Sample questions

AB-900 practice questions, with full explanations

75 sample AB-900 (Microsoft 365 Copilot and Agent Administration Fundamentals) questions, each with the correct answer, a breakdown of why every other option is wrong, a memory hook, and the Microsoft Learn reference. Work the questions by topic; each topic names the book behind it.

Exam status: Microsoft's AB-900 study guide lists the skills measured as of October 14, 2026, and no dated change. Checked October 31, 2026. AB-900 on Microsoft Learn

The status board for every exam

The books

The AB-900 bookshelf

Going deeper on one product: 18 field guides, one product each.

Each has a sample on its Amazon page.

Work the 75 questions, one topic at a time

Each topic opens to its questions with the full explanation, the memory hook and the Microsoft Learn page behind it.

Cover of AB-900 Microsoft 365 Copilot and Agent Administration Fundamentals Complete Exam Guide
Topic 1 of 3

Identify the core features and objects of Microsoft 365 services (25 questions)

Go deeper in AB-900 Microsoft 365 Copilot and Agent Administration Fundamentals Complete Exam Guide. Get it on Amazon

Show the 25 questions and explanations

Litware assigns Microsoft 365 E3 to the security group All-Staff on the Licenses page of the Microsoft 365 admin center. All-Staff contains 200 users and a nested security group, Contractors, that holds 30 users. The 30 contractors never receive E3. How can a License Administrator license them through group-based licensing?

Correct answer: C. Assign Microsoft 365 E3 to the Contractors group as well

The Assign or unassign licenses to a group article states that group-based licensing doesn't support nested groups: when a licensed group contains another group, only users in the first-level group are assigned licenses. The Contractors members are outside the scope of the All-Staff assignment, so the fix is to assign E3 to Contractors itself. Its 30 users then become first-level members of a licensed group and get E3, as will anyone added to Contractors later.

Why the other options are wrong:

  • A. Assign Microsoft 365 E3 directly to each of the 30 contractor accounts Direct assignment would license the 30 users, but it isn't group-based licensing, and each new contractor would need the same manual step.
  • B. Wait for background processing to reach the nested group's members Waiting doesn't help, because group-based licensing never extends a group's licenses to members of a nested group.
  • D. Select Reprocess for All-Staff on the E3 product's Errors & issues tab Reprocess retries assignments that failed with an error, but the contractors never failed; they're outside the group's scope, so reprocessing changes nothing.

Memory hook: Group-based licensing reaches first-level members only: license the nested group itself.

Microsoft Learn: Manage group licenses

A user at Litware signs in to Microsoft 365 by plugging in a FIDO2 security key and touching its fingerprint sensor. She doesn't type a password at any point. Which two authentication factor types does this sign-in combine?

Correct answer: D. Something you have and something you are

Learn's How it works: Microsoft Entra multifactor authentication page defines MFA as two or more of something you know, something you have and something you are, naming a hardware key as something you have and a fingerprint as something you are. The security key holds the private key and the fingerprint unlocks it, so this passwordless sign-in still counts as MFA; Learn's passkeys page says passkeys serve as MFA when combined with biometrics or a PIN.

Why the other options are wrong:

  • A. Something you know and something you have Something you know would be a password or PIN, and she entered neither; her fingerprint took the place of a knowledge factor.
  • B. Something you are, used as a single factor Skipping the password makes the sign-in look single-factor, but the fingerprint only unlocks a private key held on the physical key, which adds possession.
  • C. Something you know and something you are No password or PIN was checked, so nothing she knows was used, and this pairing leaves out the security key she physically holds.

Memory hook: Key in hand is have, finger on the sensor is are, and no password is needed for MFA.

Microsoft Learn: Concept MFA howitworks

Fabrikam's CISO suspects employees keep company files in online storage services that IT never approved. She wants to upload traffic logs from the company's firewalls and proxies, see which SaaS services are in use and by how many users, and review a risk score for each service. Which product should she use?

Correct answer: A. Microsoft Defender for Cloud Apps

Cloud Discovery in Defender for Cloud Apps analyzes traffic logs, uploaded manually from firewalls and proxies as a snapshot report or forwarded continuously, against a catalog of over 31,000 cloud apps. Each app is ranked and scored on more than 90 risk factors, giving visibility into shadow IT, the users behind it and its risk, per the Cloud app discovery overview. That covers each part of the CISO's request.

Why the other options are wrong:

  • B. Microsoft Entra ID Protection ID Protection evaluates sign-in and user risk in Microsoft Entra ID. It doesn't discover which online services employees reach across the network.
  • C. Microsoft Defender Vulnerability Management Defender Vulnerability Management assesses vulnerabilities and misconfigurations on your assets. It doesn't analyze firewall logs to find unapproved SaaS services.
  • D. Microsoft Defender for Endpoint Defender for Endpoint can feed cloud discovery from managed devices, but it doesn't take firewall or proxy log uploads or score SaaS apps. That analysis happens in Defender for Cloud Apps.

Memory hook: Shadow IT from firewall and proxy logs, with a risk score per app: Defender for Cloud Apps.

Microsoft Learn: Set up cloud discovery

In Fabrikam's Identity Secure Score, the recommendation to protect all users with MFA is worth up to 10.71 percent. An administrator enforces MFA for the 50 members of the IT department, out of 500 users in the tenant, and makes no other change. How does Identity Secure Score credit this change?

Correct answer: C. About 1.07 percent at the next daily recalculation, as a share of users protected

Some recommendations score all or nothing, but this one is scored as a percentage of the total configuration. The What is Identity Secure Score FAQ uses this case: protecting all users with MFA is worth up to 10.71 percent, so protecting part of the user base earns that part of the maximum. Here 50 of 500 is 10 percent of 10.71, about 1.07 percent. The score is recalculated every 24 hours, so the credit shows after the next daily evaluation.

Why the other options are wrong:

  • A. About 1.07 percent, but only after the monthly recalculation at the start of next month Identity Secure Score is recalculated every 24 hours, not monthly, so the partial credit appears after the next daily evaluation.
  • B. No credit until all 500 users are protected, since recommendations score all or nothing Some recommendations are binary, but MFA coverage is scored as a share of users protected, so partial completion earns partial credit.
  • D. The full 10.71 percent at the next daily recalculation, since MFA is now enforced The full value needs every user protected, or a Resolved status set on the action. Enforcing MFA for a subset earns only the matching share.

Memory hook: MFA for a tenth of the users earns a tenth of the points, within a day.

Microsoft Learn: Concept identity secure score

Woodgrove Bank's users have Microsoft 365 E5 and use Microsoft Copilot Chat (formerly Microsoft 365 Copilot Chat) for web-grounded prompts at no additional charge. A director wants Copilot to reason across all the mail, meetings, chats and files she can access, grounded through Microsoft Graph, with priority access. What should the admin do?

Correct answer: B. Assign her a Microsoft Copilot add-on license, formerly Microsoft 365 Copilot

The Compare Microsoft Copilot in Microsoft 365 E3, E5, and E7 article lists Microsoft Copilot as an add-on for E3 and E5 and as included only in E7. The Copilot Chat overview shows Copilot Chat without Graph API, Work IQ and semantic indexing grounding and without priority access. Its organizational grounding is limited to content users provide, supported in-app experiences such as Outlook, and agents that have access to organizational content. The licensed Microsoft Copilot experience reasons across the mail, meetings, chats and files a user can access and adds priority access, so the director needs the add-on license.

Why the other options are wrong:

  • A. Assign her Microsoft Entra ID P2 so Copilot Chat can ground on Graph data Entra ID P2 adds identity features such as ID Protection and Privileged Identity Management, E5 already includes it, and it doesn't unlock Graph grounding for Copilot.
  • C. Turn on Microsoft Security Copilot, which E5 includes for eligible customers Security Copilot is a generative AI tool for security and IT analysis; it isn't the Microsoft Copilot license that grounds Copilot in a user's mail, meetings, chats and files.
  • D. Nothing more, because Microsoft 365 E5 already includes the Microsoft Copilot license E5 doesn't include Microsoft Copilot; Learn lists it as an add-on for E3 and E5 and as included only in Microsoft 365 E7.

Memory hook: Copilot Chat comes with the subscription; Graph-grounded Copilot needs the add-on, included only in E7.

Microsoft Learn: Compare Copilot e3 e5 e7

Tailspin Toys licenses its store associates with Microsoft 365 F1. An associate notices that an Exchange mailbox exists for her account and asks to use it for customer email. The admin must stay within the license terms. What should the admin do?

Correct answer: C. Assign her Microsoft 365 F3, since F1 carries no rights to use that mailbox

The Changing from a Microsoft 365 E plan to a Microsoft 365 F plan article explains that F1 includes the Exchange Kiosk plan only to enable the Teams calendar: a mailbox is provisioned, but F1 users aren't entitled to use it, and Microsoft recommends disabling Outlook on the web for them. F3 includes a 2 GB mailbox that users open in Outlook on the web, so assigning F3 gives her email within the license terms.

Why the other options are wrong:

  • A. Turn on the Exchange Online Kiosk service plan in her F1 assignment The Kiosk plan is what provisioned the mailbox she already sees, and under F1 it still carries no mailbox rights.
  • B. Let her use it in Outlook on the web, since F1 includes a 2 GB mailbox The 2 GB mailbox comes with F3; F1 provisions a mailbox only to support the Teams calendar and grants no right to use it for email.
  • D. Give her Full Access to a shared mailbox, which needs no license of its own A shared mailbox needs no license, but anyone who opens one needs their own licensed Exchange Online mailbox, which F1 doesn't provide.

Memory hook: F1 has no mailbox rights even when a mailbox exists; F3 adds a 2 GB mailbox.

Microsoft Learn: Switch from enterprise to frontline

Litware wants its Marketing staff to see their own logo and colors in the Microsoft 365 suite header, while everyone else keeps the organization's default theme. Marketing staff belong to a Microsoft 365 group, Marketing-Team, and to a mail-enabled security group, Marketing-Sec. What should a Global Administrator do?

Correct answer: C. In Custom themes, add a group theme and assign it to the Marketing-Team group

The Customize the Microsoft 365 theme for your organization article places themes under Settings, then Org settings, then the Organization profile tab, then Custom themes. Besides the default theme that applies to everyone, a Global Administrator can create up to four group themes and assign each to up to five Microsoft 365 groups. Security groups and distribution groups can't be used, so the group theme goes to Marketing-Team and everyone else keeps the default.

Why the other options are wrong:

  • A. In Custom themes, add a group theme and assign it to the Marketing-Sec group Group themes map only to Microsoft 365 groups; security groups and distribution groups can't be assigned, so Marketing-Sec won't work.
  • B. In Custom themes, edit the default theme to use Marketing's logo and colors The default theme applies to everyone in the organization, so editing it would rebrand the suite header for all users, not only Marketing.
  • D. In the SharePoint brand center, create a theme and apply it to Marketing's site Brand center themes are for SharePoint sites and experiences through Change the look; the suite header theme comes from Custom themes in Org settings.

Memory hook: Suite header themes: one default plus up to four group themes, each for Microsoft 365 groups only.

Microsoft Learn: Customize your organization theme

Tailspin Toys' security architect documents three controls. Conditional Access checks each sign-in's user, device health and location before granting access. Administrators activate their roles just in time, for a few hours at a time. The network is segmented so that a compromised device can't reach every workload. Which mapping of these controls to the Zero Trust principles is correct?

Correct answer: A. Conditional Access: verify explicitly; just-in-time roles: use least privilege access; segmentation: assume breach

Learn's Zero Trust identity and device access article says its identity and device access policies, built with Conditional Access, address verify explicitly: authenticate and authorize every request based on all available data points. Use least privilege access limits users with just-in-time and just-enough access, which is what time-boxed role activation does. Assume breach means minimizing blast radius and segmenting access, so network segmentation belongs under it.

Why the other options are wrong:

  • B. Conditional Access: verify explicitly; just-in-time roles: assume breach; segmentation: use least privilege access Conditional Access is placed correctly, but the other two are swapped. Just-in-time activation narrows privilege, and segmentation limits blast radius, which Learn files under assume breach.
  • C. Conditional Access: assume breach; just-in-time roles: use least privilege access; segmentation: verify explicitly This reverses the first and last controls. Conditional Access checks a request before granting access rather than containing an attacker, and segmentation contains damage rather than verifying a request.
  • D. Conditional Access: use least privilege access; just-in-time roles: verify explicitly; segmentation: assume breach This swaps the first two controls. Conditional Access weighs signals before granting access, which is verify explicitly, while time-boxed role activation limits how long privilege exists, which is least privilege.

Memory hook: Check the request: verify explicitly. Shrink the access: least privilege. Contain the attacker: assume breach.

Microsoft Learn: Zero trust identity device access policies overview

Twelve days ago, a License Administrator at Adatum removed Microsoft 365 E3 from the wrong user. Her account still exists, but her mailbox no longer opens in Outlook. What should the admin do to bring back the mailbox with its existing mail?

Correct answer: C. Reassign Microsoft 365 E3 to her with the Exchange Online service plan on

The Delete or restore user mailboxes in Exchange Online article says that when an Exchange Online license is removed, the mailbox data is held for 30 days, and adding the license back during that grace period restores access and makes the mailbox fully active again. At day 12, reassigning E3 with Exchange Online turned on brings back the mailbox and its mail. After 30 days, the data is deleted and can't be recovered unless retention holds it.

