Technology
Microsoft is retiring Q&A in Power BI and points to Copilot, which needs Fabric capacity. What goes, what the alternative needs and what about your documents?
Microsoft is retiring Power BI Q&A, the old tool that lets you type a question such as "revenue by region in 2025" into a report or dashboard and get a chart back. From December 2026 you can no longer create new Q&A visuals, and existing visuals in reports, dashboards, the mobile app and embedded reports stop working and are removed.
That is stated in the announcement Microsoft published in the Microsoft 365 Message Center on 16 January 2026 (MC1218421, archived by Merill.Net). In it, Microsoft itself calls Q&A its "legacy" tool for plain-language questions. The settings behind it go too: the synonyms, the linguistic relationships and the feature you used to teach Q&A new terms. According to the same announcement, Q&A and its settings will be fully retired by the end of December 2026.
A third party reports that in September 2026 Microsoft moved the date to February 2027 to give customers more time (Solv Systems, October 2026). We could not find this in the Microsoft sources we were able to read on 12 November 2026. We therefore assume December 2026. Check in the Message Center of your own tenant which date applies to you, and plan as if it is December: a few extra weeks of grace are welcome, but you do not want to discover in January that your dashboard is empty.
If you never switched Q&A on, you will notice nothing. If you do use it, you lose three things.
A Q&A visual in a report stops and disappears. The same applies to the question box at the top of a dashboard and to questions in the mobile app. Colleagues who are used to typing a question instead of filtering themselves will need to find another way.
Teams that made Q&A work well often spent time on synonyms and linguistic relationships, so that "client" and "debtor" meant the same thing. Those settings go with Q&A. For preparing data, Microsoft points to a separate part of Copilot, "Prep data for AI". Expect to redo that work there.
If you embed Power BI reports in your own application or portal, the Q&A parts in them stop as well. The same third-party source reports that Microsoft currently plans no Copilot replacement for this. We could not verify that with Microsoft. If you use embedded Q&A, check with your supplier or your Microsoft partner.
Microsoft recommends moving to Copilot in Power BI. That is a language model that works with your reports and datasets: it can summarise a page, suggest a visual or answer a question about the figures. In substance that is a step forward from Q&A.
The hurdle is the licence. According to Microsoft's documentation on Copilot in Power BI (updated in August 2026, consulted on 12 November 2026), you need paid Fabric capacity, from F2, or Power BI Premium from P1. A Power BI Pro or Premium Per User licence alone is not enough, and trial capacities and free SKUs are not supported. On top of that, an administrator must have the Copilot setting switched on, and the capacity must be in a region where Copilot is available.
For an organisation that uses Power BI with Pro licences, this means: Q&A was simply included, Copilot is not. If you want to keep asking questions in plain language, you add a capacity with its own costs and its own administration. What that capacity costs depends on your contract and the size you choose. Ask your Microsoft partner before you decide.
Copilot in Power BI uses Azure OpenAI models. If your tenant or capacity is outside the United States and Microsoft's EU data boundary, Copilot is off by default until an administrator allows data to be processed outside your own region. That does not apply to a tenant in the EU, but it is worth checking which settings are switched on in yours. What Microsoft further documents about processing and training in Microsoft 365 is covered in our article on Copilot and your SharePoint.
This is the real question behind the end of Q&A. Q&A was built for figures: tables of revenue, hours, quantities. That is a small, well-structured corner of what an organisation knows.
Most of the knowledge of an accountancy firm, law firm or consultancy is not in a dataset but in documents. A colleague wants to know which notice period is in a lease, what was agreed with a client last year about the annual accounts, or in which case files a certain clause appears. Q&A could never do anything with that, and neither can Copilot in Power BI: it works with your reports, not with your SharePoint library.
| Kind of question | Example | Where the answer is | Suitable tool |
|---|---|---|---|
| About figures | How many hours did we log per department in October? | A dataset in Power BI | Copilot in Power BI, or a report with filters |
| About an overview | Which clients grew fastest? | A dataset in Power BI | Copilot in Power BI |
| About a document | Which notice period is in the contract with client A? | A document in SharePoint | Searching documents with an AI assistant |
| About a case file | What was agreed in 2025 about the annual accounts? | Emails, notes and papers in a case file | Searching documents with an AI assistant |
| About a pattern | Which case files contain a penalty clause? | Spread over many documents | Searching documents with an AI assistant |
As at November 2026. The split is a guideline: some questions need both sources.
