A few years ago, building a Power BI report meant hours of writing DAX by hand, searching forums for a Power Query error, and documenting measures at the end of a project, if there was time left. AI has changed that.
Today, AI assistants such as ChatGPT and Claude are part of my daily Power BI work. More recently, Claude Code combined with MCP has gone a step further: the AI can work directly inside a Power BI model, not just suggest code in a chat window.
In this article I'll share how I use these tools in real projects, what they are good at, and where you still need a person who understands the business.
Level 1: AI as your Power BI assistant
The simplest way to start is a chat assistant like ChatGPT or Claude. You describe what you need, and it suggests the formula or explains the code. This is where most people see value within the first day:
- Writing DAX measures from a plain-language description, such as "year-to-date sales compared with the same period last year"
- Explaining existing DAX that someone else wrote, line by line
- Fixing Power Query errors by pasting the error message and the M code
- Translating Excel formulas into DAX or Power Query steps
- Drafting documentation, measure descriptions and user guides
- Turning requirements into a plan: a list of tables, relationships and measures for a new report
The quality of the answer depends heavily on the quality of the question. Give the assistant your table and column names, explain the business rule, and say what the result should look like:
I have a Power BI model with tables Sales (Date, ProductKey, Amount)
and Date (Date, Year, Month). The Date table is marked as a date table.
Write a DAX measure for sales year-to-date, and a second measure for
the same period last year. Our financial year ends on 30 June.A prompt like this gets you a working answer in seconds. A vague prompt such as "give me YTD DAX" gets you a generic answer you still have to adapt.
Level 2: Claude Code and MCP, AI that works inside your model
Chat assistants have one big limitation: they can't see your model. You copy code in and out, and the AI only knows what you paste.
This is where MCP (Model Context Protocol) comes in. MCP is an open standard that lets AI assistants connect to other software in a safe, structured way. Microsoft has released an MCP server for Power BI semantic models, which means an AI agent can connect to a model open in Power BI Desktop and work with it directly.
Claude Code is Anthropic's AI agent that runs on your computer. Connected to Power BI through MCP, it can:
- Read the whole model: tables, columns, relationships and existing measures
- Create and update measures in bulk, for example a full set of time intelligence measures for every base measure
- Add descriptions to every measure and column, so the model documents itself
- Organize the model: display folders, consistent naming and format strings
- Run DAX queries to test that new measures return the expected numbers
- Review the model for common problems, such as unused columns or missing relationships
In one recent project, documenting more than a hundred measures, which used to take most of a day, took minutes. I reviewed every description, but I didn't have to type them.
Claude Code also works well with the Power BI Project (PBIP) format, where a report and model are saved as text files. That makes it possible to keep the model in version control, review every change the AI makes, and roll back if something isn't right.
Where AI helps most, and where it doesn't
AI is excellent at:
- Repetitive work, such as creating many similar measures or descriptions
- First drafts of DAX and Power Query code
- Explaining code and error messages
- Spotting inconsistencies in a large model
AI still needs a person for:
- Business rules. It doesn't know how your company defines revenue, allocates costs or closes a month
- Data modeling decisions. A poor star schema produces wrong numbers, however good the DAX is
- Validation. Every number must be checked against a trusted source before a report goes live
- Judgment. Knowing which question the report should answer in the first place
In other words, AI makes a skilled Power BI developer much faster. It doesn't replace the need to understand data modeling and DAX. If anything, those skills matter more, because you have to recognize when the AI is wrong.
Using AI safely with company data
A few simple rules keep AI use safe:
- Don't paste confidential data into consumer AI tools. Use business plans with clear data-protection terms, and follow your company's AI policy.
- Share structure, not data, wherever possible. Table and column names are usually enough for the AI to help.
- Work on a copy or use version control, so every AI change can be reviewed and undone.
- Test everything. Treat AI output like work from a fast but new team member.
How to get started
If you're new to AI in Power BI, start small:
- Use ChatGPT or Claude to explain a measure you don't fully understand.
- Ask it to write the next measure you need, then check the result.
- Once you're comfortable, try an AI agent such as Claude Code with MCP on a copy of a model, starting with documentation.
Final thoughts
AI-powered development is quickly becoming the normal way to build Power BI solutions. Teams that learn to use it well deliver faster, document better and spend more time on the analysis that actually matters.
At Tecrionix, our Power BI Training program includes a module on using AI assistants in Power BI work. To learn more, contact us.