Why the other options are wrong:

  • A. Export her mailbox with Content search, then import it into a new mailbox Once the license is removed, the mailbox is no longer searchable with eDiscovery tools such as Content search, so this export can't run.
  • B. Restore her account from the Deleted users page in the admin center The account was never deleted, so it isn't on the Deleted users page; only the license was removed.
  • D. Wait, since the mailbox reconnects when the 30-day grace period ends The grace period works the other way: when 30 days pass without a license, Exchange Online permanently deletes the mailbox data.

Memory hook: License removed: reassign it within 30 days and the mailbox comes back intact.

Microsoft Learn: Delete or restore mailboxes

A network engineer at Litware proposes skipping MFA and device checks for every sign-in that comes from the head office network, because that network sits behind the corporate firewall. The CISO rejects the proposal as contrary to Zero Trust. Which approach matches Zero Trust guidance for sign-ins from the head office?

Correct answer: D. Treat head office sign-ins as untrusted and check identity and device signals on each request

Learn's Zero Trust overview says the model replaces trust in network boundaries by treating every access request as untrusted regardless of origin and deciding on real-time context. A request from the head office is still checked on identity and device signals; location can be one signal, but it never earns trust by itself. The other approaches all grant trust because of where a request comes from, which is the perimeter thinking Zero Trust replaces.

Why the other options are wrong:

  • A. Block sign-ins from outside the head office and trust every request that comes from inside it Blocking outside access feels strict, yet it trusts every request inside the firewall, and Learn notes that modern attacks such as phishing and session hijacking don't depend on network location.
  • B. Mark the head office IP ranges as a trusted named location and exclude it from the MFA policy Conditional Access can exclude trusted locations from an MFA policy, so this looks like a supported design, but it grants trust by network origin, the assumption the Zero Trust overview rejects.
  • C. Allow head office sign-ins without checks and review their sign-in logs every week for misuse Weekly log review sounds like assume breach, but office requests still get in unchecked, and detection after the fact doesn't replace verifying each request.

Memory hook: Inside or outside the firewall, every request is checked: Zero Trust grants nothing for location alone.

Microsoft Learn: Zero trust overview

Tailspin Toys' warehouse staff are standard users whose only MFA method is a text message code delivered by Microsoft. In September 2026, an admin set the temporary opt-out from automatic passkey enablement, and Tailspin changes nothing else. What happens the first time one of these users has to complete MFA after February 1, 2027?

Correct answer: A. The user must register a passkey during that sign-in before continuing

Learn's article on passkeys by default and the retirement of Microsoft-provided SMS and voice authentication says Microsoft-provided SMS and voice delivery retires on February 1, 2027 for all users except Global Administrators and external users. After that date, an in-scope user whose only MFA method is SMS or voice gets a blocking prompt to register a passkey during sign-in. The temporary opt-out covered only the September 1, 2026 to February 1, 2027 changes, and there's no opt-out from enforcement.

Why the other options are wrong:

  • B. The user is locked out until the helpdesk issues a Temporary Access Pass Learn's FAQ says users aren't locked out on the retirement date; they get a blocking passkey registration prompt, so no helpdesk pass is needed to continue.
  • C. The user keeps getting text message codes because the opt-out still applies The opt-out delayed automatic passkey enablement and the registration campaign only until February 1, 2027, and Learn states there's no opt-out from enforcement.
  • D. The user keeps getting text message codes until July 1, 2027 July 1, 2027 is the later date for Global Administrators and external users only; standard warehouse staff fall under the February 1, 2027 retirement.

Memory hook: SMS-only after February 1, 2027 means register a passkey to get in; Global Administrators and external users get until July 1.

Microsoft Learn: Concept SMS voice retirement

Tailspin Toys has Microsoft 365 Business Premium. Before turning on a Conditional Access policy that requires MFA for all users, the identity administrator wants a per-user list showing who is already capable of MFA, so the help desk can contact everyone who isn't before enforcement begins. Which tool in the Microsoft Entra admin center provides this list?

Correct answer: C. Authentication methods activity, on its Registration tab

Authentication methods activity (Entra ID, then Authentication methods, then Activity) has a Registration tab that counts users capable of MFA, passwordless sign-in and self-service password reset, and its user registration details list marks each user MFA Capable or Not Capable. The report needs Microsoft Entra ID P1 or P2, and Business Premium includes P1. The data can lag by up to 36 hours, so pull the list a day or two before the policy is turned on.

Why the other options are wrong:

  • A. Sign-in logs, filtered to sign-ins that required MFA Sign-in logs show only sign-ins that happened. A user who hasn't been prompted for MFA leaves no record of whether a method is registered, so the list would be incomplete.
  • B. Risky users, in Microsoft Entra ID Protection Risky users lists accounts that ID Protection flagged for risk. Risk says nothing about whether a user has registered a method that can satisfy MFA.
  • D. The Conditional Access What If tool, run once for each user What If shows which policies would apply to a simulated sign-in. It doesn't report whether a user has registered an MFA method.

Memory hook: Who is ready for MFA? Authentication methods, then Activity, then Registration.

Microsoft Learn: Authentication methods activity

In the Microsoft Defender portal, a Contoso analyst opens one item that brings together related alerts from Defender for Endpoint, Defender for Office 365 and Defender for Identity, with a timeline of those alerts, the impacted users, devices and mailboxes, and a summary of the whole attack. What kind of object is she viewing?

Correct answer: C. An incident

The Defender portal's two main signal types are alerts and incidents. An alert flags one malicious or suspicious event. An incident is a container that the correlation engine builds from related alerts, even across products, to tell the full attack story: timelines of alerts, involved users, devices and other resources, evidence, logs of automated investigation and response, and a textual summary. The Incidents and alerts page in Defender XDR draws this line.

Why the other options are wrong:

  • A. A custom detection rule A custom detection rule is a saved hunting query that runs automatically and can raise alerts. It isn't the grouped view of an attack.
  • B. An automated investigation An automated investigation is a process that an incident can start to reach verdicts on evidence and remediate threats. Its logs appear inside the incident, which is the container of the related alerts.
  • D. An alert An alert is one signal of a malicious or suspicious event. Related alerts from several products are gathered into an incident.

Memory hook: Alert: one signal. Incident: the attack story built from related alerts.

Microsoft Learn: Incidents overview

Contoso's on-premises Active Directory domain synchronizes to Microsoft Entra ID. The security team wants alerts for reconnaissance against Active Directory, Golden Ticket activity and attempts to move laterally toward domain administrator accounts. Which Defender product provides these detections?

Correct answer: D. Microsoft Defender for Identity

Defender for Identity monitors identity signals from on-premises Active Directory, through sensors on identity infrastructure such as domain controllers, and from Microsoft Entra ID and other identity providers. Its detections follow the attack stages of reconnaissance, compromised credentials, lateral movement and domain dominance, which includes Golden Ticket activity, per the Defender for Identity overview. Its alerts then join correlated incidents in Defender XDR.

Why the other options are wrong:

  • A. Microsoft Defender for Cloud Apps Defender for Cloud Apps covers SaaS apps, shadow IT discovery and OAuth apps, not on-premises Active Directory.
  • B. Microsoft Defender for Endpoint Defender for Endpoint protects and investigates devices. The Active Directory attack-stage detections described here come from Defender for Identity.
  • C. Microsoft Entra ID Protection ID Protection scores user risk and sign-in risk for Microsoft Entra sign-ins. It doesn't watch domain controllers for Golden Ticket activity or on-premises reconnaissance.

Memory hook: Attacks on on-premises Active Directory: Defender for Identity, fed by sensors on the domain controllers.

Microsoft Learn: What is

Woodgrove Bank hires an external auditor to review configuration settings in the Microsoft 365 admin center, the Exchange admin center and the SharePoint admin center for two weeks. The auditor must be able to see the settings but must not be able to change any of them. Which Microsoft Entra role should the auditor be assigned?

Correct answer: C. Global Reader

Learn's Microsoft Entra built-in roles reference describes Global Reader as the read-only counterpart to Global Administrator: it reads settings and administrative information across Microsoft 365 services but can't take management actions, and Learn recommends it for planning, audits and investigations. It works in the Microsoft 365, Exchange and SharePoint admin centers, so it covers every place the auditor needs to look.

Why the other options are wrong:

  • A. Security Reader Security Reader is read-only too, which makes it tempting, but it reads security features such as the Defender portal and ID Protection, not the general settings in the Microsoft 365, Exchange and SharePoint admin centers that the auditor must review.
  • B. Global Administrator Global Administrator would show every setting, even for a two-week assignment, but it also lets the auditor change them, which breaks the requirement and least privilege.
  • D. Reports Reader Reports Reader sees usage reports and sign-in and audit logs, and Learn states it can't access product admin centers like Exchange, so the auditor couldn't review those settings.

Memory hook: An auditor who looks but never touches gets Global Reader.

Microsoft Learn: Permissions reference

Tailspin Toys' service desk resets forgotten passwords for employees who hold no admin roles, and it does nothing else in Microsoft Entra ID. The security team wants the agents to have the least privileged built-in role for that task. Which Microsoft Entra role should the agents be assigned?

Correct answer: B. Password Administrator

Learn's Least privileged roles by task page names Password Administrator as the least privileged role to reset the password of non-admins and lists User Administrator only as an additional role. Helpdesk Administrator and User Administrator can also reset these passwords but carry more, such as invalidating refresh tokens or creating and deleting users, so assigning either one grants more than the task needs.

Why the other options are wrong:

  • A. Privileged Authentication Administrator Privileged Authentication Administrator is the role for resetting privileged admins' passwords; agents who only reset standard users' passwords don't need that reach.
  • C. User Administrator User Administrator can reset non-admin passwords, yet it also creates, deletes and updates users, which is far more than this task needs.
  • D. Helpdesk Administrator Helpdesk Administrator sounds made for a service desk and can reset these passwords, but it also resets passwords for more roles and invalidates refresh tokens, so it isn't the least privileged fit.

Memory hook: Standard user's password: Password Administrator. Privileged admin's password: Privileged Authentication Administrator.

Microsoft Learn: Delegate by task

Adatum's endpoint admin needs one group that holds 300 company-owned Windows laptops, so that an Intune device configuration profile and an app can be assigned to the devices themselves rather than to the people who use them. Which type of group should the admin create?

Correct answer: A. A security group

Security groups grant access to resources and can contain users or devices, and the Compare types of groups in Microsoft 365 article notes that a security group of devices can be used with mobile device management services such as Microsoft Intune. Its membership can be assigned, or dynamic from device attributes such as the operating system version, and devices in a dynamic device group need no Microsoft Entra ID P1 license.

Why the other options are wrong:

  • B. A distribution group Distribution groups exist to send email to a set group of people, such as everyone in one building; they aren't the object for assigning profiles and apps to devices.
  • C. A Microsoft 365 group Microsoft 365 groups are built for collaboration and their members can only be users, so they can't hold the laptops.
  • D. A mail-enabled security group A mail-enabled security group works like a security group for access and adds email, but it can't contain devices and can't use dynamic membership.

Memory hook: Devices go in security groups; Microsoft 365 groups and mail-enabled security groups hold no devices.

Microsoft Learn: Compare groups

Fabrikam's Europe service desk must reset passwords for Europe employees only, none of whom hold admin roles. An admin created an administrative unit named Europe, added the security group Europe-Staff, which contains every Europe employee, and assigned the desk the Helpdesk Administrator role scoped to the unit. The desk staff have Microsoft Entra ID P1 licenses, yet they can't reset any Europe employee's password. What should the admin do?

Correct answer: B. Add the Europe employees' user accounts directly to the administrative unit

Learn's Administrative units article explains that adding a group to an administrative unit brings the group into the unit's management scope but not the group's members. An admin scoped to the unit whose role covers groups can rename the group or change its membership, yet no scoped admin can reset the members' passwords until those users are added directly as members of the unit. Once they are, the scoped Helpdesk Administrator assignment covers their passwords and still reaches no one outside Europe.

Why the other options are wrong:

  • A. Assign a Microsoft Entra ID P1 license to every Europe employee in the group Licensing isn't the gap: Learn requires P1 for each administrative unit administrator, which the desk already has, and members need P1 only when the unit uses dynamic membership rules.
  • C. Reassign the Helpdesk Administrator role to the desk at the tenant scope Tenant scope would let the desk reset the Europe passwords, but also the password of every other non-admin user, which breaks the Europe-only requirement.
  • D. Assign the desk the Groups Administrator role scoped to the administrative unit Groups Administrator scoped to the unit manages the group object, such as its name and membership, and still can't reset the passwords of the group's members.

Memory hook: A group in an administrative unit brings the group, not its people.

Microsoft Learn: Administrative units

Fabrikam's legal director must fully manage the Legal communication site, including its site collection features, and must be able to open every item, even items whose permission inheritance was broken. She must not get access to the SharePoint admin center or to any other site. What should a SharePoint Administrator do?

Correct answer: C. Add her as a site admin of the Legal site on its Membership tab in Active sites

Site admins, formerly site collection administrators, have the same Full Control as a site owner plus rights such as managing search, the recycle bin and site collection features, and they can reach any item in the site even where inheritance is broken, per the Admin center site permissions reference. The role covers only that site and doesn't give access to the SharePoint admin center, which matches the director's scope.

