AI can perform a lot of interesting tasks these days. It can check your email. It can draft an email response. And it can do many actions within the Power BI environment — including creating a complete Power BI report entirely from a prompt window, without touching anything inside Power BI Desktop.
This is possible today with the combination of Report Authoring Skills, the Power BI Desktop Bridge, and AI tools like GitHub Copilot.
In this article, I walk through exactly how this works — the pieces you need, how they fit together, and a live demo of building a full report from a single prompt.
Published by Reza Rad | RADACAD
📺 Watch the full video on YouTube: No-Code Power BI Report Authoring with GitHub Copilot + Fabric Skills + Power BI Desktop Bridge
🌐 RADACAD Blog: radacad.com
The Era of Manual Report Building Is Ending (Video: 0:46)
I think the era of us doing manual report building is pretty much getting over. We are in the age of AI — you can ask AI to do the tasks for you. And one of the tasks AI is genuinely good at is writing code. A Power BI report, behind the scenes, is a code base.
I have talked about the Power BI report structure — PBIR and PBIP — in another video, which I recommend checking out. That is a file structure: within that folder you have different files, different sections, for each Power BI visual. There is a piece of code for every visual, some configuration around it — and this is exactly the kind of structured, well-defined work that AI is good at.
For the full explanation of PBIX, PBIT, PBIP, and PBIR, see my article: Power BI File Formats Explained: PBIX vs PBIT vs PBIP vs PBIR.
There are Report Authoring Skills that know how to write this code and what to use, combined with tools such as the Power BI Modeling MCP Server — which I also explained in another video — that connects to your existing Power BI semantic model, whether local or the remote MCP server connecting to the semantic model in the service. Combined together, these can actually build a Power BI report for you, entirely through prompts.
For a full walkthrough of the Power BI Modeling MCP Server, see: No Code Power BI: How to Use Claude and the Power BI Modeling MCP Server.
Let me show you exactly how this works.
The Three Elements You Need (Video: 1:16)
1. VS Code and an AI Assistant
You need a development environment. In this demo I am using Visual Studio Code (VS Code) — a free tool you can download and install. You can use other tools of your choice, but this is what I use.
Alongside VS Code, there is a chat window where you ask questions of your AI assistant. In this example, I am using GitHub Copilot. This is similar to Claude Code or Codex — you can use any of these to generate code for you. I am demonstrating with GitHub Copilot, but the exact same thing works with other tools as well.
2. Fabric Skills — Specifically Report Authoring Skills
You need Fabric Skills installed — and specifically, the skill set you need for this is the Report Authoring Skills.
To find it: search for “Microsoft Fabric Skills” — you can even search specifically for “Report Authoring Skills,” although you do not necessarily need to search that specifically, because Fabric Skills is the whole skill set including everything. There is a Microsoft Learn page that explains how this works, and a GitHub library where you can download the entire skill set — everything from Warehouse and Lakehouse skills, to Fabric IQ, to data integration and Dataflow skills, and a dedicated set of skills specifically for Power BI report authoring.
I have explained the exact install command in other videos — you run it once, and once installed, the skills are ready to use for anything.
There are actually a few different Report Authoring Skills within this set — one for designing the report, one for planning it, and one for authoring the report itself. Each skill file, when you open it, contains instructions for the AI tool on how to go and work with this and create a report. It also has information about specific visual types — what this visual type is, what that visual type is — so the AI has everything it needs to build the report correctly.
3. Power BI Desktop Bridge
The third piece is having the Power BI Desktop Bridge enabled.
If you have the latest version of Power BI Desktop — and I would encourage you to download it if you don’t already have it — you enable this from Options and Settings → Options → Preview Features. In the Preview Features section, you will find:
“Enable external tool to access Power BI Desktop through secure local APIs”
This means your AI assistant — Claude Code, GitHub Copilot, or whatever you are using — can look into Power BI Desktop, see what report is currently open, and perform actions on it. It can reload Power BI Desktop, apply changes, and do everything related to working with the Desktop file — all without you touching it.
