Show the 30 questions and explanations
Fabrikam licenses Microsoft Agent 365. On one day, a single analyst is the only user of an Agent Builder agent. She chats with it from 9:00 to 9:10 AM, returns at 9:25 AM and chats until 9:30 AM, then returns at 10:15 AM for one more exchange. With the agent's Activity tab date range set to that day, what does the tab report?
Correct answer: B. One active user and two sessions, because only a gap of over 30 minutes starts a session
The Activity tab counts a user as active after a single interaction in the date range, however often she returns, so the analyst is one active user. A session is a back-and-forth conversation, and a new one starts after 30 minutes of inactivity. The 15-minute gap before 9:25 AM keeps her in the first session, and the 45-minute gap before 10:15 AM starts a second. The tab needs E7 or Agent 365 and currently covers Agent Builder, SharePoint, and Microsoft 365 Agents Toolkit agents.
Why the other options are wrong:
- A. Three active users and three sessions, because every session adds one more active user Active users are unique people who interacted at least once in the date range, so one analyst counts once no matter how many sessions she has.
- C. One active user and three sessions, because each return to the agent starts a new session Returning alone doesn't start a session. The 15-minute gap before 9:25 AM is under the 30-minute inactivity threshold, so that chat continues the first session.
- D. One active user and one session, because a session covers a user's whole day with the agent Sessions aren't daily buckets. A new session starts after 30 minutes of inactivity, so the 45-minute gap before 10:15 AM splits her day into two.
Memory hook: Active user: one person, counted once. New session: 30 minutes of silence.
Microsoft Learn: Agent details
Contoso's tenant is in the EU. Last year its admin opted in to Anthropic models under Anthropic's own commercial terms. Claude has since disappeared from Researcher and the Copilot model picker, and the AI providers operating as Microsoft subprocessors page shows Anthropic set to No users. Only the Research security group should get Claude. What should the AI Administrator do?
Correct answer: A. Opt in to Anthropic again on the subprocessors page and assign the Research security group.
Anthropic now operates as a Microsoft subprocessor, and in the EU, EFTA, and the UK its setting defaults to No users. The older independent processor setting was decommissioned on May 1, 2026, so organizations that opted in under Anthropic's own terms must opt in again on the AI providers operating as Microsoft subprocessors page. Access is assigned per provider to users or security groups, so assigning Research returns Claude to that group across Copilot experiences.
Why the other options are wrong:
- B. Turn on the Anthropic model family on the Features page of a Research environment in Power Platform. The Power Platform setting governs external models only in Copilot Studio and Power Platform, and it can't be turned on until the provider is enabled in the Microsoft 365 admin center.
- C. Opt in to Anthropic models with Data Retention and choose the Research security group for them. Models with Data Retention are a separate, off-by-default category in which Anthropic acts as an independent processor under its own terms. Opting in doesn't restore the standard Claude models.
- D. Turn on the Copilot in Microsoft 365 apps with Anthropic models setting for the whole tenant. That EU, EFTA, and UK setting lets Copilot in Word, Excel, and PowerPoint use Anthropic models. It doesn't affect Researcher or the model picker in other Copilot features.
Memory hook: EU, EFTA, and UK start Anthropic at No users, and old opt-ins must opt in again.
Microsoft Learn: Connect to AI subprocessor
Fabrikam's Cowork spending policy draws on prepaid capacity packs and on an Azure subscription that carries a Copilot Credit pre-purchase plan (P3). Pay-as-you-go is enabled on the same subscription, and both prepaid sources still hold credits this month. In what order does Cowork usage draw on these sources?
Correct answer: A. Prepaid capacity packs first, then P3 commit units, then pay-as-you-go billing
The managing usage-based billing page states that when a policy uses both prepaid capacity packs and a subscription with a Copilot Credit pre-purchase plan, billing follows a fixed order to keep spending predictable: prepaid capacity packs, then the P3 plan, then pay-as-you-go. Pay-as-you-go only picks up what the prepaid sources can't cover, and with no prepaid credits left and pay-as-you-go off, requests are blocked.