Why the other options are wrong:

  • A. Assign her the SharePoint Administrator role in the Microsoft 365 admin center The SharePoint Administrator role opens the SharePoint admin center for the whole organization, which she must not have, and it doesn't even give automatic access to site content.
  • B. Add her to the Legal site's Members group from the site's Site permissions page Members get the Edit permission level, which can't manage permissions or site collection features and doesn't reach items whose inheritance was broken.
  • D. Add her to the Legal site's Owners group from the site's Site permissions page Owners get Full Control, but site admins hold extra rights such as site collection features and access to items even where inheritance was broken, which the director needs.

Memory hook: One site, every item, no admin center: make them a site admin.

Microsoft Learn: Site permissions

On Contoso's Policies communication site, a person in the site's Members group deleted an entire document library by mistake. The site owner wants members to keep adding, editing and deleting documents and list items but no longer create or delete lists and libraries, with no other rights added. Which permission level should the owner give the Members group in place of Edit?

Correct answer: B. Contribute

Learn's Permission levels in SharePoint article lists Edit, the Members group's default, as adding, editing and deleting lists as well as items, through the Manage Lists permission. Contribute keeps view, add, update and delete for list items and documents but leaves out Manage Lists, so members keep working on content but can no longer create or delete lists and libraries.

Why the other options are wrong:

  • A. Read Read only views pages and items and downloads documents, so members could no longer add, edit or delete documents.
  • C. Design Design sounds narrower than Edit, but it can create lists and document libraries, edit pages and apply themes, so members could still delete a library.
  • D. Limited Access SharePoint assigns Limited Access automatically when someone is given access to a single item; it can't be assigned to a user or group directly and grants no rights to content.

Memory hook: Edit equals Contribute plus Manage Lists; drop Manage Lists to protect libraries.

Microsoft Learn: Understanding permission levels

In Tailspin Toys' 40-member Product team, five people need a channel for pricing discussions that the rest of the team can't see. Dana, a team member who isn't an owner, will create it, and one of the five is a guest who already belongs to the team. They don't want another team to manage. What should Dana create?

Correct answer: D. A private channel in the Product team

Private channels are focused spaces for a subset of a team, visible only to their owners and members, as the Private channels in Microsoft Teams article describes. By default any team owner or team member can create one, and anyone who already belongs to the team, guests included, can be added. That fits Dana's role, the guest on the list and the wish to avoid managing another team.

Why the other options are wrong:

  • A. A shared channel in the Product team A shared channel would also hide the discussion from the rest of the team, but guests can't be added to a shared channel, so the guest among the five couldn't join it.
  • B. A standard channel in the Product team A standard channel is open to every member of the team, so the other 35 people would see the pricing discussions.
  • C. A new private team for the five people A separate team brings another team, group and site to manage, which the five people want to avoid; a private channel keeps them inside the Product team.

Memory hook: Subset of the team, guests allowed: private channel. People from outside the team: shared channel.

Microsoft Learn: Private channels

Contoso engineers must work in one Teams channel with eight engineers from Fabrikam, a partner that also uses Microsoft 365. Contoso doesn't want guest accounts created for them, and the Fabrikam engineers must not see the rest of Contoso's Engineering team. Both organizations have configured cross-tenant access settings in Microsoft Entra ID. What should an Engineering team owner create?

Correct answer: B. A shared channel, with the Fabrikam engineers added as external participants

Learn's Shared channels in Microsoft Teams article explains that people outside the organization take part in a shared channel through Microsoft Entra B2B direct connect, without guest accounts, once both organizations configure cross-tenant access. People can be added to a shared channel without being added to the team, so the Fabrikam engineers reach that channel only and never join Contoso's Engineering team.

Why the other options are wrong:

  • A. A standard channel, with the Fabrikam engineers invited to the team as guests Guests in a standard channel are members of the whole team, so they need guest accounts and would see every standard channel in it.
  • C. A private channel, with the Fabrikam engineers invited to the team as guests Private channel members must already belong to the team, so each Fabrikam engineer would need a guest account and would join Contoso's Engineering team.
  • D. A new project team, with the Fabrikam engineers invited to it as guests A new team still depends on guest accounts for the Fabrikam engineers, which Contoso doesn't want; only shared channels use B2B direct connect.

Memory hook: Partner organization, no guest accounts: shared channel through B2B direct connect.

Microsoft Learn: Shared channels

Adatum's Global (Org-wide default) Teams policy allows private channel creation. An admin assigns the Contractors security group a custom Teams policy that blocks it, at rank 1, and later assigns the Project Leads security group a custom Teams policy that allows it, at rank 2. Ravi is a direct member of both groups and has no Teams policy assigned to him directly. Which policy decides whether Ravi can create private channels?

Correct answer: C. The Contractors policy, because rank 1 is the highest group assignment ranking

For each policy type, a direct assignment wins; without one, a member of several groups inherits the policy from the group assignment with the highest ranking, where a smaller number ranks higher and 1 is highest; only with no direct or group assignment does the Global (Org-wide default) policy apply. That is the precedence in the Assign policies to users and groups article, so Ravi's effective policy is the rank 1 Contractors policy, which blocks private channels.

Why the other options are wrong:

  • A. The Global (Org-wide default) policy, because nothing is assigned to Ravi directly The Global (Org-wide default) policy applies only when a user has no direct assignment and isn't in any group with an assignment of that type; Ravi is in two such groups.
  • B. The Project Leads policy, because a rank 2 assignment outranks a rank 1 assignment Rankings run the other way: the smaller the number, the higher the ranking, with 1 the highest, so the rank 1 Contractors assignment wins.
  • D. The Project Leads policy, because it is the most recent group policy assignment The order in which the group assignments were made doesn't decide anything; the ranking set on each group assignment does.

Memory hook: Direct beats group, rank 1 beats rank 2, and Global applies only when nothing else does.

Microsoft Learn: Assign policies users and groups

Woodgrove Bank has 600 branch tellers, frontline workers who need one consistent set of Teams policies across several policy types. The admin wants to assign the whole set at once and later change a setting once so that every teller gets the update. Woodgrove has no Teams Premium licenses. What should the admin use in the Teams admin center?

Correct answer: B. The Frontline worker policy package, assigned to the tellers

Learn's Managing policy packages in Teams article describes a policy package as a collection of predefined policies and settings for users with similar roles, and Teams includes a Frontline worker package. Packages cover policy types such as messaging, meeting, app setup and calling, and when a policy in a package changes, everyone assigned the package gets the update. Learn ties the Teams Premium requirement to custom packages, one license per targeted user, so the built-in package fits Woodgrove.

Why the other options are wrong:

  • A. A custom messaging policy assigned to a security group of tellers A messaging policy covers only one policy type, so the tellers' other policy types would still need separate assignments and separate updates.
  • C. Edits to the Global (Org-wide default) policy of each policy type Global (Org-wide default) policies apply to everyone without another assignment, so every other employee would get the teller settings too.
  • D. A custom policy package that bundles the policies for the tellers Each user targeted by a custom policy package needs a Teams Premium license, and Woodgrove has none.

Memory hook: One role, several policy types, one assignment: a policy package (custom packages need Teams Premium).

Microsoft Learn: Manage policy packages

Tailspin Toys has an internally built web app that already uses the company's on-premises AD FS for single sign-on. The team will move the app's federation to Microsoft Entra ID next year, but for now it wants the app to appear as a tile in users' My Apps portal without changing how the app authenticates. Which single sign-on option should the admin configure?

Correct answer: C. Linked SSO, pointing the My Apps tile at the app's current sign-in URL

Linked SSO is the choice when an app is configured for SSO in another identity provider service: it sets the target location users reach from their portal, and the single sign-on deployment plan gives a custom web app that uses AD FS federation as its example. It doesn't sign users in with Microsoft Entra credentials, so the app keeps authenticating through AD FS, which suits a phased migration until full SSO moves to Entra ID.

Why the other options are wrong:

  • A. Header-based SSO, publishing the app through Application Proxy Header-based SSO is for on-premises apps that use headers for authentication, published through Application Proxy; it would change how this federated app signs users in.
  • B. Password-based SSO, storing each user's credentials for the app Password-based SSO is for apps with an HTML sign-in page and no federation support; this app already federates with AD FS.
  • D. SAML-based SSO, making Microsoft Entra ID the app's identity provider now SAML-based SSO would move the app's federation to Microsoft Entra ID now, the change the team has postponed to next year.

Memory hook: Federated elsewhere and only needs a tile: Linked SSO, which signs no one in.

Microsoft Learn: Plan SSO deployment

Topic 2 of 3

Understand data protection and governance tasks for Microsoft 365 and Copilot (29 questions)

Go deeper in AB-900 Microsoft 365 Copilot and Agent Administration Fundamentals Complete Exam Guide. Get it on Amazon

Show the 29 questions and explanations

Fabrikam has Microsoft 365 E5 and assigned Microsoft 365 Copilot licenses to 200 employees a month ago. They use Copilot in Word, Outlook and Teams on Windows 11 devices. In Microsoft Purview Data Security Posture Management (DSPM), the reports and activity explorer show no Copilot interactions. In the Audit solution, a banner offers to start recording user and admin activity. What should the compliance admin do?

Correct answer: D. Turn on auditing in the Microsoft Purview portal

The banner means auditing is off for the tenant. Learn's activity explorer events table for DSPM says the AI interaction event for Microsoft 365 Copilot and Copilot Chat requires auditing to be turned on, and turning on auditing is DSPM's initial setup task, required to see meaningful data. Auditing is on by default for enterprise tenants but can be turned off; turning it on can take up to 60 minutes, and DSPM needs about a day before data appears.

Why the other options are wrong:

  • A. Deploy the Microsoft Purview browser extension The browser extension is a prerequisite for discovering visits to third-party generative AI sites, not for Microsoft 365 Copilot interactions in Word, Outlook and Teams.
  • B. Onboard the Windows 11 devices to Microsoft Purview Device onboarding gives visibility into sensitive data shared with third-party AI sites and enables endpoint DLP; Microsoft 365 Copilot interactions don't depend on it.
  • C. Create the one-click collection policy for Copilot experiences That collection policy captures prompts and responses from Copilot in Fabric and Security Copilot; Microsoft 365 Copilot interactions rely on auditing, which is off here.

Memory hook: No audit log, no Copilot insight: DSPM reads Microsoft 365 Copilot activity through auditing.

Microsoft Learn: Data security posture management considerations

Contoso has Microsoft Defender for Office 365 Plan 2. The security team worries that attackers will send email containing hidden instructions meant to hijack Copilot when users ask it to summarize their inbox. They want a protection that checks inbound messages for those instructions during mail flow, before delivery, with no extra configuration. Which protection fits?

Correct answer: C. Prompt injection protection, which marks detections as High confidence phishing

Defender for Office 365 Plan 2 detects prompt injection content in inbound email as part of the same mail flow inspection that handles phishing and malware, so no extra configuration is needed. The Prompt injection protection page says detections get the High confidence phishing verdict with the detection technology value Prompt injection protection, which Threat Explorer and Advanced Hunting can filter on. This layer acts before delivery, ahead of Copilot's own runtime safeguards.

Why the other options are wrong:

  • A. Copilot's jailbreak and cross-prompt injection classifiers, applied at prompt time Copilot's jailbreak and cross-prompt injection attack (XPIA) classifiers are real safeguards, but they run inside Copilot at prompt time, after the message is already in the mailbox.
  • B. A DLP rule for the Copilot location that excludes email from external senders Excluding external email does reduce prompt injection risk, but that DLP rule is in preview, needs a policy, checks only the sender domain and acts when Copilot grounds a response, after the mail is delivered.
  • D. Safe Links, which scans each URL in the message and checks it again at click Safe Links does scan URLs in inbound mail before delivery and again at click, but it evaluates links, not message text, so instructions hidden in the body or markup aren't what it checks.

Memory hook: Injected instructions in mail: Defender for Office 365 Plan 2 screens for them in mail flow, no setup needed.

Microsoft Learn: Prompt injection protection Defender for Office 365

Tailspin Toys just assigned its first Microsoft Copilot licenses, and its new SharePoint Administrator doesn't know where to start on governance. She wants one action that checks every site for inactivity, missing owners, broken permission inheritance, sharing with Everyone except external users and risky sharing links, then recommends fixes. What should she use?

Correct answer: D. The content management assessment under Advanced Management

Learn's content management assessment article calls it the SAM hub for assessing content management: in the SharePoint admin center, the admin selects Advanced Management, then Start assessment. It runs the inactive sites (180 days), site ownership, broken inheritance, EEEU and sharing links reports across all sites, takes 2 to 72 hours depending on tenant size, recommends remediation, takes no action without the admin's consent, and can be rerun every 30 days.

Why the other options are wrong:

  • A. A change history report for the organization's sharing settings Change history reports show who changed settings and when over the last 180 days; they don't assess sites or recommend fixes.
  • B. A site ownership policy run in simulation mode across all sites A site ownership policy checks only whether sites meet a minimum owner or admin count, so it misses inactivity, inheritance and sharing problems.
  • C. The Site permissions for users report run for every site owner The users report lists the sites specific users can reach; it doesn't scan every site for inactivity, ownership or sharing risks.