Setting Up the Connection to a Semantic Model (Video: 6:45)
After having these three pieces ready, the next step is creating a connection to an existing Power BI semantic model. I could even have GitHub Copilot create that connection for me — but I wanted to show you how this connects to an existing repo, so I did it manually first.
I connected to a semantic model already inside a Fabric workspace. This does not have to be a Fabric capacity workspace — it could just as easily be a Premium Per User workspace. This feature has nothing to do with Fabric capacity specifically; it works in a Power BI Pro workspace too.
Once signed in, “Get data from Power BI semantic model” lists every semantic model you have access to. I connected to a semantic model with two tables — a Customer table and a FactInternetSales table. I did not build any visualization — this is a thin report, a report with a live connection to that semantic model. Just a blank canvas at this point.
Saving as PBIP — Not PBIX
I saved the report — and this is an important detail. Saving it as a normal PBIX file does not help, because the format needs to be PBIP (Power BI Project) for this workflow to work. PBIP saves the report as PBIR format behind the scenes — the folder structure that AI can actually read and modify.
Once saved as PBIP, you have a proper Power BI project — with a report section and other sections, all as structured files rather than one binary file.
Building the Report From a Single Prompt (Video: 9:41)
Back in VS Code, I confirmed the Fabric Skills were installed and enabled — in the Extensions section, you can see the installation and enable or disable individual skills, including Report Authoring. It is enabled by default once you install it.
Here is the prompt I used:
“I want to create a new report layout for Report 6, which is already open in Power BI Desktop. I want it professional looking. I want to use the same dataset — sales analysis by customer demographic and also by the order date.”
Notice what I did NOT specify: I did not say which visual types to use, I did not specify a layout, I did not give exact field names. Just a general request.
What Happened Behind the Scenes
Once I submitted this prompt, Copilot first scanned what skills and MCP servers it had available — because connecting to and analysing an existing Power BI semantic model requires an MCP server. It correctly identified the Power BI Authoring Skills to use for building the report, and used the MCP server connection to analyse the semantic model structure.
I said “allow all commands in this session” — but this is important: this is only safe because I was working in my test workspace and test folder. Whenever you give AI this level of authority, make sure you are working in a test environment with good version control in place. All of a sudden, an AI agent could wipe out hours or even days of work if something goes wrong. Good folder structure, good software development lifecycle practices, and good version control are essential before you give an AI this kind of access.
Copilot analysed the report — found it was currently a blank canvas — and worked out what would be good candidates for building the report, based on analysing the semantic model structure. I let it run for a little while.
The Result
Once it finished:
- The report was created
- Copilot automatically switched to Power BI Desktop and showed me the result — I never asked it to do that, it did so on its own
- It automatically reloaded the report — one of the core functions the Desktop Bridge enables
- The report included slicers for gender and marital status
- It included visuals for sales by customer demographic and by order date
- Some dollar values were not formatted correctly — because the underlying semantic model did not have currency formatting applied on those measures. Had the formatting existed in the model, it would have carried through automatically.
All of this — from a single prompt.
💡 This ability to work with Power BI Desktop alongside GitHub Copilot, without me even touching it, is fantastic. This is the combined effort of the Power BI Desktop Bridge and the Report Authoring Skill working together.
There was a small error in the initial build, which Copilot automatically detected and fixed on its own — I did not have to ask it to.
Refining the Report With a Follow-Up Prompt (Video: 15:24)
After seeing the first result, I sent a follow-up prompt:
“That’s great, thanks. I would like to have shadow applied on all visuals. And I would like one of the visuals to be a donut chart.”
Note — I did not specify which visual should become a donut chart. I just gave that general instruction and let the AI decide.
The result: Copilot applied a shadow effect to every visual in the report from that single instruction. It also slightly adjusted the layout — which is fine, but if I had wanted the layout to stay exactly the same, I could have specified that. It also converted one of the visuals into a donut chart, showing that it is genuinely familiar with what Power BI visuals look like and how to construct them — all from two prompts total.