Why the other options are wrong:
- B. Prepaid capacity packs first, then pay-as-you-go, with P3 applied at renewal P3 commit units are consumed before any pay-as-you-go overage. They aren't a true-up at renewal.
- C. Whichever source the billing method lists first, then the others in listed order The order is fixed by the service, not by how an admin arranges the billing method.
- D. P3 commit units first, then prepaid capacity packs, then pay-as-you-go billing P3 is discounted, which tempts candidates to spend it first, but Learn puts capacity packs ahead of P3.
Memory hook: Credits draw down packs, then P3, then pay-as-you-go.
Microsoft Learn: Usage based billing manage Copilot credits
Woodgrove Bank is preparing for an ISO certification audit of the management system that governs how it develops and uses AI, including agents run under Agent 365. The compliance manager wants a Compliance Manager assessment whose controls match that standard. Which regulatory template should the assessment use?
Correct answer: D. ISO/IEC 42001:2023
ISO/IEC 42001 specifies requirements for establishing, implementing, maintaining, and continually improving an AI management system, and Compliance Manager offers a premium template for it. The Compliance Manager assessments page lists it among the four AI regulation templates under Premium AI templates, with the EU Artificial Intelligence Act, ISO/IEC 23894:2023, and NIST AI RMF 1.0. Among the choices, only 42001 is an ISO standard for AI, so its template matches the audit.
Why the other options are wrong:
- A. ISO/IEC 27001:2022 ISO/IEC 27001:2022 specifies an information security management system. It's a premium template under Premium templates, not a standard for managing AI.
- B. ISO/IEC 27701:2019 ISO/IEC 27701:2019 specifies a privacy information management system that extends ISO/IEC 27001, so it governs personal data handling, not AI.
- C. NIST AI Risk Management Framework (RMF) 1.0 NIST AI RMF 1.0 is one of the AI templates, but it's a NIST framework, not an ISO standard, so it doesn't match an ISO certification audit.
Memory hook: ISO/IEC 42001 manages AI; 27001 manages security; 27701 manages privacy.
Microsoft Learn: Offering iso 42001
Contoso has Microsoft 365 E7. On the All agents page, an agent's Risks column shows 4, and its Risk details pane lists Microsoft Purview among the platforms contributing signals. A Global Administrator selects the Microsoft Purview link to review the Insider Risk Management alerts but can't view them. What should the administrator do?
Correct answer: A. Assign themselves the IRM Analyst or IRM Investigator role, which that Purview link requires.
Each platform named in the Risk details pane is a deep link, and what opens depends on the viewer's roles. The agent registry page's table accepts Global Administrator, Global Reader, Security Reader, Security Administrator, or AI Administrator for the Entra and Defender links, but only IRM Analyst or IRM Investigator for Purview IRM alerts, and says Global Administrator alone is insufficient. In Purview those are the Insider Risk Management Analysts and Investigators role groups, which a Global Administrator can add themselves to.
Why the other options are wrong:
- B. Wait up to an hour, since Purview can trail the admin center before it shows a counted signal. The documented lag runs the other way: the admin center's counts can trail the security portals by up to an hour. The alert already exists; the viewer lacks the role.
- C. Assign themselves Security Reader, the role the Risk details pane checks for every source link. Security Reader does open the Entra and Defender links, which is why it tempts, but it isn't one of the roles the Purview IRM alerts link accepts.
- D. Assign an Agent 365 license to their own account, since risk details are licensed per viewer. The Risks column needs an E7 or Agent 365 license in the tenant, which Contoso has since the counts show. A license doesn't grant access to Purview IRM alerts.
Memory hook: Purview IRM deep links want IRM Analyst or IRM Investigator, even from a Global Administrator.
Microsoft Learn: Agent registry
Litware's records team wants members of the Research group to stop creating new Copilot Pages and Copilot Notebooks in the Microsoft Copilot app. Loop workspaces must keep working for everyone, and nobody outside Research should be affected. What should the Office Apps admin do?
Correct answer: D. In Cloud Policy, disable Create and view Copilot Pages and Copilot Notebooks in a configuration for Research.