Memory hook: Don't know where to start? Advanced Management, then Start assessment.

Microsoft Learn: Content management assessment

Woodgrove Bank's legal team is responding to a lawsuit. It must find a loan officer's Microsoft 365 Copilot prompts and responses from the past three months, make sure they can't be lost while the matter is open, and give copies to outside counsel. Which Microsoft Purview solution meets all three needs?

Correct answer: B. eDiscovery, with a case that covers the officer's mailbox

Copilot prompts and responses are stored in the user's mailbox, so an eDiscovery case can find them (for example, with a condition that selects Copilot activity), preserve them with a hold, and export them for counsel. Learn's Microsoft 365 Copilot guidance says to create an eDiscovery case when you need to preserve, collect, review or export these interactions.

Why the other options are wrong:

  • A. DSPM activity explorer, filtered to the officer's interactions DSPM shows AI interaction events, with prompt text for the right role, but it can't place a hold or produce an export package for counsel.
  • C. Communication Compliance, with a policy that covers Copilot Communication Compliance routes policy matches in prompts to reviewers for action; it isn't a preservation and export tool for litigation.
  • D. Audit, with a search of the officer's Copilot activity records Audit records when and how users interacted with Copilot, kept for the audit retention period; it can't place content on hold or package it for counsel.

Memory hook: Evidence means eDiscovery: search, hold and export work on Copilot prompts too.

Microsoft Learn: AI Microsoft 365 Copilot

Ravi holds the Microsoft Entra Compliance Administrator role at Fabrikam. He can open the Data explorer tab in the Microsoft Purview portal but sees no list of items. He needs to see each matching item and its location, but he must not be able to view the contents of the files. What should be added to his account?

Correct answer: A. Membership in the Data Explorer List viewer role group

Data explorer keeps three grants apart. Entra roles such as Compliance Administrator and Global Administrator open the tab only. The Data Explorer List viewer role group (data classification list viewer role) shows each item and its location, while the Data Explorer Content viewer role group shows file contents and item names. The two groups aren't cumulative, so List viewer alone meets Ravi's need. Older Learn pages call these groups Content Explorer List Viewer and Content Explorer Content Viewer.

Why the other options are wrong:

  • B. Membership in the Information Protection Readers role group Information Protection Readers gives view-only access to DLP and label reports; it doesn't include the data classification list viewer role, so the item list stays empty.
  • C. The Microsoft Entra Global Administrator role Global Administrator is one of the roles that opens the Data explorer tab but doesn't show the list of items or their contents, and it's far broader than the task needs.
  • D. Membership in the Data Explorer Content viewer role group Content viewer lets members view the contents of each item, which Ravi must not have; it's the grant to add only when reading files is required.

Memory hook: Admin role opens the tab, List viewer shows the list, Content viewer opens the file.

Microsoft Learn: Data classification data explorer

Contoso uses Microsoft Purview Data Security Posture Management (DSPM). Pay-as-you-go billing is set up, and the one-click policies that extend insights for data discovery are created. Reports show Microsoft 365 Copilot interactions, but nothing for employees who paste customer data into Google Gemini in Chrome on their Windows 11 devices. What must the admin do to get that visibility?

Correct answer: A. Onboard the Windows 11 devices to Microsoft Purview and deploy the Microsoft Purview browser extension

The DSPM setup tasks spell out the signal for third-party generative AI sites: devices must be onboarded to Microsoft Purview to show sensitive information shared with those sites, and the Microsoft Purview browser extension must be deployed to Windows users, which endpoint DLP also needs on Windows when people use Chrome. Copilot data already appears, so auditing and Copilot licensing are in place; the missing piece is the device and browser signal for Gemini in Chrome.

Why the other options are wrong:

  • B. Deploy an Edge configuration policy that activates the Microsoft Purview integration in the browser An Edge configuration policy activates the Purview integration in Microsoft Edge, but these employees use Chrome, which needs onboarded devices and the browser extension.
  • C. Assign Microsoft 365 Copilot licenses to the employees who paste data into Google Gemini A Microsoft 365 Copilot license lights up the Copilot and agents signal only; it gives no visibility into a third-party site such as Gemini.
  • D. Turn on Microsoft Purview Audit for the organization so that activity on AI sites starts being recorded Auditing is on by default and is clearly working here, because Copilot interactions already appear in the reports, so turning it on changes nothing for Gemini.

Memory hook: Third-party AI sites in DSPM: onboard the device, deploy the Purview browser extension.

Microsoft Learn: Data security posture management setup

Adatum has Microsoft 365 E3 and Microsoft 365 Copilot licenses but doesn't use Microsoft Purview Data Loss Prevention. The payroll team wants the files on its SharePoint site kept out of the tenant-level semantic index that Copilot uses. What should the admin do?

Correct answer: D. In the site settings, set Allow this site to appear in search results to No

The semantic indexing page says organizations without a DLP solution can exclude a SharePoint site from the tenant-level index by setting Allow this site to appear in search results to No, under Search and offline availability in the site settings. The exclusion covers Microsoft Search and the semantic index together; there's no option to drop the site from only one, so payroll content also disappears from organization-wide search results.

Why the other options are wrong:

  • A. On the Search and Intelligence page, turn off item insights for the tenant Item insights do feed recommendations in Copilot results, but turning them off only reduces people and item suggestions; it doesn't remove a site's files from the index.
  • B. Apply a sensitivity label to the site so the index skips the site's files Labels feel like the protection layer for sensitive files, but Learn says information protection offers no other way to exclude data from semantic indexing, and a site's label isn't inherited by its files.
  • C. In the Microsoft 365 admin center, turn off semantic indexing for the tenant An admin center switch sounds like the obvious control, but Microsoft enables semantic indexing automatically as an improvement to Microsoft 365 Search, and it can't be disabled.

Memory hook: Hide a site from search and it leaves the semantic index too; the two can't be split.

Microsoft Learn: Semantic index for Copilot

Adatum has Microsoft Copilot licenses assigned, and salary files from its HR site have started appearing in Copilot answers for employees outside HR. While HR owners review permissions this month, the SharePoint Administrator must keep the site out of organization-wide search and Copilot without changing anyone's access, and HR staff must still search within the site. Adatum never enabled Restricted SharePoint Search. What should the admin do?

Correct answer: A. Turn on Restricted Content Discovery on the HR site's Settings tab

Learn's Restricted Content Discovery article describes a site-level setting that keeps a site's content out of organization-wide search results and Copilot responses while permissions stay the same: users still open content they can reach, and searches that start inside the site are unaffected. It is a temporary control for reviews, turned on with Restrict content from Microsoft Copilot on the site's Settings tab. Learn points to it in place of Restricted SharePoint Search, whose new enablement is blocked starting July 31, 2026.

Why the other options are wrong:

  • B. Turn off Allow this site to appear in Search results for the HR site Learn notes that turning off this setting blocks the site's content from both organization-wide search and site search, so HR staff would lose search within the site.
  • C. Turn on Restricted SharePoint Search without the HR site on its allow list Restricted SharePoint Search is retiring, and Learn states that starting July 31, 2026, new enablement is blocked, so a tenant that never turned it on can't start now.
  • D. Apply restricted access control to the HR site with the HR security group Restricted access control changes who can open the site, denying everyone outside the HR group, which breaks the rule that no one's access changes.

Memory hook: Hide, don't block: Restricted Content Discovery changes discovery, never access.

Microsoft Learn: Restricted content discovery

Contoso publishes a Copilot agent for its sales staff. The agent labels its answers with "AI-generated content may be incorrect" and opens each conversation with a short note on what it can and can't do. Which responsible AI principle do these design choices most directly support?

Correct answer: C. Transparency

Microsoft's Responsible AI for agent design page defines transparency as making an AI system's capabilities and limitations clear, so users understand what the agent can do, how information is used and how outputs are generated. It names in-product disclosures, such as a label saying AI-generated content may be incorrect, as practical ways to implement transparency. Telling users up front what the agent can and can't do serves the same principle.

Why the other options are wrong:

  • A. Inclusiveness Inclusiveness covers accessibility, language variation and different levels of technical skill, none of which a disclosure label addresses.
  • B. Fairness Fairness asks for consistent, equitable outcomes without unfair treatment based on sensitive attributes, which a disclosure label doesn't change.
  • D. Accountability A disclosure can feel like taking responsibility, but accountability is about clear human oversight and the team answerable for the agent, not what the agent tells users.

Memory hook: Say it's AI and say what it can't do: transparency.

Microsoft Learn: Responsible AI

Fabrikam has Microsoft Copilot licenses assigned. Its chief legal officer, a member of the Legal-Team security group, keeps case files in her OneDrive. The SharePoint Administrator must ensure that only Legal-Team members can open content in that one OneDrive, even people who already hold a sharing link, while every other employee keeps full use of their own OneDrive. What should the admin use?

Correct answer: D. Restricted access control on her OneDrive with the Legal-Team group

Learn's article on restricting a user's OneDrive content says a site access restriction policy can limit one OneDrive to up to 10 Microsoft Entra security groups or Microsoft 365 groups; people outside those groups can't access the content even with earlier permissions or a sharing link. Site access restriction must be enabled for the organization first, and the owner must belong to one of the groups or loses access; the chief legal officer is in Legal-Team, so she keeps hers.

Why the other options are wrong:

  • A. The tenant's Restrict OneDrive access setting, scoped to Legal-Team That tenant-wide setting limits OneDrive itself to the listed security groups, so every employee outside Legal-Team would lose access to their own OneDrive.
  • B. Restricted Content Discovery applied to her OneDrive and its files Restricted Content Discovery applies only to SharePoint sites, not OneDrive, and it changes discovery in search and Copilot rather than who can open content.
  • C. A site access review that asks her to remove the old sharing links Site access reviews support only SharePoint sites, not OneDrive accounts, and a review asks an owner to clean up rather than enforcing who can open files.

Memory hook: Lock one OneDrive to a group with restricted access control, and keep the owner inside the group.

Microsoft Learn: OneDrive site access restriction

Elena is a member of the Data Explorer List viewer role group at Litware. In Data explorer she sees items that match the U.S. Bank Account Number sensitive information type and their locations, but the item names are hidden and she can't open the files. An investigation now requires her to open those unencrypted documents from within Data explorer. What should she be granted?

Correct answer: B. Membership in the Data Explorer Content viewer role group

The Data Explorer Content viewer role group (data classification content viewer role) lets members view the contents of each item, and the Data explorer page notes it's also required to see item names in list view, because names can contain sensitive data. Data explorer's role groups supersede permissions assigned locally on the items, so site permissions don't change what she can open there. Even with this role, preview and download are disabled for encrypted documents; these files open because they're unencrypted.

Why the other options are wrong:

  • A. Site member permissions on each SharePoint site that stores the files Data explorer permissions supersede the permissions assigned on the items themselves; site membership might let her open the files in SharePoint, but it doesn't unlock names or contents in Data explorer.
  • C. The Role Management role in the Microsoft Purview portal Role Management holders can assign data explorer role group membership, but the role doesn't itself show item names or contents.
  • D. The Microsoft Entra Compliance Data Administrator role Compliance Data Administrator is one of the roles that only opens the Data explorer tab; it grants neither the item list nor the contents.

Memory hook: Item names and contents both need Data Explorer Content viewer.

Microsoft Learn: Data classification data explorer

Tailspin Toys has Microsoft 365 E5. It onboarded its Windows devices to Microsoft Purview, deployed the Microsoft Purview browser extension and, two weeks ago, created the one-click policy DSPM for AI - Detect when users visit AI sites. Employees visit AI sites daily. Lena, in the Microsoft Purview Compliance Administrator role group, sees AI interaction events but no AI website visit events in Data Security Posture Management (DSPM) activity explorer. What should the admin do?

Correct answer: C. Add Lena to the Insider Risk Management Analysts role group

AI site visits are detected by an insider risk management policy, and DSPM's permissions page shows that every admin and view-only role, the Compliance Administrator role group and Global Administrator included, sees activity explorer events except the browse to URL (AI Visit) events from insider risk management. Those events need the Insider Risk Management Analysts or Investigators role group, so adding Lena to Analysts shows them.

Why the other options are wrong:

  • A. Assign Lena the Microsoft Entra Global Administrator role Global Administrator has the same gap as the Compliance Administrator role group: DSPM hides AI website visit events from both.
  • B. Assign Lena the Purview Data Security AI Content Viewer role The AI Content Viewer role adds prompt and response text to AI interaction events; it doesn't reveal the insider risk AI website visit events.
  • D. Add Lena to the Insider Risk Management Admins role group Lena's Compliance Administrator role group already carries the same insider risk permissions as Insider Risk Management Admins, and DSPM lists only Analysts or Investigators for AI site visit events.

Memory hook: AI site visits are insider risk data: only IRM Analysts or Investigators see them in DSPM.

Microsoft Learn: Data security posture management permissions

Litware's legal department must not receive Copilot or Copilot Chat responses grounded in web content, and legal users must not be able to turn web search back on. Every other department keeps web search. What should the Microsoft 365 admin configure?