Before you buy a capacity, make a simple list of the questions colleagues now ask Q&A, and of the questions they would actually like to ask. If most are about figures, Copilot in Power BI is the logical next step. If most are about documents, a Fabric capacity solves the wrong problem.
The obvious move is to drag a document into ChatGPT or Claude and ask a question. For a single public document that is fine. For a client file it is different: it contains names, addresses, financial data and sometimes health data of people who did not choose to be in it. As soon as you upload such a file, the provider processes that personal data, and then the rules of the GDPR apply: a lawful basis, a data processing agreement, security and a clear purpose.
There is something else. A standalone chatbot does not know who is asking. It does not know your SharePoint permissions and cannot check whether a colleague may see a document. If you want to make case files searchable for an AI assistant, you need something that sits between the documents and the assistant and does answer those questions. What goes wrong when it is missing is covered in our article on uploading documents to an AI chatbot.
PrivaSearch does not replace Power BI Q&A and does nothing with your dashboards or datasets. It is meant for the other half: the documents.
You connect a SharePoint library read-only, or you add documents to a project. PrivaSearch detects the personal data in them and replaces it with pseudonyms; a colleague checks that before a document becomes searchable. Within PrivaSearch, project membership determines who sees what, and on every request PrivaSearch checks live whether the user may still see the document in SharePoint. How that works step by step is on the page How PrivaSearch works.
Then you connect the AI assistant you already use or want to start using. That connection works with Claude, ChatGPT Business and Mistral Vibe, via the open MCP standard. What that is, you can read in our article what is an MCP server. The assistant only receives released, pseudonymised passages with a source reference, never the original file. You choose which language model you use and you are not tied to one supplier.
Two caveats. Pseudonymised data remains personal data for your organisation: it reduces the risk, it is not an exemption from the GDPR. And whatever you type into the assistant yourself goes straight to the provider of that assistant. Which assistants you can connect and what you need to arrange for that is on our page on AI chatbots and the GDPR.
If you want to see how asking questions about your own documents works with the assistant you choose, request a demo.
According to Microsoft's announcement in the Message Center (MC1218421, January 2026), Q&A ends in December 2026 and is fully retired by the end of December. A third-party source reports a move to February 2027, but we could not find that with Microsoft. Check in the Message Center of your own tenant which date applies.
They stop and are removed from reports, dashboards, the mobile app and embedded reports. You can no longer create new ones either. The settings behind them, such as synonyms and linguistic relationships, go with them. Before the end date, note which reports use Q&A, so you know what you need to replace.
No. According to Microsoft's documentation, a Pro or Premium Per User licence alone is not enough. Copilot in Power BI requires paid Fabric capacity from F2 or Power BI Premium from P1, and does not work on trial capacity. An administrator must also have the Copilot setting switched on.
No. PrivaSearch does nothing with dashboards or datasets in Power BI. It makes documents searchable, such as case files and contracts in SharePoint, for an AI assistant you connect yourself. If your questions are about figures, Copilot in Power BI or a good report is the suitable tool. If they are about documents, PrivaSearch helps.
Claude, ChatGPT Business and Mistral Vibe, which was called Le Chat until May 2026. The connection uses the open MCP standard, and you approve it after signing in at your organisation's address. The assistant only receives released, pseudonymised passages with a source reference, never the original document.
Not without preparation. A case file contains personal data, and as soon as you upload it the provider processes that data. You then need, among other things, a lawful basis, a data processing agreement and appropriate security. Pseudonymisation reduces the risk, but the data remains personal data. Record in your AI policy what is and is not allowed.
Why software that finds names, addresses and numbers in client files always misses something or masks too much, and how a human review catches it.
A small window in the workspace shows in which project your assistant searched and which documents it found. The assistant itself does not get the names.
GreenPT is working on MCP with sign-in. What a connection with PrivaSearch would then offer: searching your client files with pseudonymised passages.
We show it with fictitious client files from your sector: upload, detection, review, search and the connection with your AI assistant. Afterwards you'll know exactly what your team can use safely.
Book a demo →