Creation of Copilot Pages and Copilot Notebooks is governed by the Create and view Copilot Pages and Copilot Notebooks setting in Cloud Policy, which is enabled by default and can be scoped to a group. Disabling it for Research stops new Pages and Notebooks, hides the Notebooks module in the Copilot app, and deletes nothing that already exists. Pages and Notebooks are independent of Loop, so Loop workspaces keep working.
Why the other options are wrong:
- A. In Cloud Policy, disable Enable code previews for AI-generated content in a configuration scoped to Research. Code previews only control whether AI-generated code runs as a preview in Copilot Chat and Pages. Creation of Pages and Notebooks stays on.
- B. On the Copilot settings page, restrict Advanced package uploads to users outside the Research group. Advanced package uploads govern who can upload agent and plugin packages that use actions or MCP servers. They don't touch Pages or Notebooks.
- C. In Cloud Policy, disable Create Loop workspaces in Loop in a policy configuration scoped to Research. Loop and Copilot Pages share a storage container, which makes this tempting, but the Loop policy blocks the workspaces that must keep working and leaves Pages and Notebooks creation on.
Memory hook: Pages and Notebooks share one Cloud Policy switch, and it's separate from Loop.
Microsoft Learn: Cpcn admin configuration
Fabrikam's security team wants to stop administrators from using Copilot inside the Microsoft 365, Exchange, SharePoint, and Teams admin centers, while Copilot for licensed end users stays exactly as it is. Several of the administrators hold no Copilot license, yet they already use Copilot in those admin centers. What should the AI Administrator do?
Correct answer: C. Turn off Microsoft Copilot in admin centers on the User access tab of Copilot settings.
Microsoft Copilot in admin centers, on the User access tab under Copilot then Settings, decides whether users with an admin role can use Copilot in the Microsoft 365, Exchange, SharePoint, and Teams admin centers. It is allowed for all admins by default, which is why unlicensed admins already use it. Turning it off removes Copilot from those admin centers and leaves end-user Copilot alone. The admin experience needs no extra Copilot license, so licenses aren't the lever.
Why the other options are wrong:
- A. Remove the Microsoft Copilot licenses from every account that holds an administrator role. Licenses feel like the obvious switch, but the admin-center experience needs no extra Copilot license, as the unlicensed admins show, and stripping licenses would also change Copilot for admins who use it as end users.
- B. Set Copilot Frontier to No access so preview Copilot features stay off for every administrator. Frontier only governs early access to preview features and already defaults to No access. Copilot in the admin centers is a generally available setting with its own control.
- D. Turn off Pin Microsoft Copilot Chat on the User access tab of Copilot settings. Pinning decides whether Copilot Chat is pinned in Teams, Outlook, and the Copilot app. It would change the end-user experience and leaves Copilot in the admin centers untouched.
Memory hook: Copilot in the admin centers is on for every admin by default. The off switch sits on the User access tab.
Microsoft Learn: Microsoft 365 Copilot page
Woodgrove Bank has a Conditional Access policy, set to On, that blocks every agent identity derived from its trading blueprint from All resources. Later, the hosting service uses the blueprint to call Microsoft Graph and create two new agent identities, and an existing trading agent calls a partner pricing service with a static API key that isn't registered in Microsoft Entra ID. What happens?
Correct answer: B. Both succeed; the blueprint's creation token and API-key calls are outside Conditional Access.
The Conditional Access for agents page lists its boundaries: policies don't apply when an agent identity blueprint acquires a Microsoft Graph token to create an agent identity or an agent's user account, and Conditional Access protects only resources secured by Microsoft Entra ID, so a call made with an API key bypasses token issuance entirely. The block still applies once the new agent identities request tokens for Entra-protected resources.
Why the other options are wrong:
- A. The creation is blocked, but the API-key call succeeds because the service isn't in Entra ID. Blueprint token acquisition for creating agent identities is a documented exception, so the creation isn't blocked.
- C. Both are blocked, because a blueprint-scoped policy covers every token the blueprint requests. A blueprint-scoped policy covers the agent identities derived from the blueprint, not the blueprint's own Graph token for creating them.
