Microsoft 365 and AI Services Administrator Associate (AB-650)

A is correct.

Explanation: An AI Administrator or Global Administrator can reassign an ownerless or active shared Copilot Studio agent using the Assign new owner action. Starting or stopping a Foundry agent is a separate Azure action that provisions or deallocates compute and strictly requires the Azure AI Owner role.

B is incorrect: Reassignment is not supported through this route for Foundry agents, and reassigning an owner does not stop the underlying compute.

C is incorrect: Add owner is an Agent Builder action for multi-owner continuity, and the AI Administrator role does not grant the Foundry start or stop action.

D is incorrect: Deletion is documented specifically for Agent Builder agents, not Copilot Studio. The Foundry action is tied to Azure AI Owner, not Global Administrator.

C is correct.

Explanation: For a low-adoption observation, you must state the metric and data-through time, name at least two plausible explanations, and request evidence that separates them. Then select one scoped action and define the follow-up metric, cohort, and observation window. An organizational message suits only a diagnosed awareness gap.

A is incorrect: The absence of a posted issue means only that no matching issue is posted. It does not prove end-to-end Copilot health.

B is incorrect: An organizational message raises awareness, but it does not fix a license, incident, report-latency, workflow, or quality problem on its own.

D is incorrect: The three pillars organize governance and are not a single status screen. Changing them does not diagnose the cause of low adoption.

B is correct.

Explanation: The Uses eligible update channel value is the authoritative report value for the observed user and device. When the value is No, you must corroborate the device in the Microsoft 365 Apps admin center Inventory before you plan a managed update. Activity columns and the candidate flag do not show channel eligibility.

A is incorrect: The Suggested candidate flag only marks the top nonlicensed users by app usage, and it does not indicate update channel eligibility.

C is incorrect: The activity columns are adoption signals, not proof of channel eligibility; the eligibility column and Inventory provide the required evidence.

D is incorrect: The Copilot readiness report has no composite ready status, so waiting for one would never resolve the administrator's rollout decision.

C is correct.

Explanation: Blocking the server prevents invocation across supported client surfaces and serves as the durable retirement option today. Because no public block-propagation timing is stated for these clients, the administrator must refresh the Registry and verify the observed state instead of promising immediate enforcement to the team.

A is incorrect: Deleting a BYO server is not supported in the current lifecycle limits, so it cannot be used to retire the server.

B is incorrect: Republishing a new version of a BYO server is not supported, so the administrator cannot create an inactive version.

D is incorrect: The Agent 365 service principal is a registration prerequisite. Removing it is not a documented or supported way to retire a specific server.

B is correct.

Explanation: Approval records the registration decision only. Microsoft Entra consent is the separate authorization step that makes an approved server available to supported agent-building surfaces. After approval and consent, Copilot Studio discovery can take up to 30 minutes across all environments, so the administrator must grant the exact consent first.

A is incorrect: Approval is already recorded, and repeating it does not grant the separate Microsoft Entra consent that the server requires for client discovery.

C is incorrect: Unblocking restores availability for an existing blocked tool, but this server was never blocked in the scenario and still requires consent.

D is incorrect: Republishing a new version of a BYO server is not supported, and it would not provide the missing Microsoft Entra consent.

B is correct.

Explanation: The declared lookup capability might be proportionate, but required trust evidence is missing. This requires a Conditional review classification. An administrator cannot approve the request or grant consent until the endpoint owner, rotation owner, and data boundary are provided, so return the request for evidence.

A is incorrect: A reasonable operation does not replace missing ownership, rotation, data boundary, and security evidence, so an Accept decision is not supported.

C is incorrect: The backend uses API-key authentication, and registered-tool approval and required Microsoft Entra consent remain separate checks, so Reject is incorrect.

D is incorrect: Approving the request now would approve it based on incomplete evidence. The approval decision must wait until the required evidence is complete.

D is correct.

Explanation: The third policy state turns web search off in Work mode, Researcher, and Cowork, while Web mode and Chat keep web search available. The admin policy sets the ceiling, and the user toggle can only lower the setting, so the user can successfully use web search in Chat within that limit.

A is incorrect: Disabling web search in both Work and Web modes turns off Chat, but this specific policy state keeps Chat web search available.

B is incorrect: Changing the toggle later does not disable Chat. The toggle can lower access but cannot override the established admin policy ceiling.

C is incorrect: The user toggle never overrides the policy. It can turn search off but cannot enable it beyond what the admin policy allows.

C is correct.

Explanation: Consumption is almost never distributed evenly, and heavy use is not automatically a cost problem. Checking the output of heavy users shows whether they produce proportionate value. If they do, their practices should be shared. If they consume heavily without corresponding value, targeted coaching or a scoped limit is appropriate.

A is incorrect: A lower limit applied before checking value can punish proficient users and rarely reaches the actual inefficiency causing the high consumption.

B is incorrect: A tenant-wide limit is a blanket restriction that affects everyone and does not distinguish proportionate value from wasteful use.

D is incorrect: Retiring agents applies to deployed agents that are unused or no longer deliver value. It does not address heavy user consumption patterns.

C is correct.

Explanation: A failed or missing check is a no-go. The team must stop assigning licenses to the next group, hold the rollout at its current population, and return to validate until the evidence closes the gap. The denied-user test failing shows the control does not match the intended audience.

A is incorrect: Positive sentiment alone does not pass the gate, and the access test result must match the intended audience before expansion.

B is incorrect: Expansion is allowed only after validation meets its thresholds, so licensing the next group first skips the required validation gate.

D is incorrect: A pilot supports conclusions only about its selected users and content, not tenant-wide access correctness for the entire organization.

A is correct.

Explanation: The Microsoft Entra owner is the technical administrator for the object. Owners can edit authentication properties and manage blueprint credentials when they own that object. The sponsor stays accountable for business purpose and lifecycle decisions, so the two roles remain separate and distinct.

B is incorrect: The Agent Registry owner manages metadata and settings in the Microsoft 365 admin center. It does not govern the underlying Microsoft Entra identity credentials.

C is incorrect: Sponsors cannot edit authentication settings or manage credentials. Adding another sponsor provides business accountability, but it does not grant the required technical administration permissions.

D is incorrect: Role-assignable groups are not supported as sponsors for agent identities or blueprints. Furthermore, a sponsor relationship does not grant the permissions to manage credentials.