Microsoft 365 Certified: Copilot and Agent Administration Fundamentals (AB-900) practice questions

A governance lead asks how Microsoft 365 Copilot decides which documents it can use when answering a prompt. Which statement correctly describes the access model Copilot applies?

• ACopilot uses the signed-in user's existing permissions and can only reference content that the user already has rights to open.check_circle Correct
• BCopilot is granted a dedicated service identity with tenant-wide read access so it can reason over every document in the organisation.
• CCopilot can read any file whose sensitivity label is lower than the user's clearance level, regardless of sharing permissions.
• DCopilot indexes content under an administrator identity, then filters results to the user only at display time.

Why A is correct: Correct. Copilot has no access model of its own; it queries content through Microsoft Graph in the user's security context, so it surfaces only what that user could already retrieve.

Why B is wrong: Tempting because Copilot can seem to 'know everything', but it holds no elevated tenant-wide identity; it always operates within the requesting user's permissions.

Why C is wrong: Sensitivity labels influence protection and usage rights, but access is decided by the user's file permissions, not by a clearance comparison, so this is wrong.

Why D is wrong: This sounds plausible given central indexing, but retrieval is trimmed by permission in the user's context during the query, not by a late display-time filter over admin-read data.

An organisation runs a workload where employees only occasionally use a custom agent, and it wants to be charged for the consumption those interactions generate rather than committing to a fixed per-user subscription. Which Microsoft 365 Copilot commercial model matches this requirement?

• AThe Microsoft 365 Copilot per-user monthly licence, which grants each assigned user unlimited use for a fixed recurring fee
• BA Microsoft Entra ID P2 subscription, which includes Copilot agent usage as part of its identity governance features
• CPay-as-you-go billing, which meters agent consumption and charges the organisation for the messages actually usedcheck_circle Correct
• DA Microsoft 365 E5 add-on that converts unused Copilot licences into consumption credits automatically

Why A is wrong: This is a real model and is tempting because it is the headline Copilot offer, but it charges a flat fee per assigned user regardless of usage, which is the opposite of paying only for occasional consumption.

Why B is wrong: Entra ID P2 covers identity protection and governance such as Privileged Identity Management; it does not include or meter Copilot agent consumption, so it cannot fulfil the billing requirement.

Why C is correct: Pay-as-you-go bills on metered consumption of agent messages, so an organisation with occasional use pays for what it consumes rather than a fixed per-user fee.

Why D is wrong: This sounds plausible because E5 bundles many add-ons, but no such automatic licence-to-credit conversion add-on exists, making it an invented mechanism.

An administrator wants to understand the role of the semantic index in Microsoft 365 Copilot. Which description best captures what the semantic index provides?

• AA separate copy of all tenant files moved into the Copilot service so responses can be generated without touching Microsoft 365 data.
• BA conceptual, meaning-based map of tenant content and interactions that helps Copilot find relevant grounding data while respecting each user's permissions.check_circle Correct
• CA keyword lookup table that matches prompt terms to file names and returns any document containing the exact words.
• DA compliance log that records which prompts each user submitted so that auditors can review Copilot activity.

Why A is wrong: Tempting as an efficiency story, but the semantic index does not export a parallel data store outside the tenant boundary; grounding data stays within Microsoft 365.

Why B is correct: Correct. The semantic index represents content and relationships by meaning rather than exact keywords, improving grounding relevance while retrieval stays permission-trimmed per user.

Why C is wrong: This describes classic keyword search; the semantic index is meaning-based and captures relationships, so a plain keyword table understates what it does.

Why D is wrong: Audit and activity records exist elsewhere in Microsoft Purview; the semantic index is a relevance structure for grounding, not an audit log.