- D. The creation succeeds, but the API-key call is blocked because the policy targets All resources. All resources means resources protected by Microsoft Entra ID; a service reached with an API key never asks Entra for a token, so the policy never evaluates it.
Memory hook: Blueprint creation tokens and API keys sit outside Conditional Access.
Microsoft Learn: Agent ID
At Tailspin Toys, six platform engineers in an assigned-membership security group must be able to edit an agent identity blueprint's settings and manage its credentials. Adding the group as an owner of the blueprint fails, and the team won't accept any tenant-wide admin role. What should the administrator do?
Correct answer: B. Add each of the six engineers individually as an owner of the blueprint.
The owners, sponsors, and managers page says individual users, including guests, and service principals can be owners, while groups aren't supported as owners. Owners hold administrative rights scoped to their blueprint or agent identity: they edit settings, manage credentials, change configurations, and add more owners. Adding each engineer as an owner meets the need with no directory role, and multiple owners also give backup coverage.
Why the other options are wrong:
- A. Add the security group as a sponsor of the blueprint instead of an owner. Assigned-membership security groups are rejected as sponsors, and sponsorship wouldn't grant technical access even for an allowed group type.
- C. Convert the group to a role-assignable group, then add it as an owner. No group type can be an owner, and role-assignable groups are excluded from sponsorship as well.
- D. Add each of the six engineers individually as a sponsor of the blueprint. Sponsors carry business accountability with nondestructive lifecycle rights only; they can't modify application settings or credentials.
Memory hook: Owners are people or service principals, never groups.
Microsoft Learn: Agent owners sponsors managers
Fabrikam bills Copilot Studio agents through a Power Platform billing plan and Cowork through its default spending policy, both on one Azure subscription. In Azure Cost Management, finance sees all of it under a single Microsoft Copilot Studio service and wants Cowork charges separated from next month on. What should the admin do?
Correct answer: C. Bill the Cowork policy to its own Azure subscription or resource group.
The article comparing admin center views with the Azure bill says Cowork, Copilot Studio, and Work IQ API consumption is billed through one Copilot Credits service labeled Microsoft Copilot Studio, and that attributing cost to one service means separating it by Azure subscription or resource group. A Global or Billing Administrator can change an existing spending policy's billing method without recreating the policy.
Why the other options are wrong:
- A. Filter Azure Cost Management on a Microsoft Copilot Cowork service name. Azure has no separate Cowork service today. Cowork consumption posts under Microsoft Copilot Studio.
- B. Use the Consumption tab export in Copilot Cost Management as the cost record. The admin center counts Copilot Credits, not currency, and Microsoft says to reconcile against the monthly billing record rather than the dashboards.
- D. Turn off Auto-apply new services so Cowork charges post as their own service. Auto-apply new services only decides whether newly supported services join a policy. It doesn't change how Azure labels charges.
Memory hook: One Copilot Studio line in Azure: split cost by subscription or resource group.
Microsoft Learn: Usage based billing compare dashboard views
Tailspin Toys' insider risk team wants Insider Risk Management to detect prompt injection attacks and access to protected materials when employees interact with agents managed by Agent 365. Which policy template should the team create its policy from?
Correct answer: D. The Risky AI usage template
The Purview page for Agent 365 points Insider Risk Management to the Risky AI usage policy template to detect risky usage that includes prompt injection attacks and access to protected materials. The policy templates page adds that this template scores user prompts and AI responses containing sensitive information in Microsoft 365 Copilot, Microsoft Copilot, and agents, and that its detections feed user risk scoring in Adaptive Protection. The other three templates score file, departure, and device signals rather than AI interactions.
Why the other options are wrong:
- A. The Data theft by departing users template Data theft by departing users scores exfiltration near resignation or end dates from an HR connector or account deletion, so it never evaluates prompt injection in agent interactions.
- B. The Data leaks template Data leaks sounds right for protected materials, but it scores SharePoint downloads, file sharing, printing, copying to personal cloud services, and high-severity DLP alerts, not risky prompts to agents.
- C. The Security policy violations template Security policy violations scores Microsoft Defender for Endpoint alerts about defense evasion and unwanted software on devices, not AI interactions.
Memory hook: Prompt injection or protected materials in agent chats: Risky AI usage.