What This Setup Actually Requires — The Full Stack
To recap, here is everything working together:
- VS Code — the development environment
- GitHub Copilot — the AI assistant (or Claude, Codex, or any equivalent tool)
- Power BI Desktop Bridge — the connection that lets the AI see and control the open report in Desktop, apply changes, and reload automatically
- Fabric Skills — specifically Report Authoring Skills — the knowledge layer that tells the AI what a Power BI report structure looks like, what visuals exist, and how to build them
- Power BI Modeling MCP Server — the connection to your semantic model, which the AI uses to analyse fields, tables, and relationships to decide what makes sense to visualise
Together, these let you build a Power BI report front end entirely through prompts — no dragging, no dropping, no manual canvas work.
How to Get Even Better Results
The result I showed came from a very simple, general prompt. You can dramatically improve the outcome by providing more context:
📸 Provide screenshots of professional-looking reports you like — “I want all of my pages to look like this.”
✏️ Sketch the layout yourself. Draw it on paper, take a photo, and say: “This is my report page. These are the sections I want.” Pass that image alongside your prompt.
📋 Pass any business requirements you have gathered — field definitions, what your end user actually needs from the report, documentation from stakeholder conversations.
The more information you provide, the better the outcome. At the time of this demo, I gave a very simple prompt and got a good — but definitely improvable — result. Combine screenshots, sketches, and documentation with your prompts, and you get significantly better, more tailored results.
The Future of Power BI Reporting
I believe there will be far fewer people manually building Power BI reports going forward. It will primarily be built using prompts. Prompt-first reporting may well be the term we use going forward.
This is still a preview capability at the time of recording this video — but it works, it is available today, and it represents a genuine shift in how Power BI report development happens.
Key Takeaways
- Three components make this work: VS Code + AI assistant, Fabric Skills (specifically Report Authoring Skills), and the Power BI Desktop Bridge
- Save your file as PBIP, not PBIX — this is what gives the AI a readable, structured file format (PBIR) to work with
- Enable the Desktop Bridge in Power BI Desktop Options → Preview Features → “Enable external tool to access Power BI Desktop through secure local APIs”
- Always test in a non-production workspace with good version control before giving an AI agent broad command permissions
- The AI can operate Power BI Desktop directly — switching to it, reloading it, and fixing its own errors — without you touching it
- Better inputs produce better outputs — screenshots, sketches, and business requirements dramatically improve the quality of AI-generated reports
- This is Preview technology today, but fully usable — and represents where Power BI report development is heading
Related Resources
- 📝 Power BI File Formats Explained: PBIX vs PBIT vs PBIP vs PBIR
- 📝 No Code Power BI: How to Use Claude and the Power BI Modeling MCP Server
- 📝 No Code, No Tools: Documentation for Power BI and Fabric Using Fabric Skills and GitHub Copilot
- 📝 No Code Dataflow Gen 2 in Microsoft Fabric Using GitHub Copilot and Fabric Skills
- 📝 Fabric MCP Servers, CLI & AI Agents — Fabric Insider Ep. 8 with Hasan Abo-Shally
- 📝 Skills Needed for a Power BI Developer
- 📝 Microsoft Certification Exams for Power BI and Fabric
- 🛠️ Fabric Skills on GitHub
- 🤖 GitHub Copilot
- 🖥️ VS Code
Reza Rad is a Microsoft Regional Director, Data Platform MVP (16 consecutive years), Author, and Trainer. He is the co-founder of RADACAD and the author of multiple books on Power BI, Power Query, and Microsoft Fabric. You can follow him on LinkedIn and subscribe to the RADACAD YouTube channel.
Reza is author of more than 14 books on Microsoft Business Intelligence, most of these books are published under Power BI category. Among these are books such as Power BI DAX Simplified, Pro Power BI Architecture, Power BI from Rookie to Rock Star, Power Query books series, Row-Level Security in Power BI and etc.
He is an International Speaker in Microsoft Ignite, Microsoft Business Applications Summit, Data Insight Summit, PASS Summit, SQL Saturday and SQL user groups. And He is a Microsoft Certified Trainer.
Reza’s passion is to help you find the best data solution, he is Data enthusiast.
His articles on different aspects of technologies, especially on MS BI, can be found on his blog: https://radacad.com/blog.