Microsoft announced Copilot in Database Hub in public preview on November 11, 2026. The Microsoft Fabric experience answers natural-language questions using the inventory, issues, and telemetry already available in Database Hub. Microsoft announcement.
My proposed pilot would measure whether an engineer can move from a question to verified resource evidence more efficiently. Record the expected investigation outcome and the information the reviewer must confirm before any operational change.
Establish What the Answer Can See
Microsoft says Copilot works within the user's existing Azure and Fabric permissions and does not change databases. Its role is to summarize, explain, and prioritize the information available in the hub. Permissions and action boundary.
I would choose a small, approved set of resources whose inventory and known issues the pilot team can independently review. Preserve the resource identifiers and observation times alongside the question used in each test.
Ask the reviewer to distinguish the visible estate from the application's complete dependency picture. Record any missing resource, unclear ownership, or stale observation as a limitation of the investigation rather than silently treating the summary as complete.
Rehearse One Investigation End to End
The announcement describes following answers to affected resources and performance charts, including review of security issues, performance symptoms, and existing migration assessments. Investigation workflow.
For a hypothetical slowdown, I would define the resource and time window, ask the pilot question, and preserve the resulting explanation. The reviewer would then check the underlying observations and document which parts support the proposed interpretation.
Record competing explanations that remain plausible. A useful pilot result shows how the engineer investigated uncertainty and which evidence would be needed before proposing a configuration change or escalating to another team.
Review Enablement as a Tenant Decision
Copilot in Database Hub is opt-in. Microsoft identifies required tenant settings covering Azure OpenAI-powered experiences, OpenAI as a Microsoft subprocessor, and processing beyond the capacity's geographic or compliance boundary; settings can target security groups. Enablement requirements.
I would have the Fabric administrator document the proposed pilot group, current settings, intended scope, and the decision owner. Retain the rationale with the evaluation so the later reviewer can establish which users and experiences were included.
Microsoft's subprocessor guidance describes processing and compliance considerations and says OpenAI-operated models are currently unavailable in government and sovereign clouds. Fabric subprocessor considerations.
Review that documented boundary against the organization's requirements before enabling the pilot. Give any unresolved data-processing question a named owner and keep its outcome attached to the tenant decision.
Measure the Cost of the Complete Investigation
The announcement lists paid Fabric F2 or higher, or Power BI Premium P1 or higher, and says token-based Copilot usage consumes Fabric capacity units. Capacity and usage requirements.
My measurement would pair the relevant usage observation with engineer time and the quality of the verified conclusion. Compare equivalent investigation scenarios and record the observation window used for each.
Include the effort needed to identify an incomplete or incorrect answer. The pilot owner should be able to explain whether the experience helped the operator reach a defensible conclusion within the team's time and capacity expectations.
Preserve the Human Decision Record
I would keep the pilot question, answer, checked evidence, remaining uncertainty, and final recommendation together. Record the person responsible for accepting the conclusion and the existing process used for any later database change.
Ask another engineer to review a sample without coaching from the original investigator. Use their questions to improve the evidence record and decide whether the proposed investigation workflow is ready for wider use.
Practical Cloud Engineer Takeaway
Define a small resource scope and independently known investigation cases.
Check linked observations before accepting the summary's interpretation.
Review tenant settings and subprocessor requirements before enablement.
Compare capacity usage, operator effort, and verified answer quality.
Keep operational changes within the established review process.
Who Should Care?
Azure database administrators, Fabric administrators, estate operations teams, and engineers evaluating AI assistance for database investigations.
Bottom Line
The preview adds a conversational route into Database Hub's existing information. A useful adoption decision demonstrates reliable evidence review, appropriate tenant scope, and a measurable benefit for the engineers responsible for the estate.
No Fabric tenant setting, Copilot access, database, capacity allocation, or Azure resource was changed for this article.
Sources
Stay radical, stay curious, and keep pushing the boundaries of what is possible in the cloud.
Chriz
Beyond Cloud with Chriz
Comments