Correct answer: C. The Allow web search in Copilot policy in Cloud Policy, scoped to legal users

The Allow web search in Copilot policy, available only in Cloud Policy service for Microsoft 365, turns web search on or off for users or groups across both Copilot and Copilot Chat. When the policy disables web search, the Web content toggle is turned off and dimmed, so legal users can't turn it back on, while users outside the policy's scope keep web search. The web search privacy page documents this policy and its three settings.

Why the other options are wrong:

  • A. A DLP policy with the Performing Web Searches action, scoped to legal users A DLP rule looks like an enforceable block, but that action stops web search only for prompts that contain the selected sensitive information types, so other legal prompts still use web content.
  • B. The Web content toggle in Copilot settings, turned off for each legal user Turning the toggle off does stop web content for that user, but it's the user's own preference, and anyone allowed web search can switch it back on.
  • D. The optional connected experiences privacy setting in each legal user's apps Learn states that the optional connected experiences privacy setting users see in their Microsoft 365 apps has no effect on whether web search is available.

Memory hook: Web search off for a group, enforced: the Allow web search in Copilot policy in Cloud Policy service.

Microsoft Learn: Manage public web access

Fabrikam has Microsoft 365 E5 and Endpoint DLP on onboarded Windows 11 devices. A user prints a file that matches two DLP rules: one audits printing, and a stricter one blocks it. In activity explorer, filtering on the blocking rule shows the print event, but filtering on the audit rule shows nothing for that user. What explains this?

Correct answer: A. Endpoint events show only the most restrictive DLP rule they matched

Learn's activity explorer article notes that for endpoint events only the most restrictive DLP rule appears, and that the filters you apply also work from that rule. The print matched both rules, so the event carries the blocking rule, and a filter on the audit rule finds nothing even though that rule matched too. Filter on the blocking rule, or on the user and activity, to see the event.

Why the other options are wrong:

  • B. Audit-only rules don't send endpoint events into activity explorer Audited matches do reach activity explorer. Learn's table shows an audited rule producing both an activity event and a rule-match event.
  • C. The device went offline, and the audit rule's event will backfill later The print event is already in activity explorer under the blocking rule, so the device reported it. Offline backfill doesn't explain the missing rule.
  • D. Endpoint DLP reports print events only for files with a sensitivity label The print is already recorded under the blocking rule, so the event was reported. Endpoint events aren't limited to labeled files.

Memory hook: One endpoint event, one rule shown: the most restrictive one.

Microsoft Learn: Data classification activity explorer

Litware has Microsoft 365 E5. Legal asks who removed the Confidential sensitivity label from a contract in a SharePoint library about six weeks ago. In activity explorer, the compliance admin can't find any activity that old. What should the admin use to answer Legal?

Correct answer: A. A search of the audit log in Microsoft Purview Audit for the label removal

Activity explorer reports on up to 30 days of data, so a six-week-old event is outside its window. Its records come from the unified audit log, which keeps them far longer: 180 days under Audit (Standard), and one year by default for SharePoint, Exchange and Entra ID activity by E5 users. An Audit search for the Removed sensitivity label from file activity shows who removed the label and when.

Why the other options are wrong:

  • B. A new DLP policy for the Confidential label, then a review of the alerts it raises A new DLP policy acts only on activity from now on. It can't report a label removal that happened six weeks earlier.
  • C. Export-ActivityExplorerData in Security & Compliance PowerShell with a 60-day start time Export-ActivityExplorerData exports activity explorer's own data, and activity explorer reports on up to 30 days, so a 60-day start time can't reach the event.
  • D. Data explorer, opening the contract to review its sensitivity label history Data explorer is a current snapshot of where labeled and sensitive items sit. It shows today's label, not who removed one six weeks ago.

Memory hook: Older than 30 days? Leave activity explorer and search Audit.

Microsoft Learn: Data classification activity explorer

The HR team at Litware wants harassing or threatening language in Teams chats and Exchange email detected after messages are sent, so that designated reviewers can examine each match and send a notice to the sender. Messages must not be blocked. Which Microsoft Purview solution should the compliance admin configure?

Correct answer: A. Communication Compliance

Communication Compliance detects, captures and lets designated reviewers act on potentially inappropriate messages in Exchange email, Teams, Viva Engage, Copilot and third-party sources. Built-in classifiers for threats, harassment and profanity flag matches after a message is sent, and reviewers can resolve, tag, escalate or notify the user. It reviews conduct rather than blocking it, which fits a requirement that messages are never blocked, as the Learn about Communication Compliance page explains.

Why the other options are wrong:

  • B. eDiscovery eDiscovery finds, preserves and exports content for legal investigations; it has no classifiers that flag harassment for reviewers as messages are sent.
  • C. Data Loss Prevention (DLP) DLP watches for sensitive information such as card numbers and acts with policy tips or blocks; it doesn't route harassing language to reviewers.
  • D. Insider Risk Management Insider Risk Management scores risky user activity such as downloads and exfiltration; it can take signals from Communication Compliance, but it doesn't hand messages to reviewers.

Memory hook: Conduct in messages, reviewed after sending: Communication Compliance. Sensitive data stopped in flight: DLP.

Microsoft Learn: Communication compliance

Fabrikam has Microsoft 365 E5 and Microsoft 365 Copilot licenses, and it runs a Communication Compliance policy built from the Detect Microsoft Copilot interactions template. A user enters a prompt in Microsoft 365 Copilot, and both the prompt and Copilot's response match the policy's conditions. What will the policy's reviewer find?

Correct answer: D. Two items on the policy's Pending tab, one for the prompt and one for the response

Learn's article on Communication Compliance for generative AI says a matching prompt or response appears as a policy match on the policy's Pending tab, with separate entries for prompts and responses; if only one of them matches, only that one appears. A Copilot response shows Copilot as the sender and the user as the recipient. Detecting Microsoft 365 Copilot interactions needs no pay-as-you-go billing.

Why the other options are wrong:

  • A. No items until pay-as-you-go billing is turned on for Copilot interactions Pay-as-you-go billing applies to non-Microsoft 365 AI data such as Copilot Studio, Security Copilot and other connected AI apps. Learn states there's no such requirement or charge for Microsoft 365 Copilot data.
  • B. One item for the prompt only, because the policy doesn't evaluate Copilot's responses The policy evaluates what the user enters and what Copilot returns, and a matching response gets its own entry with Copilot as the sender.
  • C. One item on the Pending tab that holds the prompt and the response together Prompts and responses aren't merged. Learn lists them as separate entries on the Pending tab, so the reviewer can act on each one.

Memory hook: Copilot in Communication Compliance: prompt and response are two pending items, with no pay-as-you-go for Microsoft 365 Copilot.

Microsoft Learn: Communication compliance Copilot

Tailspin Toys must stop employees from sending email that contains customer credit card numbers to recipients outside the organization. When a message is stopped, the sender must see a notice that explains why. Which Microsoft Purview capability should the compliance administrator configure?

Correct answer: D. A data loss prevention policy for Exchange email

Data loss prevention (DLP) acts on what users do with sensitive items. A DLP policy for Exchange email can combine the built-in credit card number sensitive information type with a condition for content shared with people outside the organization, then block the message and show the sender a policy tip that explains why. The DLP overview lists both blocking the share and showing a policy tip among the protective actions a DLP policy can take.

Why the other options are wrong:

  • A. A Data Lifecycle Management retention policy for Exchange A retention policy only decides how long email is kept and when it's deleted. It can't stop a message from being sent or tell the sender anything.
  • B. A Communication Compliance policy for Exchange email Communication Compliance raises alerts when sent messages match a policy, and human reviewers then decide what to do. It supervises conduct after the fact rather than stopping a message as it leaves.
  • C. An Insider Risk Management policy for data leaks Insider Risk Management correlates signals to surface risky users for analysts to investigate. Blocking an individual email at send time is a DLP action, not an Insider Risk Management one.

Memory hook: Stop it as it leaves: DLP. Review it after it's sent: Communication Compliance.

Microsoft Learn: Purview: DLP

Woodgrove Bank has Microsoft 365 E5. A junior analyst must open the DLP Alerts dashboard in the Microsoft Purview portal, assign alerts, add comments and change alert status. Bank policy says the analyst must never see the matched sensitive content or content previews in those alerts. Which role group should the admin assign to the analyst?

Correct answer: C. Security Operator

The DLP Alerts dashboard needs the Manage alerts role plus DLP Compliance Management or its view-only version, and Learn's DLP alerts article lists Security Operator among the role groups that can use it. Security Operator holds Manage Alerts and View-Only DLP Compliance Management but not Data classification content viewer, the role behind content preview and matched sensitive content. Learn suggests this split when someone needs alerts without the sensitive context.

Why the other options are wrong:

  • A. Global Reader Global Reader is read-only. It carries View-Only Manage Alerts but not Manage Alerts, so the analyst couldn't assign alerts, add comments or change their status.
  • B. Information Protection Investigators Information Protection Investigators can work DLP alerts, but the group includes Data classification content viewer, so the analyst would see the matched content the bank forbids.
  • D. Content Explorer Content Viewer Content Explorer Content Viewer grants exactly the content access the bank forbids, and it holds no alert management role for triage.

Memory hook: Seeing the alert and seeing the data are separate grants: Security Operator triages, the content viewer role reads.

Microsoft Learn: DLP alerts

Woodgrove Bank has Microsoft 365 E5. Its DLP policy for customer account numbers flags many numbers that only look like account numbers. Woodgrove keeps each customer's account number, name and date of birth in a database table and wants matches only for values in that table, without uploading the values to Microsoft in clear text. What should the administrator create?

Correct answer: B. An exact data match sensitive information type built from the table

Exact data match (EDM) classification creates a custom sensitive information type that matches exact values from your own table of sensitive data, which gives fewer false positives than a generic pattern. You hash the schema and the values with a salt before upload, so only hashed values reach the service. The EDM overview adds that the table can hold up to 100 million rows and can be refreshed daily.

Why the other options are wrong:

  • A. A custom sensitive information type with a stricter regular expression A stricter regular expression is still a pattern, so it matches any number in the right format, including numbers that belong to no Woodgrove customer.
  • C. A custom trainable classifier trained on sample customer account records A trainable classifier learns a category of documents from examples. It can't tell whether a specific number belongs to a real customer in the table.
  • D. A document fingerprint created from the bank's account opening form A document fingerprint detects documents created from a standard form. Account numbers that appear in email or other files wouldn't match the form.

Memory hook: Only the values in your own table, hashed before upload: EDM.

Microsoft Learn: SIT exact data match based SITs

Adatum has Microsoft 365 E5, and its Windows 11 devices are onboarded to Microsoft Defender for Endpoint. Adatum wants to stop users from copying files that contain Social Security numbers to USB drives, while still letting users copy other files to USB drives. What should the administrator do?

Correct answer: A. Turn on device monitoring and create a DLP policy for Devices that blocks copy to USB

Endpoint DLP extends DLP to onboarded Windows and macOS devices. The endpoint DLP article states that devices already onboarded to Defender for Endpoint appear in Purview automatically, so you only turn on device monitoring. A DLP policy for the Devices location with a Social Security number condition can then block the copy to USB removable device activity for matching files, while other files can still be copied. Endpoint DLP is included in Microsoft 365 E5.

Why the other options are wrong:

  • B. Create a DLP policy for the OneDrive accounts location that blocks sharing these files A OneDrive policy acts on files stored and shared in OneDrive. Copying a file from the device to a USB drive never passes through OneDrive, so the policy doesn't see it.
  • C. Create a Defender for Endpoint device control policy that blocks all removable storage A device control policy that blocks all removable storage stops every file from reaching USB drives, which breaks the requirement to keep copying other files. The device control overview points to endpoint DLP for blocking by file sensitivity.
  • D. Apply a sensitivity label that encrypts every file that contains Social Security numbers Encryption controls who can open a file, not where it can be copied. An encrypted file can still be copied to a USB drive.

Memory hook: Block sensitive files from USB, allow the rest: endpoint DLP on the Devices location.

Microsoft Learn: Endpoint DLP

Contoso has Microsoft 365 E5. Its security operations team wants to investigate DLP alerts where Microsoft recommends doing that work: grouped into incidents with Microsoft Defender for Endpoint alerts and open to advanced hunting across both. Where should the team investigate and manage the DLP alerts?

Correct answer: A. The incident queue in the Microsoft Defender portal

The DLP alerts article on Learn names the Microsoft Defender portal as the recommended place to investigate and manage DLP alerts, and the Microsoft Purview portal as the place to create and edit DLP policies. Under Incidents & alerts, then Incidents, Defender groups DLP alerts into incidents, correlates them with alerts from Defender for Endpoint and other solutions, and adds advanced hunting. Since early October 2026, DLP signals show there as alerts only after a built-in alert tuning rule is disabled.

Why the other options are wrong:

  • B. Audit search in the Microsoft Purview portal Audit search returns individual audit records of user and admin activity. It doesn't group alerts into incidents or offer advanced hunting.
  • C. The DLP Alerts dashboard in the Microsoft Purview portal The Purview dashboard can triage, assign and resolve DLP alerts, but it holds DLP alerts only and doesn't correlate them with Defender for Endpoint alerts in one incident. Learn recommends Purview for building and editing DLP policies.
  • D. Activity explorer in the Microsoft Purview portal Activity explorer lists label and DLP activity drawn from the unified audit log for up to 30 days. It has no incidents and no cross-product correlation.

