Analysing Power BI Data Using Copilot In Excel

This week there was another big Copilot-related announcement for Power BI – but on the Excel blog. You can now analyse data from Power BI semantic models using Copilot in Excel. Full details on how to enable this are in the docs here, but what does it look like?

Using the same semantic model from last week’s blog post on doing ABC analysis in the M365 Copilot app, I opened Excel, opened the Copilot pane and entered the same prompt:

Use this semantic model <insert URL of semantic model here> to do an ABC analysis. 
The upper boundary for the first group is £290000 and the upper boundary for the second group is £790000. 
Filter the transactions to just 20th January 2025.

…and fairly soon after saw two worksheets created. One had a few thousand rows from the fact table in my semantic model:

…and the other had the ABC Analysis I asked for:

Interestingly, unlike in my blog post last week, Copilot did not use the DAX UDF in the semantic model to do the ABC analysis. Instead it did what most Excel users would prefer and grabbed the raw data, did some of the analysis in the DAX query that got that raw data, then did most of the analysis using Excel formulas. It got the answer I was expecting though and the logic was consistent with the logic in my DAX UDF.

The data on the transactions worksheet was a static copy and I would love it if, in future, Copilot created Excel tables linked to a DAX query and a connection back to the semantic model so that when the data in the model changed then the workbook would update automatically.

Next, just to see if it worked, I used the brandkit skill (also announced this week) to make it look more corporate with the following prompt:

 @brandkit make this look like a Microsoft branded report

Luckily there was already a Microsoft brand kit available to me; the result was a report which (I guess) was a bit more Microsoft-blue and which had a Microsoft logo in the top right-hand corner:

Job done, report created.

So what? The functionality exists and it works well, which is good to know. But there’s a more interesting question to consider here: who cares? Is anyone going to use this? After all, “AI helps you analyse data” is not exactly headline news anymore. But because Excel Copilot:

• Works well enough – which this seems to do
• Works well enough before other competing tools which also work well enough become commonplace on corporate desktops
• Produces analysis in a way – Excel formulas – that is more accessible and understandable to more people than, say, Python code
• Does this analysis directly inside the tool that most people already know and love, Excel, without needing anything extra installed and not in some other tool – even if that tool can generate Excel workbooks

…then I think there’s a very good chance that this will be popular. Probably more popular as a way of analysing data from Power BI semantic models than the M365 Copilot app, and maybe even more popular than Power BI reports as a way of analysing data from Power BI semantic models.It’s Export to Excel for the age of AI!