Power BI Q&A Retirement: Migration Guide to Copilot

The Power BI Q&A retirement is on the calendar, and if any of your reports or dashboards rely on typing a question into a box, you have a short window to act. According to Microsoft's announcement, Q&A, Power BI's legacy natural language tool, will be fully retired by the end of December 2026. After that, new Q&A visuals cannot be created and existing ones stop working.

This guide explains what is changing, how to find out whether you are affected, and how to replace Q&A with Copilot in a sensible order. It is written for Power BI admins, report owners and technical leads. There is no need to panic, but there is a good reason to start early: a Q&A visual that stops working does not warn anyone first.

Quick answer

• What is changing: Q&A experiences and Q&A Setup retire by the end of December 2026, according to Microsoft's announcement.

• Who is affected: any organisation using Q&A in reports, dashboards, mobile or embedded analytics.

• What to do first: list every place Q&A is used, decide the Copilot replacement for each, prepare the semantic model, then test with real questions.

What the Power BI Q&A retirement means in practice

Microsoft published the retirement notice in the Microsoft 365 Message Center (MC1218421) and on the Power BI blog under "Deprecating Power BI Q&A". Both point users to Power BI Copilot as the way to query data in natural language, and both mention "Prep data for AI" as the alternative to Q&A Setup.

Q&A can appear in four places, so your audit needs to cover all of them:

• Report visuals: a Q&A visual placed on a report page.

• Dashboards: the Q&A question box at the top of a dashboard in the Power BI service.

• Mobile: the Q&A experience at the bottom of a dashboard in the Power BI mobile app.

• Embedded analytics: Q&A added to custom apps through APIs or embedded tokens.

Microsoft's blog suggests a replacement for each:

Where Q&A is used Microsoft's suggested replacement Reports Copilot with Power BI reports Dashboards The standalone Copilot experience Mobile Copilot in Power BI mobile Embedded Copilot for SaaS scenarios; PaaS support was described as coming later

One practical note on dates. The two announcements use slightly different wording ("December 2026" in one place, "by the end of December 2026" in another). Treat 1 December as your working deadline and you will not be caught out.

What a "Q&A visual not working" problem will look like

Today, a Q&A visual that misbehaves is usually an ordinary issue you can troubleshoot. After retirement, a Q&A visual that stops responding is working as designed, and there will be nothing to repair. Plan to replace it rather than fix it.

Step 1: Audit every report and dashboard that uses Q&A

You cannot replace what you cannot find. Build a simple inventory before you change anything.

1. Start with your workspace list. If you already inventory your tenant with your usual admin tooling, use that list as the starting point.

2. Open the reports. Check each report page for Q&A visuals. Visual-level detail usually needs a human look.

3. Check dashboards. Open each dashboard in the service and note whether the Q&A box is used by its audience.

4. Ask owners and users. Send a short message asking who relies on typed questions and which ones they ask most.

5. Check mobile and embedded. Ask the team that owns any custom app whether it uses Q&A through APIs or embedded tokens.

Record each finding in a table the whole team can see. A simple layout works well:

Workspace Report or dashboard Owner Where Q&A is used Audience Priority Planned replacement

Priority matters. A Q&A box that executives use every week deserves attention before a forgotten test report.

Step 2: Decide what replaces each use

Not every Q&A visual needs a one-for-one swap. For each item in your inventory, choose one of four outcomes:

• Copilot in reports: best where users asked free-form questions on a report page.

• Standalone Copilot on dashboards: the route Microsoft suggests for the dashboard Q&A box.

• A redesigned report page: if people only ever asked the same five questions, a well-built page with clear visuals and slicers may serve them better than any chat box.

• Retire it: if nobody uses it, remove it and tidy up.

This is also the moment to look at the questions people actually ask. Your existing Q&A setup, such as synonyms and suggested questions, is a useful record of how your business talks about its data. It will help you in the next step.

If you want exceptional results in Power BI with 100% accuracy. Book a meeting with the C Apps team and get your job done professionally.

Step 3: Prepare the semantic model for Copilot

Copilot does not fix a messy model. It reads the model you give it, so the quality of its answers depends heavily on how well that model is prepared. This is the idea behind "Prep data for AI", and it is the core of how you prep data for AI in Power BI.

Power BI now calls datasets "semantic models", so you may see both terms in older documentation and forum posts. Everything below applies to the semantic model behind your reports, previously known as the dataset.

According to Microsoft's documentation, Prep data for AI offers four main ideas for getting a model ready:

• AI data schema: choose a simplified subset of your tables, columns and measures for Copilot to consider, so it is not distracted by helper fields and clutter.

• Verified answers: link common business questions to specific visuals, so that a question phrased in a familiar way returns an answer you have already checked.

• AI instructions: give Copilot model-level guidance, such as how your business defines a term or which fields to prefer.

