Back to All Engineering Notes
Fabric & Modernization•8 min read•2026-10-10

Power BI Q&A Is Retiring in December 2026: What to Replace It With

Every Q&A visual stops working in December 2026. Here is how to find where you use it, what Copilot needs, and how to switch without a broken experience.

Core Engineering Takeaways

  • •Microsoft announced the Q&A deprecation in December 2025; Q&A experiences and Q&A setup are fully retired in December 2026.
  • •Existing Q&A visuals in reports, dashboards, mobile and embedded scenarios will stop working — this is not a soft deprecation.
  • •Copilot is the stated replacement, but it needs paid capacity (F2+ or a P SKU, not trial) and depends on clear model names and descriptions.
  • •Not every Q&A use case should become Copilot: some are better as a curated report page, and some can simply be removed.

1. What exactly is being retired

Q&A was Power BI's original natural-language feature: a visual or dashboard box where users type a question and get a chart back. Microsoft announced its deprecation in December 2025, and the retirement lands in December 2026.

Two things are in scope:

  • Q&A experiences: Q&A in reports (the Q&A visual), Q&A on dashboards, and Q&A in Power BI Embedded.
  • Q&A setup: the tooling authors used to tune it — synonyms, linguistic relationships, row labels, hide from Q&A, teach Q&A, review questions and suggested questions.

After the retirement, existing Q&A visuals in reports, dashboards, mobile and embedded apps will no longer work, and new Q&A experiences cannot be created. If an executive's favourite dashboard has a Q&A box at the top, it will break.

Microsoft's stated replacement is Copilot and the chat with your data experiences in Power BI. Microsoft has also said it is reviewing the approach for Pro licences in this transition, so check the current licensing guidance before you commit to a plan.

The deadline is a calendar date, not a renewal date. Unlike a capacity migration, there is no grace period: plan the switch for well before December 2026.

2. Find out where Q&A is actually used

Most teams underestimate their Q&A footprint, because the most visible usage is the easiest to forget: the question box pinned to a dashboard years ago. Before deciding anything, build an inventory.

Where to look

  • Reports: any page containing the Q&A visual. In PBIP projects this is a text search across the report definition; for PBIX files, open each report or use a scanning script against the workspace.
  • Dashboards: Q&A is enabled on dashboards by default, so the question is whether anyone uses it. Usage metrics for the dashboard tell you more than the presence of the box.
  • Embedded apps: if you embed Power BI in a portal or product, check whether your application exposes Q&A to customers. This is the highest-risk category because the people affected are outside your organisation.
  • Q&A setup investment: models with many synonyms or linguistic relationships tell you someone cared about natural-language quality. That knowledge — what users call things — is worth keeping.

For each item, record the owner, how often it is used, and who would notice if it disappeared. That last column decides the priority.

3. Choose a replacement per use case

The temptation is to swap every Q&A visual for Copilot. In practice, Q&A was used for three quite different jobs, and each has a different best replacement.

  1. Exploration by curious users — people typing ad-hoc questions to see what the data says. This is Copilot's sweet spot, provided the model is well described.
  2. A shortcut to a handful of recurring answers — the same five questions every week ("sales by region this month"). A well-designed report page, or a few bookmarks, answers these faster, cheaper and more reliably than any natural-language feature.
  3. Decoration — a Q&A box nobody has used in a year. Remove it.

Splitting the inventory into these buckets typically shrinks the Copilot work to a small number of models, which matters because Copilot has real costs.

4. What Copilot needs that Q&A did not

Copilot is not a drop-in swap, for three reasons.

It needs paid capacity. Copilot in Fabric and Power BI requires an F2 or larger capacity, or a P SKU, in a supported region, and it is not available on trial capacities. If your capacity is outside the US or EU data boundary, a Fabric admin may need to enable the tenant setting that allows data to be processed by Azure OpenAI outside your region. See the Fabric SKU comparison for capacity sizes and costs.

It consumes capacity units. Every Copilot interaction uses CUs on the same capacity that serves your reports and refreshes. On a capacity that is already close to throttling, adding Copilot usage can push it over. Check headroom with the capacity estimator before rolling it out widely.

Its answers depend on your model's metadata. Copilot reads table, column and measure names and descriptions to understand what your data means. A model full of fct_sales, dim_cust_v2 and Measure 7 will produce poor answers no matter how good the AI is.

5. Prepare the semantic model

This is where most of the real work is, and it improves the model for human users too.

  • Business names. Rename tables, columns and measures to what users actually call them. Q&A synonyms are a good source: they record the vocabulary people used.
  • Descriptions. Add a one-sentence description to every measure and every column a user might ask about. Say what it means and any nuance ("Net revenue after returns, excluding tax").
  • Hide the plumbing. Hide surrogate keys, technical columns and helper measures so they cannot be picked up as answers.
  • One clear date table, marked as a date table, with the fiscal attributes people ask about.
  • Unambiguous relationships. Bidirectional or ambiguous paths confuse people and AI alike.

If your models live in PBIP/TMDL, this work is reviewable in Git: every rename and description shows up as a diff. That makes it much easier to split across a team and to keep consistent.

Carry the knowledge in your Q&A synonyms into names and descriptions before December 2026 — after the retirement, the Q&A setup tooling goes away with it.

6. Switch over and validate

With a prepared model and capacity in place:

  1. Replace Q&A visuals with the chosen alternative, starting with the most-used items in your inventory.
  2. Collect the questions users actually asked Q&A (review questions, interviews, or simply asking the owners) and test them against Copilot. Fix names and descriptions where answers are wrong.
  3. Watch capacity usage for the first weeks after rollout.
  4. Write a one-page guide for report authors on keeping models Copilot-friendly, so new content does not regress.

None of this is hard engineering, but it touches many reports and owners, and it has a fixed deadline. Start with the inventory now — it is the step that tells you how big the job really is.

If you would rather hand it off, the Q&A → Copilot readiness sprint covers the inventory, model preparation, capacity checks and switch-over at a fixed price.

Power BI Diagnostic Audit

Get ready before December 2026

Check your capacity headroom for Copilot, or hand the whole Q&A replacement to a fixed-scope sprint.

Get the next Engineering Note

Practical DAX, semantic model and Fabric notes. 1–2 a month, unsubscribe any time.

1–2 useful notes a month. Unsubscribe at any time.

Piyush JainPower BI & Fabric Engineer
#Q&A#Copilot#Deprecation#Semantic Models