Microsoft 365 Copilot And Power BI: The Easiest Way To Chat With Your Data

If you’re a regular reader of my blog you’re likely to be a Power BI developer, consultant or admin. You’re probably also under pressure from your management and your end users to do something, anything with AI that will allow them to get answers to business questions with natural language using the data that’s stored in Power BI. If you haven’t done anything about this yet, what should you do? Should you be using Power BI Copilot, M365 Copilot and Power BI, Data Agents, Ontologies or something else? What is the guidance from Microsoft? I thought it would be a good idea to write a post that gives simple, clear advice on how organisations that use Fabric and Power BI can get started with letting people chat with their data.

Let’s deal with Fabric Data Agents and Ontologies first. Both of them are powerful technologies that you’re going to want to use eventually: Data Agents can be used for doing deep data analysis for a very specific set of questions over a number of different Fabric sources including Power BI semantic models; you can think of Ontologies as a layer on top of all the data you have in Fabric, including your Power BI semantic models. I don’t think they are what you should use when you’re starting out though.

My recommendation is that you should begin your “chat with your data” journey with using M365 Copilot and Cowork‘s integration with Power BI semantic models. This was announced back in July in two quite detailed blog posts here and here that you should take the time to read, it’s now GA, and it represents a partial break with the previous Power BI Copilot strategy. What’s different is that, instead of asking data questions in a side pane next to a Power BI report or in a dedicated page in the Power BI portal, you ask your data questions in the M365 Copilot app on your desktop, in the browser or in other places where M365 Copilot lives like Copilot in Excel. What’s not different is that it still uses your existing Power BI semantic models where all your business logic is already defined and all that hard work you put in to prep your data for AI for Power BI Copilot (adding synonyms, AI Instructions and so on) is all still relevant. Only the front-end has changed and it has changed to the place where end users want to ask their questions and do their work. If you tried Power BI Copilot in the past, say last year, you should try this new Power BI integration with M365 Copilot because you’ll find that the answers you get are a lot better. Instructions on how to enable this integration in M365 and in the Fabric admin portal can be found here.

Of course M365 Copilot and Cowork are not your only options for chatting with your Power BI data – you can use lots of other tools with the Fabric IQ MCP server. I think there are some compelling reasons to choose M365 Copilot though and if your company already has the Copilot Business Premium licences necessary then it’s a no-brainer to use them. Other competing data platforms also integrate with M365 Copilot but not as deeply as Power BI does: setting it up the M365 Copilot/Power BI integration is extremely easy and when you do so, data from Power BI can be combined with other organisational data from Work IQ to answer questions in a way that other data platforms’ integrations cannot. The value of Work IQ can’t be overstated: for example if I want to know why sales of widgets increased last month then the data in my Power BI semantic model might be able to tell me that overall sales increased because sales in the teenager demographic increased, but the real explanation – that widgets went viral on TikTok last month – might be in an email I didn’t read or in the transcript of a meeting that I missed because I was on holiday, and you only get that extra context from Work IQ. This demo has some great examples of how M365 Copilot, Cowork and Work IQ come together to answer questions.

Last of all there’s the question of cost. An M365 Business Premium licence (around $20 USD per month) allows you to query any Power BI semantic model in any workspace, whether it’s Pro, PPU or on a P or F SKU capacity. When you do so the only costs you incur are for the CUs consumed by the DAX query if your semantic model is in a workspace on a capacity; if your semantic model is in a Pro or PPU workspace there’s no additional cost for querying it beyond the cost of your Pro or PPU licence. All the AI costs are covered by the Copilot Business Premium licence if you’re only using M365 Copilot, although if you use Cowork for more complex tasks you’ll need to pay for Cowork Credits too. This is in contrast to legacy Power BI Copilot where, when you asked a question in the Copilot side pane next to a report, you had to have a Fabric capacity and it consumed CUs for both the DAX queries that were run and also for the AI costs involved, which led to worries about the impact of these AI costs on your capacities. It’s also in contrast with other data platforms’ integration with M365 Copilot where you pay for the Copilot Business Premium licence and any Cowork Credits, you pay for the cost of the queries that are generated and you pay additional AI costs on the data platform. This means using M365 Copilot with Power BI semantic models can be very cost effective, especially if your organisation already has M365 Copilot Business Premium licences.

As I said at the beginning of this post, this advice is aimed at organisations that already use Power BI heavily and who are deeply invested in the Microsoft ecosystem. I’m a Microsoft employee writing a blog about Microsoft BI so naturally this is my focus. Other data platforms and other ways to chat with your data exist. My point is, though, that if you and your users already have both Power BI and M365 Copilot you should link them up: it’s easy to do, it’s surprisingly cheap and it gives you really good results when you ask data questions. Try it and see for yourself.