• Descriptions: add plain-English descriptions to tables, columns and measures so their meaning is clear.

Microsoft's guidance suggests working in that order: schema first, verified answers second, and instructions last, for fine-tuning.

Good model habits that help Copilot

• Use clear, business-friendly names for tables, columns and measures.

• Hide technical or duplicate fields that users should never need.

• Keep each measure's definition consistent, and describe it.

• Turn your most-asked questions into verified answers, starting with the ones you found in Step 2.

If the problems start further upstream, in inconsistent source data or duplicated tables, tidying the data feeding the model will help more than any Copilot setting.

One caution. At the time of writing, some Microsoft documentation lists Q&A being enabled on the model as a prerequisite for setting verified answers. Because Q&A itself is retiring, check the current wording before you rely on it. See the Prep data for AI verified answers documentation for the latest.

Step 4: Rebuild, test with real questions and tell your users

With the model prepared, put the replacement in place and test it properly.

Test with real questions. Gather a set of ten to twenty questions that your users actually ask, taken from the audit and from owner conversations. Run each one through Copilot and compare the answer with a number you already trust. Generative answers can vary, so use verified answers for the questions that must always come back the same way.

Test as your real audiences. Try the experience with accounts that represent different users, including anyone with restricted access to data.

Communicate the change. Tell users what is changing and when, show them where to ask questions now, and update any internal guides or documentation that mention Q&A. Remove or replace old Q&A visuals before the deadline, rather than leaving a broken box on a page.

Keep a record. Note what you changed and why, per report. It makes handover and later reviews much easier.

Licensing and admin prerequisites for Copilot

Copilot is not simply switched on for everyone. Before you commit to a plan, confirm the following in Microsoft's documentation, because requirements and availability can change:

• The licence and capacity requirements that apply to your organisation.

• The tenant settings an admin must enable for Copilot.

• Whether Copilot is available in your region.

• Whether Copilot covers every place you used Q&A, including mobile and embedded scenarios.

Do this early. If Copilot cannot be enabled for some of your users, you will want to know now, when a redesigned report page is still an easy alternative.

Power BI Q&A migration checklist

Copy this list into your project tracker:

• Read Microsoft's current retirement announcement and note the exact date

• Inventory every report, dashboard, mobile and embedded use of Q&A

• Record owners, audiences and priority for each item

• Decide the replacement for each: Copilot, standalone Copilot, redesigned page or retire

• Confirm Copilot licensing, capacity, tenant settings and regional availability

• Simplify the AI data schema on each semantic model in scope

• Create verified answers for the most-asked questions

• Add AI instructions and descriptions where definitions are unclear

• Test with real user questions and compare against trusted numbers

• Test with accounts that match each audience

• Tell users what is changing and where to ask questions in future

• Update documentation and internal guides

• Remove or replace Q&A visuals before the retirement date

• Re-check the list for anything missed, including embedded apps

Common mistakes

• Waiting until December. A retirement is easier to manage with weeks of runway than days.

• Auditing only reports. Dashboards, mobile and embedded apps are easy to forget.

• Treating Copilot as a drop-in replacement. It is a different, generative experience, so answers and behaviour will differ from Q&A.

• Skipping model preparation. An unprepared model gives vague answers, and users blame Copilot.

• Testing with invented questions. Use questions your users really ask.

• Not telling users. A silent change means support requests later.

• Assuming licensing is sorted. Check requirements before you build.

FAQ

Will my Q&A visuals break?

According to Microsoft's announcement, existing Q&A visuals stop working after the retirement, and new ones can no longer be created. Plan to replace them before the end of December 2026.

Is Copilot a drop-in replacement for Q&A?

No. Microsoft describes Copilot as a more advanced, generative AI experience, but it works differently. Expect to prepare your semantic model and test the answers rather than just swapping one for the other.

What replaces the Q&A box on a dashboard?

Microsoft's blog points to the standalone Copilot experience for dashboards. Check that it is available for your users before you decide.

Do I have to rebuild my semantic model?

Not necessarily. Most teams will simplify and describe what they already have, then add verified answers and instructions. A messy model may need deeper cleanup.

Get help reviewing your Q&A reports

Cee Apps is a Power BI services partner. The team takes the time to understand how your business actually uses its reports, then turns those needs into working report logic.

Cee Apps can help in three ways:

• Business intelligence reporting: report and dashboard design in Power BI, including reviewing reports that rely on Q&A and redesigning them.

• ETL and data warehousing: preparing and organising the data that feeds Power BI, so your semantic models start from a clean base.

• Scripting and automation: Zapier, Zoho Flow, Make and Google Apps Script for the workflows around your reporting.

If you would like a second pair of eyes on your Q&A reports before the deadline, Book a Meeting.