Memory hook: Write DLP policies in Purview; work DLP alerts as incidents in Defender.

Microsoft Learn: DLP alerts

A paralegal at Fabrikam, a member of the eDiscovery Manager role group, built a search in an eDiscovery case. Before anything is exported, she wants to look at a representative handful of the actual items from each mailbox, see each item's subject, date and sender, and open a few to confirm they're relevant. What should she generate when she runs the query?

Correct answer: D. A sample of items from each location

Run query offers two views. Statistics estimates hits, locations and trends; Sample returns a representative set of items, 1, 10 or 100 per location from up to 10,000 locations, listed by subject or title, date and sender or author, and selecting an item opens its source view. Sample results stay valid for 24 hours, while statistics expire after 14 days, so regenerate whichever view has aged out.

Why the other options are wrong:

  • A. Statistics, with the query keywords report Statistics give counts, and the keywords report shows hits per keyword line; neither lets her read the items.
  • B. Statistics that include partially indexed items This statistics option adds the count and locations of partially indexed items; it still shows no item details.
  • C. An export of the items report only An items report is produced by an export process, which she wants to avoid for now, and it lists metadata rather than opening items.

Memory hook: Statistics count, Sample shows: samples last 24 hours, statistics 14 days.

Microsoft Learn: eDiscovery search results

A records manager at Contoso, a member of the eDiscovery Manager role group, must quickly find every email and document that mentions Project Atlas across all mailboxes and SharePoint sites, and preview a few of them. Legal hasn't opened an investigation, so she doesn't want to create a new case. Where in the Microsoft Purview portal should she create the search?

Correct answer: C. In eDiscovery, in the Content Search case

Content search is no longer a separate tool. Learn's eDiscovery overview says all Content Search functionality now lives in a system-generated eDiscovery case, open by default to members of the eDiscovery Manager and Administrator role groups. She selects Content Search in eDiscovery, creates a search with a name that's unique in the organization, adds the mailboxes and sites as data sources and runs a keyword query, with no new case needed.

Why the other options are wrong:

  • A. In Communication Compliance, as a new policy A Communication Compliance policy routes messages that match conduct conditions to reviewers; it isn't an on-demand search of existing email and files.
  • B. In Audit, as a new audit log search An audit log search returns records of user and admin activities, not the emails and documents themselves.
  • D. In Information Protection, in activity explorer Activity explorer shows actions taken on labeled and sensitive content; it doesn't run keyword searches across mailboxes and sites.

Memory hook: Content search now lives inside eDiscovery, in the system-generated Content Search case.

Microsoft Learn: Purview: eDiscovery

Fabrikam has Microsoft 365 E3, with no E5 licenses or E5 add-ons. An eDiscovery manager's search in a case returns the emails that legal asked for, and legal wants the items in hand for its review. Which next step is available to Fabrikam with its current licenses?

Correct answer: C. Export the search results and download the packages

Learn's eDiscovery feature table puts search, statistics and samples, holds and exporting search results in the standard eDiscovery column, while review sets, tagging, analytics and optical character recognition are premium features that need Office 365 E5, Microsoft 365 E5 or a related E5 add-on. Exporting needs a Microsoft 365 E3 or E5 license, so the manager exports and downloads the packages from the Process manager within 14 days.

Why the other options are wrong:

  • A. Add the search results to a review set for tagging Review sets, and the tagging done in them, are premium eDiscovery features that need E5-level licensing.
  • B. Run near duplicate detection and email threading Near duplicate detection and email threading are review set analytics, a premium feature that E3 alone doesn't include.
  • D. Use OCR in a review set to search text in images OCR on review set content is a premium eDiscovery feature, so it isn't available with E3 alone.

Memory hook: E3 eDiscovery: search, hold and export. Review sets and analytics need E5.

Microsoft Learn: Purview: eDiscovery

Legal at Adatum reports that an engineer may be leaking design files. The engineer is in scope of a Data leaks policy that uses a DLP policy as its triggering event, but there has been no trigger, so none of the engineer's activity has been scored. The insider risk team wants that recent activity scored now and shown on the User activity timeline. What should they do?

Correct answer: B. Use Start scoring activity for users to add the engineer to the policy temporarily

Policies wait for a triggering event before they score a user. Start scoring activity for users, on the Policies page, adds users to policies for a set period and bypasses the triggering requirement. The solution then scores the user's activities for the previous 90 days and adds them to the User activity timeline; the user can take several hours to appear on the Users dashboard, and those 90 days of activity can take up to 24 hours to display.

Why the other options are wrong:

  • A. Run an insider risk analytics scan from the Insider Risk Management Overview page Analytics scans evaluate potential insider risks across the organization without any policy and report insights after up to 48 hours; they don't score one user's activity into a policy.
  • C. Add the engineer to a priority user group in Insider Risk Management settings Priority user groups feed the priority-user templates and a risk score booster; membership doesn't waive the triggering event in a Data leaks policy, so scoring still wouldn't start.
  • D. Set the triggering DLP policy's incident reports to Low severity so it fires sooner Insider risk alerts aren't generated from DLP policies whose incident reports are set at Low or Medium; only a High severity DLP alert triggers the policy, so this removes the trigger.

Memory hook: No trigger, no score. Start scoring manually and IRM looks back 90 days.

Microsoft Learn: Insider risk management policies

Adatum applies a retention policy to its Finance SharePoint site that retains items for seven years and then deletes them. Two months later, a user deletes a budget workbook from the site's document library. Where is a copy of the workbook kept so that compliance tools such as eDiscovery can still find it for the rest of the seven years?

Correct answer: C. In the site's Preservation Hold library, a hidden location for retained files

When a user deletes an item that a retention policy covers in SharePoint or OneDrive, the original is copied to the site's Preservation Hold library, a hidden system location that compliance tools such as eDiscovery use to reach the content. The delete removes the workbook from the document library, but the copy stays in the Preservation Hold library until the retention period ends, as the SharePoint and OneDrive retention page explains. By contrast, Exchange keeps retained copies in Recoverable Items, and Teams and Copilot use the hidden SubstrateHolds folder.

Why the other options are wrong:

  • A. In the Recoverable Items folder of the Exchange Online mailbox that belongs to the user Recoverable Items is where Exchange keeps retained email; a file deleted from a SharePoint library doesn't move into anyone's mailbox.
  • B. In a hidden SubstrateHolds folder under Recoverable Items in the site owner's mailbox SubstrateHolds holds retained Teams, Viva Engage, Copilot and AI app messages, not files from SharePoint sites.
  • D. In the site's second-stage Recycle Bin, where it stays until the seven years end The Recycle Bin keeps deleted items for 93 days and isn't indexed for eDiscovery search; retention keeps its copy in the Preservation Hold library instead.

Memory hook: SharePoint and OneDrive: Preservation Hold library. Exchange: Recoverable Items. Teams and Copilot: SubstrateHolds.

Microsoft Learn: Retention policies SharePoint

Adatum, which has Microsoft 365 E5, creates a team for each acquisition project. Every project team must be private, its owners must not be able to add guests, and its SharePoint site must not allow sharing with people outside Adatum. Adatum doesn't need the files inside these teams to carry a label. What should the administrator use?

Correct answer: A. A sensitivity label scoped to Groups & sites, applied to each team

A sensitivity label with the Groups & sites scope protects the container itself. Its settings include privacy (public or private), external user access, which controls whether owners can add guests, and external sharing from the labeled SharePoint site, so applying it to each team enforces all three requirements. Learn notes that a container label doesn't label the items inside the container, which fits Adatum's plan. The scope becomes available after you enable sensitivity labels for containers.

Why the other options are wrong:

  • B. A DLP policy for Teams chat and channel messages that blocks external sharing DLP for Teams inspects message and file content for sensitive information. It doesn't make a team private or stop its owners from adding guests.
  • C. A sensitivity label scoped to files, set as each library's default label A document library's default label applies to files uploaded or edited in that library. It sets nothing on the team, so privacy, guest access and site sharing stay unmanaged.
  • D. A retention policy for each team's Microsoft 365 group mailbox and site A retention policy decides how long the team's content is kept or when it's deleted. It has no settings for privacy, guests or external sharing.

Memory hook: Protect the team itself: a Groups & sites label. Protect the file: an item label.

Microsoft Learn: Sensitivity labels Teams groups sites

Fabrikam's finance team emails quarterly forecast files to outside partners. Each file must stay readable only by approved people and must carry visible Confidential text, even after a partner downloads it or forwards it to someone else. What should the administrator use?

Correct answer: A. A sensitivity label that applies encryption and a footer

A sensitivity label from Microsoft Purview Information Protection is stored in the file's metadata, so it stays with the content wherever the file is saved or sent. The label can apply encryption that controls who can open the file and what they can do with it, plus content markings such as a footer. Because the protection travels with the file, it still applies after a partner downloads or forwards it, as the sensitivity labels overview describes.

Why the other options are wrong:

  • B. A Communication Compliance policy that reviews outbound finance email Communication Compliance lets reviewers examine messages that match a policy after they're sent. It doesn't encrypt attachments or add text to them.
  • C. A retention label that keeps the forecast files for seven years A retention label decides how long an item is kept and whether it's deleted. It applies no encryption and no visible text, so an unapproved person could still open a forwarded file.
  • D. An Insider Risk Management policy for data leaks by finance users Insider Risk Management scores user activity and raises alerts for analysts to investigate. It doesn't encrypt or mark files, so a forwarded forecast stays readable.

Memory hook: Protection that travels with the file: a sensitivity label.

Microsoft Learn: Sensitivity labels

Topic 3 of 3

Perform basic administrative tasks for Copilot and agents (21 questions)

Go deeper in AB-900 Microsoft 365 Copilot and Agent Administration Fundamentals Complete Exam Guide. Get it on Amazon

Show the 21 questions and explanations

Contoso has 2,500 users, and 300 of them hold Microsoft 365 Copilot licenses. Pay-as-you-go billing isn't set up. Every user can add files to the marketing team's SharePoint site. The site owner asks who can create and use agents in SharePoint on that site. What should the admin tell her?

Correct answer: D. Only the 300 licensed users can create and use agents; the others can't use them

Learn's Get started with agents in SharePoint page says agents in SharePoint need no activation. Creating one takes a Copilot license plus the ability to add files to the site, and using one takes a Copilot license or pay-as-you-go billing set up for the user. Its comparison table covers this tenant: licensed with no pay-as-you-go means licensed users can create and use agents, while unlicensed users can't create or use them.

Why the other options are wrong:

  • A. All users can use agents on the site, but only the 300 licensed users can create them Unlicensed users can use agents in SharePoint only when pay-as-you-go billing is set up, and here it isn't.
  • B. All users can create and use agents, and unlicensed use is billed to Azure With no pay-as-you-go billing connected, nothing bills unlicensed use, so those users can't use agents at all.
  • C. No one can use agents until a SharePoint Administrator activates them for the site Agents in SharePoint don't require activation; they become available to licensed users on eligible sites where those users have permissions.

Memory hook: SharePoint agents need a license or a meter to use; with no meter, only seat holders get them.

Microsoft Learn: SharePoint agents

Adatum wants one administrator to approve, block and reassign agents in the Microsoft 365 admin center and to review Copilot Studio agents, agent flows and environments in the Power Platform admin center inventory. Under least privilege, that administrator must not see canvas apps or cloud flows in the inventory. Which Microsoft Entra role fits?

Correct answer: B. AI Administrator

AI Administrator and Global Administrator are the roles that can install, approve and manage agents in the Microsoft 365 admin center, including assigning owners. Power Platform inventory access follows Microsoft Entra roles, and AI Administrator, like AI Reader, is scoped there to agents, agentic apps, agent flows, environments and environment groups. Canvas apps, model-driven apps and cloud flows sit outside that scope, which is what Adatum asks for.

Why the other options are wrong:

  • A. Global Administrator Global Administrator can do all of it but sees every inventory resource, and Learn says to limit this highly privileged role to emergencies.
  • C. Power Platform Administrator Power Platform Administrator sees every inventory resource, canvas apps and cloud flows included, and governs agents only inside its own product.
  • D. AI Reader AI Reader has the same AI-scoped inventory view, but in the Microsoft 365 admin center it's read-only and can't approve, block or reassign agents.

Memory hook: In Power Platform inventory the AI roles see only AI resources, and AI Administrator is the one that can also act.

Microsoft Learn: Power platform inventory

Contoso's security team wants users to stop installing agents and plugins from external publishers, including any that publishers release in the future. Agents built by Microsoft and agents built by Contoso must stay available to everyone. What should the AI Administrator do in the Microsoft 365 admin center?

Correct answer: A. Under Agents, then Settings, then Agent and plugin access, clear the external publishers option

Agent and plugin access controls which types of agents and plugins users can view and install, with separate options for those built by Microsoft, by your organization and by external publishers. Learn notes that when you disable an option, agents and plugins of that type don't appear for users in the Agent Store. Because the control works by publisher type rather than agent by agent, clearing only the external publishers option also covers external agents released later, while the Microsoft and organization options stay on.

