A while back I wrote about a smarter Q&A tool for Power BI, built entirely from scratch. That original project replaced Power BI's native Q&A visual with something much more focused: a drop–down list of preset questions, each answered instantly and accurately using the data already sitting inside the report.
That setup was never meant to cover every possible question under the sun. The whole point was speed. Instead of digging through filters and slicers to find a key figure, you simply picked a question from the list and got your answer a second later. For the handful of things people ask for time and time again, that fixed list did the job brilliantly.
Now I have built on top of that foundation to take things a step further. I put together a new version that lets people type their question in plain, natural English, however they would normally phrase it, and still get a clean, accurate text answer. No drop–down menus, no preset options. Just a simple text box and a direct answer.
How it works
The core idea is straightforward, even if getting it to work smoothly took a bit of effort. You type a question, something like "top 3 brokers in retail by deal count", and within a second or two you get a clear answer written out in the same clean style as the original version.
Behind that simple surface, a lightweight AI model reads your question and figures out what calculation is needed. Crucially, it never sees your actual data. It only ever looks at the question you typed and the names of the fields in your dataset - things like broker, property type, or deal amount. From that, it works out the steps: group by broker, filter to retail, count the deals, sort from highest to lowest, then take the top three.
That set of instructions is sent right back to the visual inside your report, which performs the actual calculation using the data it already has loaded. The AI never touches your numbers. It simply decides which calculation to run, and the report handles the heavy lifting.
This detail matters far more than it might sound. It means none of your property names, deal values, client details, or sensitive figures ever leave the report or get sent anywhere else. The only information traveling outside is your typed question and a list of basic field names, nothing that could ever identify a client, a deal, or a meaningful figure on its own.
It also means the feature runs perfectly on a standard Power BI Pro licence. There is no need to upgrade to Fabric or Premium capacity, because none of this relies on Microsoft's AI infrastructure. It is a completely self-contained setup running alongside your report.
The end result looks and feels just like the original visual: same dark header, same clean fonts, same answer cards and ranked lists. The only difference on the surface is that a text box has replaced the drop–down menu, and you are no longer limited to a fixed list of seven questions. You can ask almost anything the dataset is capable of answering, using whatever words come naturally to you.
What it costs to run
This is the bit people usually expect to be complicated, but it really is not.
Each question costs a tiny fraction of a penny. The AI model reads a brief description of the dataset fields along with your question, then replies with a short set of instructions — nothing more. That complete round trip works out to roughly a tenth of a US cent per question (about $0.0011).
If you scale that up to a typical office using it occasionally, say ten to fifteen people asking a handful of questions a week, you are looking at around $0.39 a month. Even in a busy team asking hundreds of questions every day, the total cost stays well under two dollars a month.
There are no hidden licence fees sitting on top of that either. No Fabric capacity, no per–user charges, no monthly subscriptions. Just that tiny usage cost. Even hosting the lightweight backend infrastructure behind it is free at this scale, as it fits easily inside the free tier offered by most cloud providers.
All in all, the total running cost for a whole team using this daily comes to well under five dollars a month.
Why this matters right now
Here is the reason this shifts from a nice technical experiment into something genuinely worth paying attention to.
Microsoft is officially retiring the legacy Power BI Q&A feature, along with all the tools used to configure it, in December 2026. This is not a slow phase–out where things gradually degrade. Any native Q&A visual currently sitting in a report, dashboard, mobile app, or embedded report will simply stop working. There is no backup option and no fallback mode. It goes straight from working to gone.
Microsoft's suggested replacements are Power BI Copilot and Microsoft Fabric Data Agents, both built around heavy generative AI. On paper that sounds like an upgrade, but in practice it brings a completely different cost model. The old Q&A visual was entirely free. These new tools are billed against Fabric Compute Capacity and, in many cases, require extra per–user licensing on top.
To put some real numbers on that, the entry-level F2 capacity starts at $262.80 a month on pay–as–you–go, or $156.34 a month if you commit to a one-year reservation. And that is just to turn the light on. On top of that base fee, every single question your team types consumes compute units from your pool. The more people use it, the more capacity it burns through.
For anyone who built reports around the free native Q&A visual, December 2026 is a hard deadline. Once that date hits, the feature disappears, and Microsoft's official way back in will cost hundreds of dollars every month before a single question has even been typed.
What this means for you
When you put those two options side by side, the argument for a custom solution becomes very clear.
On one hand, you have Microsoft's official path: a monthly capacity bill starting at well over $150, rising with every question asked, plus potential extra user licences. On the other hand, you have a tailored visual that costs a fraction of a penny per question and runs on the basic Power BI licence you already pay for.
Yet the cost saving is only half the story. The bigger benefit is independence. When you own the solution, nobody else gets to decide when it changes, what it costs, or whether it disappears. You are no longer tied to Microsoft's product roadmap, sudden pricing updates, or another deprecation notice landing in your inbox a few years down the line. It fits your report design perfectly, looks exactly how you want it to look, and keeps working reliably for as long as you need it to.
For reports still relying on the old native Q&A visual, December 2026 is coming up fast. It is definitely worth getting a plan in place now, rather than waiting until things stop working.