• AI Features Enabled: Go to File > Options and settings > Options. Under the "Global" section, navigate to "Preview features." Ensure that "Quick measure suggestions" is checked. You may need to restart Power BI Desktop after enabling it.
• A Sample Dataset: An AI model is only as good as the data it has to work with. For our examples, we'll imagine a simple sales dataset with columns like 'Sales Amount', 'Order Date', 'Region', and 'Product Category'. Having clearly named columns is crucial for the AI to understand your requests.
• For Copilot Features: Using the full-scale Copilot for creating entire reports and detailed narratives requires a Microsoft Fabric subscription for your organization. However, the foundational natural language features for DAX measures are available more broadly.

Generating DAX Expressions with Natural Language

This is arguably the most powerful AI addition for the everyday Power BI user. It takes the pain out of one of the tool's most difficult aspects: writing DAX.

The Challenge with DAX

Anyone who has spent time in Power BI has stared at a complex DAX formula with frustration. Functions like CALCULATE, SUMX, and FILTER have their own specific syntax and rules. A single misplaced comma or parenthesis can break your entire report. Calculating a "Year-over-Year Growth Percentage" or the "Rolling 3-Month Average Sales" can take time, research, and a lot of trial and error.

The natural language feature removes this barrier. Now, you just describe the calculation you need.

Step-by-Step: Creating a DAX Measure with AI

Let's create a measure to calculate the total sales for a specific region, "West."

1. Load Your Data: Open Power BI Desktop and load your sample sales data.
2. Open Quick Measures: Once your data is loaded, navigate to the Modeling tab in the ribbon at the top of the screen. Click on the Quick measure button.
3. Select 'Suggestions': A new window will appear. At the bottom, you'll see a toggle for "Suggestions." Make sure it's turned on. You should now see a text box where you can type your request.
4. Write Your Prompt: In the text box, describe the calculation you want. Let's keep it simple:

sum of sales amount for the west region

1. Generate and Review: Click the Generate button. In a few seconds, the AI will analyze your prompt, look at your data model, and generate a DAX formula. It will look something like this:

SUM('Sales'[Sales Amount]) FILTER('Sales', 'Sales'[Region] = "West")

1. Add to Your Report: If the formula looks correct, just click the Add button. Power BI will create a new measure in your data panel with that formula. You can now drag this measure onto a card visual or into a chart just like any other field.

You’ve just created a filtered measure without writing a single line of DAX code. The real power here is that it also serves as a learning tool. You can see how Power BI constructs the formula, helping you understand DAX logic over time.

Tips for Writing Effective DAX Prompts

While the AI is smart, you can help it deliver better results by being clear in your requests. It’s not so much about “prompt engineering” as it is about plain communication.

• Use Column Names: Refer to your columns and tables precisely as they're named in your data model. If your sales column is named [Sales Amount], use that in your prompt.
• Be Specific About Calculations: Use explicit terms like "sum," "average," "count," "max," or "min." A prompt like "show deals" is vague, but "count of deals" is clear.
• State Your Filters Clearly: If you need to filter the data, be direct. For example: "average sales amount for Canada where product category is 'Bikes'" is much better than "Canada bike sales."

More Prompt Examples:

• "percentage change in sales amount from last year"
• "running total of revenue over order date"
• "average profit per order for the last 90 days"

Automating Report Summaries with Copilot

Once your visuals are built, the next step is explaining what they mean. Writing executive summaries or highlighting key takeaways takes time and can feel repetitive. This is where the Copilot-powered "Narrative" visual comes in.

From Data to Story

Imagine your report page has a bar chart showing sales by country and a line chart displaying sales trends over the past year. Your task is to write a short paragraph summarizing what these charts show for a leadership presentation. This is the exact task generative AI can automate.

Step-by-Step: Adding an AI-Powered Narrative

1. Arrange Your Report Page: Create a report page with at least two or three related visuals. For instance, a bar chart of Sales by Region and a line chart of Sales over time. The narrative AI will analyze all the visuals on the current page to create its summary.
2. Add the Narrative Visual: In the Visualizations pane, find the icon for the Smart Narrative or Narrative with Copilot visual (the exact name and icon have evolved, but it typically looks like a document with a spark icon). Click it to add it to your report canvas.
3. Generate the Summary: When you add the visual, Copilot will automatically analyze the report page and generate a text summary. It can dynamically pull values from your visuals, such as: “Total Sales Amount was $12.5M. The West region was the highest performing, contributing 34% of total sales. Sales showed a strong upward trend during Q4, increasing by 15% from the previous quarter.”
4. Customize the Summary: The initial summary is a great starting point, but you can tailor it further. Within the new Narrative visual, an interface will appear that lets you guide the AI. You can pre-select from topics like "key takeaways" or "outliers," or you can ask a custom question in a prompt box. For instance, you could type:

Highlight the areas where sales are declining and suggest possible reasons.

Copilot will re-analyze the data and regenerate the narrative to be more focused on your specific question.

Editing and Customizing the Narrative

The text generated by the AI is fully editable. You can click into the text box and rephrase sentences, remove points, or add your own conclusions. Think of the AI's output as a first draft written by a very fast but very literal-minded junior analyst. Your job is to add the crucial business context and human insight that only you can provide.

Beyond the Basics: Where Do These Features Fit In?

Generative AI in Power BI isn’t about replacing analysts, it's about making them - and everyone else - more efficient.

For someone new to Power BI, it works as a fantastic co-pilot for learning. By seeing the DAX created from a natural language prompt, you can begin to understand the underlying logic and patterns of the language.

For seasoned professionals, it's a massive time-saver. Instead of spending ten minutes hammering out a complex date calculation, you can generate it in seconds and move on to the more strategic part of your work: interpreting the data and making recommendations. It automates the mundane, freeing up mental energy for the critical thinking that drives business decisions.

Final Thoughts

Power BI's generative AI tools, especially the natural language to DAX feature and Copilot narratives, significantly lower the tool's learning curve. These features help automate routine reporting tasks, making data analysis faster and far more accessible for everyone on your team, regardless of their technical background.

While an AI assistant inside Power BI is a great way to handle tedious work like writing DAX formulas, getting to that point still requires you to set up data models, configure permissions, and manage a workspace. At Graphed, we've focused on using natural language AI to eliminate all that upfront complexity from the start. We designed Graphed to connect to your sales and marketing data sources in seconds so you can build real-time dashboards and reports simply by describing what you want - no modeling, no complex BI tools to learn, just plain English prompts that generate the live reports you need instantly.