Forum Discussion
Questions on "Prep data for AI"
I have been starting doing some tests on this to see how we are going to use it.
I have some questions related to my findings.
1. I have noticed Q&A needs to be activated to be able to configure this. And that will go away in February.
What will happen after that?
2. Do we always need to use large semantic models to get a verified answer in the service?
3. When deploying dataset to our dev/test workspace I don't get all headings activated even if I use large model storage format. What can be the reason for that? This workspace is configured with deployment pipeline, is that the reason?
If I publish to another workspace I get all headings activated.
4. Related to Chris blog post.
What is your recommendations related to this. Are you using M365 instead of Copilot in Power BI?
3 Replies
- BeaBF
Super User
Hi TobNyg , Great questions! Let me share a few considerations that might help (i work at microsoft italia).
- Q&A retirement
Power BI Q&A experiences are scheduled for retirement in February 2027, with Copilot in Power BI becoming the recommended natural-language experience.
Since Prep data for AI currently still has a dependency on Q&A for certain features, it would be useful to get clarification from the product team on how this prerequisite will evolve.
- Large semantic model storage
Large semantic model storage format shouldn't be confused with the actual size of your dataset.
Based on your screenshots, enabling this format appears to unlock additional authoring capabilities. However, I would distinguish this from the requirements for Verified Answers themselves, which are also dependent on Q&A being enabled and the appropriate Copilot configuration.
- Deployment pipelines and workspace differences
One thing I would check first is whether your dev/test workspace and the workspace where everything works are assigned to the same capacity type.
Copilot requires a supported Fabric or Premium capacity. If your dev/test workspace is on shared capacity while the other workspace is assigned to a supported Fabric or Premium capacity, this could explain the different behavior.
I would also verify:
- Whether Copilot is enabled for both workspaces.
- Whether the semantic models use the same storage format.
- Whether you have the required authoring permissions.
I wouldn't necessarily attribute the issue to deployment pipelines themselves.
- Copilot in Power BI vs. Microsoft 365 Copilot
This is particularly interesting given the evolving integration between Power BI Copilot and Microsoft 365 Copilot.
The direction is not to replace Copilot in Power BI with Microsoft 365 Copilot, but rather to bring the two experiences closer together.
They serve complementary purposes:
- Copilot in Power BI provides a dedicated analytical experience for exploring reports, semantic models, and business insights directly within Power BI.
- Microsoft 365 Copilot provides a broader entry point, allowing users to discover and interact with Power BI insights alongside their everyday Microsoft 365 workflows.
The integration also works in the other direction: Power BI semantic models can provide a trusted, governed semantic foundation for AI experiences accessed through Microsoft 365 Copilot.
This is why I believe Prep data for AI is becoming increasingly important.
By defining AI data schemas, verified answers, and AI instructions, we can improve how Copilot understands business terminology, measures, and relationships, regardless of where users interact with their data.
My recommendation would be to focus on preparing and governing semantic models for AI rather than choosing between Copilot in Power BI and Microsoft 365 Copilot.
Ultimately, the goal is to make trusted Power BI insights accessible wherever users work, while maintaining a consistent semantic foundation.
Hope this helps!
BBFThanks for quick response!
Related to
3. Deployment pipelines and workspace differences
- When large model size is active all headings is active in Service, but what I noticed I do not see verified answers in this workspace as I do in the other workspace I deployed to.
(Can see "Simplify the data schema" and "Add AI instructions" data)
Both workspaces is in a F256 capacity and I am admin on both workspaces.
Verified answers not visible in the service on workspace1And in the other workspace I can see them.
- Copilot activated on workspace level? I guess you mean semantic model level or?- BeaBF
Hi TobNyg,Thanks for the additional details!
Since both workspaces are running on F256 capacities and you have Admin permissions on both, we can rule out the capacity SKU and workspace permissions as the likely causes.
One important point I found in the official documentation is that Verified Answers currently don't support CI/CD capabilities, including deployment pipelines. This might explain the inconsistent behavior you're experiencing between the two workspaces.
You can find this limitation here:
https://learn.microsoft.com/en-us/power-bi/create-reports/copilot-prepare-data-ai-verified-answersRegarding Copilot activation, I was referring to the tenant/capacity-level settings, rather than a setting on the semantic model itself. Copilot needs to be enabled through the appropriate tenant or delegated capacity settings, and the workspace must be assigned to a supported capacity.
A couple of things I'd suggest checking:
- Is Power BI Q&A enabled on the semantic model in both workspaces?
- If you create a new Verified Answer directly in the source workspace, does the option become available?
- Were the Verified Answers originally created in Desktop, directly in the Service, or before deployment?
These checks could help us understand whether the issue is related to the deployment pipeline limitation or to something specific to the source semantic model.
Hope this helps!
BBF