Why the other options are wrong:

  • B. Under Agents, then All agents, filter by publisher type and block each external agent in the list Blocking works only on agents already in the registry, so an external agent released next month would stay installable until someone blocks it too.
  • C. Under Agents, then Settings, then Sharing, select No users and save the change Sharing controls who can share agents built with Agent Builder. It has no effect on installing agents and plugins from external publishers.
  • D. Under Agents, then Settings, then User access, select No users and save the change User access set to No users stops everyone from using any agents and plugins, including the Microsoft-built and Contoso-built agents that must stay available.

Memory hook: Which kinds of agents (Microsoft, your org, external): Agent and plugin access. Which people: User access.

Microsoft Learn: Agent settings

Fabrikam's users have Microsoft 365 E3 and Copilot Chat but no Microsoft 365 Copilot licenses, and no pay-as-you-go billing is set up. Four agents are proposed for company-wide use. Which agent can these users use without the organization adding licenses or setting up metered billing?

Correct answer: A. An agent grounded only in its instructions and two public government websites

Learn's Purchase and manage agents in Copilot Chat page says declarative agents grounded in instructions and public websites are included with Copilot Chat and available by default. For users without the license, agents that access shared tenant data, such as SharePoint or Copilot connector content, are billed by metered consumption, and those pay-as-you-go agents are turned off by default until billing is set up. Only the instructions and public web agent works as the tenant stands.

Why the other options are wrong:

  • B. The ready-made agent on the finance department's SharePoint site Agents in SharePoint need a Copilot license or pay-as-you-go billing, so unlicensed users without either can't use the site's agent.
  • C. An agent that answers from ServiceNow tickets indexed through a Copilot connector Content indexed through a Copilot connector is shared tenant data, which is pay-as-you-go only for users without the license.
  • D. An Agent Builder agent that answers from the HR policy library in SharePoint The HR library is shared tenant data in SharePoint, so for unlicensed users this agent is metered and stays off until pay-as-you-go billing is set up.

Memory hook: Instructions and public web: included. SharePoint or connector data: a license or a meter.

Microsoft Learn: Copilot: Agents

The IT help desk at Contoso holds Microsoft 365 Copilot licenses and already uses Copilot in Teams and Outlook. The team wants an assistant that always follows its triage instructions, answers only from the IT knowledge base site in SharePoint, and can open a ticket in the ticketing system through that system's API. What should the team use?

Correct answer: D. An agent with the triage instructions, the knowledge base as knowledge, and a ticket action

Learn's page comparing declarative and custom engine agents says Copilot's built-in capabilities often need added knowledge, data sources or applications for a specific use case, and agents fill that gap. A declarative agent configures Copilot with custom instructions, custom knowledge such as a SharePoint site, and custom actions that call an external system's API. Agents can retrieve and summarize information or take actions such as updating records, which the ticket step requires.

Why the other options are wrong:

  • A. A triage prompt saved in Copilot Prompt Gallery and shared with the help desk team in Teams A shared prompt reuses the same wording, but a prompt can't add an action that calls the ticketing system's API.
  • B. A Copilot Page that holds the triage steps as a shared canvas the help desk team can edit Copilot Pages turns responses into an editable, shareable canvas. It stores the steps but doesn't answer from the site or open tickets.
  • C. Copilot Search across Microsoft 365 content and the organization's connected data sources Copilot Search finds content across Microsoft 365 and connected sources, but it neither enforces the triage instructions nor opens tickets.

Memory hook: Instructions plus knowledge plus actions for one job: that's an agent, not built-in Copilot.

Microsoft Learn: Agents overview

Adatum selected Do not pin Copilot Chat in Microsoft 365 apps in the Microsoft 365 admin center to keep Copilot Chat away from users without Microsoft 365 Copilot licenses. Two weeks later, an audit shows unlicensed users still open Copilot Chat in Teams, in Outlook and on the web. What should the admin do?

Correct answer: D. Restrict the Copilot app to licensed users in Integrated apps and Teams app policies

Learn's Manage Copilot Chat page says users can still reach Copilot Chat on the web when Do not pin is selected, and the pinning article adds that users can add it from their app store. Removing access means managing the Copilot app in Integrated apps, such as letting only a group of licensed users install it, and allowing Copilot Chat in Teams only for licensed users through app permission policies.

Why the other options are wrong:

  • A. Disable web search for unlicensed users with the Allow web search in Copilot policy Disabling web search limits what Copilot can ground on, but users still open and use Copilot Chat.
  • B. Wait longer, since changes to Copilot Chat pinning take time to reach every user Pinning changes take up to 48 hours, and two weeks have passed. Do not pin only removes the default pin; it was never an access control.
  • C. Turn off Let users access Microsoft apps in your tenant in Integrated apps Learn says turning off that Integrated apps setting won't turn off the Copilot app, which has to be managed directly.

Memory hook: Unpinning hides Copilot Chat; Integrated apps and Teams app policies take it away.

Microsoft Learn: Copilot: Manage

An admin at Litware created a pay-as-you-go billing policy in the Microsoft 365 admin center, linked to an Azure subscription and resource group in the tenant and applied to all users. A week later, users without Microsoft 365 Copilot licenses still can't use a Copilot Chat agent that answers from ServiceNow content indexed by a Copilot connector. What must the admin do next?

Correct answer: D. Connect the billing policy to the Microsoft Copilot Chat service

The Microsoft Copilot pay-as-you-go service overview describes billing in two steps: add a billing policy, then connect it to a Copilot service such as Microsoft Copilot Chat or SharePoint agents. Creating a policy only defines the Azure subscription and the users; it doesn't complete setup. Connecting the policy is what grants the covered users access to the metered agents, and pay-as-you-go stays disabled by default until then.

Why the other options are wrong:

  • A. Add a monthly budget to the billing policy to authorize the spending A budget only sends email alerts at chosen percentages of a spending limit. It neither authorizes nor enables metered agents, and reaching it doesn't stop billing.
  • B. Enable the Allow web search in Copilot policy for the unlicensed users Web search grounds answers in public web content. It doesn't unlock an agent that reads connector content in the tenant.
  • C. Create a second billing policy scoped to a group of unlicensed users A second policy has the same gap, because no policy does anything until it's connected to a service, and the all-users policy already covers these users.

Memory hook: A billing policy is a wallet; nothing is spent or enabled until it's connected to a service.

Microsoft Learn: Pay as you go: Overview

An HR specialist at Contoso built a benefits agent in Agent Builder and shared it with the 15-person HR team. An admin also approved her org catalog submission, so other employees add the agent from the Agent Store under Built by your org. Today she edits the agent's instructions and selects Update, but she doesn't resubmit it. Who gets the new instructions?

Correct answer: A. The HR team she shared it with, while Agent Store users keep the last approved version for now

Learn's page on submitting Agent Builder agents to the org catalog says the shared version and the Agent Store version are tracked as separate entries. Selecting Update changes the shared version that the HR team uses, while the Agent Store version doesn't change until the owner resubmits the agent and an admin approves the update. Until then, users who add the agent from the Agent Store continue to see the last approved version.

Why the other options are wrong:

  • B. Only the employees who added it from the Agent Store, because the shared version waits for resubmission This reverses the two paths: the shared version takes the update, and the Agent Store version is the one that waits for an approved resubmission.
  • C. Everyone, because selecting Update refreshes both the shared version and the Agent Store listing Update changes only the shared version. The Agent Store listing changes only after an admin approves a resubmission.
  • D. No one until an admin approves the change, because an approved agent's shared version is locked After approval, the owner can keep iterating on the shared version at any time. Only the Agent Store version waits for admin approval.

Memory hook: Update reaches shared users; the Agent Store copy changes only after resubmission and admin approval.

Microsoft Learn: Agent builder submit to org catalog

An AI Administrator at Contoso is reviewing a pending request for a contract-review agent. The agent should appear in the Agent Store under Built by your org and be installable only by the Legal group, while every other agent stays available to all users as it is today. What should the administrator do?

Correct answer: D. Select Publish to store and choose the Legal group as the users who can install the agent

Learn's agent requests page says Publish to store opens a publishing wizard: the admin selects the users or groups that can install the agent, optionally selects who gets it preinstalled, applies a policy template, reviews the requested permissions and grants admin consent if appropriate, and then selects Publish. Publishing makes the agent available for installation to the selected audience, so choosing the Legal group meets the goal without touching any tenant-wide setting.

Why the other options are wrong:

  • A. Select Publish to store for all users, then block the agent for users outside the Legal group Block works for the entire organization, preventing any user from using the agent. It can't be scoped to everyone outside one group.
  • B. Approve the request from the Allow user to install state for each member of the Legal group The Allow user to install state is for Microsoft-built agents that admin policy blocks. It isn't how a submitted agent is published to a chosen audience.
  • C. Select Publish to store for all users, then set User access to Specific users or groups with Legal User access is tenant-wide and governs all agents and plugins, so limiting it to Legal would cut every other user off from every agent.

Memory hook: Scope a new agent's audience inside the Publish to store wizard, not with tenant-wide switches.

Microsoft Learn: Agent requests

At Fabrikam, tenant-wide User access is at its default. An HR benefits agent in the agent registry is available only to the HR security group. An AI Administrator opens the agent's Users tab, sets the agent to be preinstalled for the entire organization, leaves the availability scope unchanged, and saves. What is the result?

Correct answer: A. The agent is installed for every user in the organization, including users outside the HR group

Learn's agent details page treats availability (who can discover and install an agent) and installation (who has it preinstalled) as independent settings on the agent's Users tab. It states that if you install an agent to your entire organization, it installs automatically regardless of its availability scope, and that this behavior is by design: the agent is installed for the targeted users even if they weren't in the original availability audience. Everyone at Fabrikam gets the agent.

Why the other options are wrong:

  • B. The change is rejected until the availability scope is widened to all users in the organization Learn doesn't require the install audience to sit inside the availability scope. It describes the two settings as independent and the organization-wide install as working by design.
  • C. The agent is installed only for members of the HR group, because preinstalling stays inside the availability scope This assumes preinstallation is filtered by availability. Learn says the reverse for an organization-wide install: it applies regardless of the availability scope, by design.
  • D. Users outside the HR group see the agent in the Agent Store and must install it themselves When an agent is available to specific users, only those users can see and install it in the store, and here the organization-wide install puts the agent in place for everyone anyway.

Memory hook: Availability decides who may install; preinstalling for the whole organization installs for everyone regardless.

Microsoft Learn: Agent details

Megan and Raj both hold Copilot licenses at Litware. Last week Megan selected the Copilot icon on the Word ribbon several times to open the Copilot pane, then closed it each time without typing. Raj submitted one prompt to Copilot Chat in Teams. Once that week's data arrives, how does the Microsoft Copilot usage report count them?

Correct answer: A. Raj counts as an active user; Megan counts as enabled but not active

The usage report's FAQ counts a user as active in an app only after an intentional action with an AI-powered capability. Learn's own example: selecting the Copilot icon on the Word ribbon to open the pane doesn't count, while submitting a prompt in that pane does. Copilot Chat (work) in Teams is a listed user-initiated feature, so Raj's prompt makes him an active user; Megan stays an enabled user who isn't active.

Why the other options are wrong:

  • B. Megan counts as active, and Raj counts only in the Copilot Chat report Megan's clicks don't count, and the Copilot Chat usage report covers only users without a Copilot license; a licensed user's Copilot Chat prompts show in the Microsoft Copilot usage report.
  • C. Neither counts until each has used Copilot on at least three days that week Three days a week is the AI adoption score's target for habitual use. The usage report has no such threshold: trying one user-initiated Copilot feature in the period makes an enabled user active.
  • D. Both count as active users, because each opened a Copilot experience Opening a Copilot experience isn't enough. Learn uses this exact case, opening the pane from the Word ribbon, as an action that doesn't count toward active usage.

Memory hook: Opening the pane isn't using Copilot; submitting a prompt is.

Microsoft Learn: Microsoft 365 Copilot usage

Developers at Woodgrove Bank need an agent that runs on a language model and orchestration they choose. The agent must also start a review workflow on its own whenever a loan record changes, without waiting for a user to prompt it. Which kind of agent should they build?

Correct answer: A. A custom engine agent built in Copilot Studio or with the Microsoft 365 Agents Toolkit

Learn's comparison of declarative and custom engine agents says declarative agents are limited to Copilot's orchestrator and models and rely on user-initiated interactions, so proactive behavior isn't supported. Custom engine agents are fully customizable, including the choice of AI models and orchestration, and can trigger actions automatically without direct user input. They can be built with low-code Copilot Studio or with pro-code tools such as the Agents Toolkit.

Why the other options are wrong:

  • B. A declarative agent built in Agent Builder and grounded in the loan records Agent Builder builds declarative agents, which run on Copilot's own orchestrator and models and act only when a user starts the interaction.
  • C. A declarative agent built with the Microsoft 365 Agents Toolkit that includes an API action A pro-code tool doesn't change the agent type: a declarative agent still uses Copilot's orchestrator and models, and adding an API action doesn't make it proactive.
  • D. A SharePoint agent built over the document library that stores the loan records A SharePoint agent is a declarative agent that answers from chosen SharePoint content when a user asks. It can't bring its own model or start work on its own.

Memory hook: Own model, own orchestration, acts without a prompt: custom engine agent. Copilot's model, user starts it: declarative.

Microsoft Learn: Agents overview

Fabrikam's admin must find which Power Platform environments hold the most Copilot Studio agents and agent flows, filter those resources by owner to spot items owned by employees who are leaving, and download the entire list to a CSV file. Which tool fits?

Correct answer: C. Manage, then Inventory, in the Power Platform admin center

Power Platform inventory gives tenant administrators one view of the agents, apps, flows and environments built on Power Platform, including Copilot Studio agents and agent flows. Admins can filter and sort by any column, such as environment and owner, rank environments by resource count and find resources owned by departing users before they become orphaned. Download exports the entire inventory to a CSV file, not only the rows loaded in the grid.

Why the other options are wrong:

  • A. Agents, then All agents, then Registry, in the Microsoft 365 admin center The Microsoft 365 agent registry is a governance inventory of agents from every platform. Agent flows aren't agents, and ranking Power Platform environments by resource count is a Power Platform inventory task.
  • B. Licensing, then Copilot Studio, in the Power Platform admin center Licensing, then Copilot Studio, tracks Copilot Studio credit capacity and consumption by environment, agent or user. It reports what used credits, not which resources each owner holds.
  • D. Copilot, then Copilot Studio, in the Power Platform admin center The Copilot Studio page in the Copilot area shows usage insights such as active agents and the top 10 agents by monthly active sessions, not a full resource list with owners to export.

Memory hook: Power Platform resources by environment and owner, exported to CSV: Manage, then Inventory, in the Power Platform admin center.

Microsoft Learn: Power platform inventory

Woodgrove Bank connected a billing policy to Microsoft Copilot Chat so that employees without a Microsoft Copilot license can use metered agents. The admin set a $1,500 monthly budget on the policy, with alerts at 90 and 100 percent sent to a mail-enabled security group. What happens when spending under the policy reaches $1,500 in a month?

Correct answer: C. The group gets an alert, possibly hours later, and agent use and billing both continue

The Create a budget for pay-as-you-go billing page says a budget sends email alerts at the thresholds you set, and that alerts can be delayed by up to 24 hours. It also states that reaching 100 percent of the budget doesn't stop the service or billing; the only way to stop billing is to disconnect the pay-as-you-go service. Woodgrove's users keep using metered agents, and charges keep accruing, until an admin disconnects the service.

Why the other options are wrong:

  • A. Billing stops at $1,500, and users keep using metered agents until the budget resets Billing doesn't pause at the budget amount; charges keep accruing, and only disconnecting the service stops billing.
  • B. Metered agents stop working for users in the policy until the budget resets next month A budget isn't a cap: Learn says the system doesn't enforce it, so metered agents stay available after the limit is reached.
  • D. The service is disconnected from the policy automatically until an admin reconnects it Nothing disconnects on its own; disconnecting the service from the policy is a manual admin action on the Pay-as-you-go services tab.

Memory hook: A pay-as-you-go budget warns but never stops: disconnect the service to stop billing.

Microsoft Learn: Pay as you go budget

Adatum's finance analysts are licensed for Microsoft 365 E3 and use Copilot Chat, but none of them holds a Microsoft 365 Copilot license. Their manager wants them to use Analyst to turn monthly spreadsheet exports into trend charts and forecasts. What must the Microsoft 365 administrator do so these analysts can use Analyst?

Correct answer: D. Assign each analyst a Microsoft 365 Copilot license

Learn's agents admin guide lists Researcher and Analyst among the capabilities an organization gains when it buys Microsoft Copilot licenses, and the agent installation page says Microsoft deployed both to users with Copilot licenses. Pay-as-you-go in Copilot Chat meters work-based chat and agents instead. Learn's E3, E5, and E7 comparison lists Copilot as an add-on to both E3 and E5, so the license itself is the fix.

Why the other options are wrong:

  • A. Create a pay-as-you-go billing policy and connect it to Microsoft Copilot Chat Pay-as-you-go lets Copilot Chat users reach work-based chat and agents on metered billing, but Learn ties Researcher and Analyst to the Microsoft 365 Copilot license, not to a billing policy.
  • B. Upgrade the analysts from Microsoft 365 E3 to Microsoft 365 E5 Microsoft 365 E5 still lists Microsoft Copilot as an add-on. It adds security and compliance features, not the Copilot license that brings Analyst.
  • C. Open Analyst on the Agents page and add the analysts in its Edit users panel For Researcher and Analyst the Edit users panel is disabled, and admins can only block them for the whole tenant. Access follows the Copilot license, not an assignment list.

Memory hook: Researcher and Analyst ride on the Copilot license: no license, no Analyst.

Microsoft Learn: Microsoft 365 agents admin guide

Litware's security team flags a published Copilot Studio agent for a compliance issue. It wants the agent stopped in every channel while its maker keeps testing a fix in Copilot Studio, and it wants to restore access later with one action. What should an administrator do?

Correct answer: D. Block the agent from Manage, then Inventory, in the Power Platform admin center

In Power Platform inventory, Block applies to published agents and canvas apps and is reversible, so Unblock restores access at any time. Learn says makers can still see and test a blocked agent in Copilot Studio, but it can't be used in any other channel, which suits an investigation. Quarantine through the Power Platform API documents the same behavior for Copilot Studio agents.

Why the other options are wrong:

  • A. Delete the agent from Manage, then Inventory, in the Power Platform admin center Delete removes the resource from the inventory instead of isolating it, so the maker has no agent left to test and Unblock can't bring it back.
  • B. Delete the agent from Agents, then All agents, in the Microsoft 365 admin center Deleting the agent in the Microsoft 365 admin center makes it unavailable to its maker as well as its users, so the maker couldn't keep testing the fix.
  • C. Set Sharing to No users under Agents, then Settings, in the Microsoft 365 admin center Sharing applies only to agents built with Agent Builder and decides who can share them. It doesn't stop anyone using a published Copilot Studio agent.

Memory hook: Investigate a Copilot Studio agent: Block it in Power Platform inventory. The maker can test, no other channel works, Unblock restores.

Microsoft Learn: Power platform inventory

After a reorganization at Woodgrove Bank, about 140 shared agents built in Agent Builder became ownerless when their creators' accounts were deleted. The AI Administrator wants to review the affected agents, then give each one to its former owner's manager, based on the Microsoft Entra ID manager hierarchy, in one bulk action and without writing code. What should she use?

Correct answer: C. An agent management rule that reassigns ownerless agents to managers

Agent management rules, under Agents, then Settings, apply governance actions to agents in bulk after the admin reviews the affected list. The rule that reassigns ownerless agents created with Agent Builder finds agents with no valid owner and transfers each one, in a single bulk action, to the previous owner's manager in the Microsoft Entra ID hierarchy. It supports only Agent Builder agents, which these are.

Why the other options are wrong:

  • A. The Assign new owner action in each agent's details pane, one by one Assign new owner works one agent at a time and leaves the admin to look up every manager, which isn't one bulk action.
  • B. The reassign operation in the Power Platform API, once for each agent The Power Platform API reassign operation needs a call per agent with each new owner's Microsoft Entra user ID, which means writing code.
  • D. An agent management rule that blocks ownerless agents with no usage That rule blocks ownerless agents that nobody uses. It takes them out of service instead of handing them to new owners.

Memory hook: Many ownerless Agent Builder agents to their owners' managers at once: an agent management rule.

Microsoft Learn: Agent settings

Last quarter, Woodgrove Bank's AI Administrator blocked a Microsoft-built agent for the whole organization. A loan officer now opens the agent in the Agent Store and selects Request. The administrator decides to let her install it. What must the administrator do in the Microsoft 365 admin center?

Correct answer: B. Unblock the agent, then approve her request, which shows the state Allow user to install

Learn's agent requests page says users can request a Microsoft-built agent that an admin policy makes unavailable, such as a block, and the request appears under Agents, then All agents, then Requests with the state Allow user to install. If the agent is blocked, it must be unblocked before any request to install it can be accepted, so the administrator selects Unblock agent and then approves the request. Approval doesn't override organization settings that block the user's access.

Why the other options are wrong:

  • A. Select Publish to store on her request and choose her as the only user who can install the agent Publish to store is the action for submitted agents pending review. A request for a blocked Microsoft-built agent is handled by unblocking the agent and approving the request.
  • C. Add her to User access under Agents, then Settings, because requests for this agent can't be approved Requests for Microsoft-built agents can be approved through this workflow. It's third-party agents that users can't request and admins can't approve in the Microsoft 365 admin center.
  • D. Approve her request under Requests, which lets her install the agent while it stays blocked for others Learn says a blocked agent must be unblocked before any request to install it can be accepted. Approval doesn't create a per-user exception to the block.

Memory hook: Blocked Microsoft agent requested: unblock first, then approve the Allow user to install request.

Microsoft Learn: Agent requests

A maker at Tailspin Toys selects Submit for admin approval for a Copilot Studio agent so that it can be shown to everyone in the organization. Where does the AI Administrator review the agent's capabilities and data sources, and then publish it?

Correct answer: B. In the Microsoft 365 admin center, under Agents, then All agents, then Requests

Learn's page on managing agent requests says agents that members of the organization build with Copilot Studio, Microsoft Foundry or the Microsoft 365 Agents Toolkit can be submitted for admin approval, and they appear under Agents, then All agents, then Requests in the Microsoft 365 admin center. The admin selects the agent to confirm its capabilities, data sources and custom actions, then chooses Publish to store or Reject submission.

Why the other options are wrong:

  • A. In the Microsoft Entra admin center, under Enterprise apps, then Admin consent requests Admin consent requests are users' requests for admin consent to apps they can't consent to themselves. Agent publishing requests go to the Microsoft 365 admin center instead.
  • C. In Copilot Studio, under Channels, then Availability options Availability options is the maker's side of the flow, where Submit for admin approval is selected. The admin doesn't review or publish the request there.
  • D. In the Power Platform admin center, under Manage, then Inventory The Power Platform inventory lists agents and lets admins delete or block published ones, but a submission for org-wide availability is reviewed in the Microsoft 365 admin center's Requests list.

Memory hook: Maker submits in Copilot Studio; admin decides in the Microsoft 365 admin center under Agents, All agents, Requests.

Microsoft Learn: Agent requests

Litware assigned Microsoft Copilot licenses to its Legal group from Billing, then Licenses, and the admin who did it cleared several items under Turn apps and services on or off. Legal users now have Copilot in Word and Outlook but not in Microsoft Teams. Before anyone changes the assignment, an AI Administrator wants to confirm the cause for one Legal user. What should the AI Administrator do?

Correct answer: D. Run the Copilot service plan diagnostic from Support, then Help and support

Learn's Copilot service plan diagnostic page explains that a Copilot license carries several service plans and that a user missing one feature may be missing one plan, here Microsoft 365 Copilot in Microsoft Teams. An AI Administrator opens Support, then Help and support, enters copilot missing, and runs the test against the user's email address or UPN. The output lists every Copilot service plan assigned, which confirms the cause before the assignment changes.

Why the other options are wrong:

  • A. Review the user's sign-in logs in the Microsoft Entra admin center for Teams failures Sign-in logs record authentication events. A service plan that was turned off isn't a sign-in failure, so the logs can't show which Copilot plans the user holds.
  • B. Search the Microsoft Purview audit log for the user's Copilot interactions in Teams The audit log records Copilot interactions that took place; it doesn't show which service plans the user's license includes.
  • C. Open the Copilot report's Readiness tab and check that user's row in the user table The Readiness tab's user table shows only whether a Copilot license is assigned, plus update channel and app use; it doesn't list the service plans inside the license.

Memory hook: One Copilot feature missing for a licensed user? Suspect a service plan and run the copilot missing diagnostic.

Microsoft Learn: Microsoft 365 Copilot service plans

Northwind Traders connected a pay-as-you-go billing policy to the SharePoint agents service. On a project site, a manager who holds a Microsoft 365 Copilot license and a coordinator with no Copilot license, who is in the policy's security group, each ask the site's agent one complex question in SharePoint. The agent answers from the site's documents. How is this usage charged?

Correct answer: A. Only the coordinator's question is billed, for tenant graph grounding and a generative answer; the manager's has no charge

Learn's Use agents with pay-as-you-go billing page lists SharePoint agent rates for generative answers and tenant graph grounding, and its footnote says interactive use by authenticated Microsoft Copilot users is included at no extra cost, so the licensed manager's question has no charge. SharePoint agents are grounded in the tenant graph, so the unlicensed coordinator's complex prompt costs 12 Copilot Credits (the page still calls them messages), 10 for grounding and 2 for the generative answer, under the Copilot Studio meter.

Why the other options are wrong:

  • B. Both questions are billed for tenant graph grounding and a generative answer, because the policy is connected Connecting the policy bills only the users who need the meter; interactive use by authenticated licensed users is included, so the manager's question costs nothing.
  • C. Both questions are billed, but the manager's question is charged at the lower classic answer rate Licensed users' interactive use isn't moved to a cheaper rate; it carries no charge at all.
  • D. Only the coordinator's question is billed, and only for the generative answer, since site files aren't tenant graph grounding The SharePoint billing page notes that SharePoint agents are grounded in the tenant graph, so the grounding charge applies along with the generative answer.

Memory hook: Same SharePoint agent, two users: the seat holder is included, the metered user pays grounding plus answer.

Microsoft Learn: SharePoint agents Azure billing