Microsoft Learn: AI agent 365
Tailspin Toys wants makers who build agents with Microsoft Copilot Agent Builder to stop sharing them with the whole organization. Makers must still be able to share an agent directly with individual colleagues, and no group of makers should keep broad sharing rights. Which configuration under Agents then Settings meets these requirements?
Correct answer: C. Set Sharing to No users, and leave User access at its default of All users.
Sharing on the Agent settings page offers All users, No users, and Specific users. No users turns off sharing at the organization level while still letting users share directly with specific individuals, which matches Tailspin's requirement exactly. Specific users restricts broad sharing to designated groups, so a group would keep it. The page also notes that the Sharing control applies only to agents built with Microsoft Copilot Agent Builder, the maker population in this scenario.
Why the other options are wrong:
- A. Set Sharing to Specific users, and select a small group of trusted makers. Specific users feels like the cautious middle ground, but it keeps broad sharing rights for the selected group, and the stem says no group should keep them.
- B. Set User access to Specific users/groups, and select only the makers. User access gates who can use agents, not how makers share them, and it would also stop every other employee from using agents.
- D. Under Allowed agent types, turn off Allow apps and agents built by your organization. This toggle keeps org-built agents out of the Agent store for users. It doesn't govern sharing and would take the organization's agents away from everyone.
Memory hook: Sharing set to No users: no org-wide shares, direct shares with named people still work, Agent Builder only.
Microsoft Learn: Agent settings
Woodgrove Bank has 35 assigned Microsoft Copilot licenses and no Viva Insights licenses. Executives who open the Copilot Dashboard want to compare adoption with similar companies, but no benchmarks appear. What has to change before benchmarks are available?
Correct answer: D. Reach at least 50 Copilot licenses or 50 Viva Insights licenses.
The Copilot Dashboard is the strategic report in the measurement and reporting pillar of Copilot controls, and its license table decides which features a tenant gets. Tenants with 1 to 49 Copilot licenses get the Readiness page and adoption and impact metrics, but not benchmarks, agent insights, intelligent summaries, or delegation. Those arrive with at least 50 Copilot licenses or at least 50 Viva Insights licenses.
Why the other options are wrong:
- A. Have an executive delegate dashboard access to the other leaders. Delegation is itself one of the features a tenant under 50 licenses doesn't get, and it wouldn't add benchmarks.
- B. Upload an organizational data file with the Organization attribute. Organizational data adds filters such as Organization. It doesn't change which features the license count allows.
- C. Lower the dashboard's minimum group size from 10 down to 5. Minimum group size decides which groups show metrics. It doesn't unlock benchmarks, and it can't go below 5.
Memory hook: Benchmarks, agent insights, and delegation need 50 Copilot or 50 Viva Insights licenses.
Microsoft Learn: Copilot dashboard
Litware's search administrator is bulk importing 40 acronyms into Copilot Search from a CSV file. Legal must approve each definition before users can see it in search results, and the review will take several days. How should the administrator set up the import?
Correct answer: D. Set State to Draft for each row, then move each acronym to Published after approval.
The Copilot Search admin experience article gives acronyms three states: Published, which users see in Copilot Search; Draft, which stays out of search results until an admin moves it to Published; and Excluded, which prevents an acronym from appearing. State is a mandatory column in the acronym import file, so importing each row as Draft and publishing after Legal signs off matches the review requirement.
Why the other options are wrong:
- A. Set State to Excluded for each row, then remove the exclusion after Legal approves it. Excluded actively suppresses an acronym, and undoing it means deleting the excluded entry and adding the acronym again, which isn't a review workflow.
- B. Leave the State column empty so the acronyms stay hidden until an admin publishes them. State is a mandatory column in the acronym import file, so leaving it empty isn't an option.
- C. Set State to Scheduled with a publish date after Legal's review is expected to end. Scheduled is a bookmark state; acronym rows accept only Published, Draft, or Excluded.
Memory hook: Acronyms awaiting review go in as Draft; Excluded suppresses, and Scheduled is for bookmarks.
Microsoft Learn: Microsoft 365 Copilot search admin experience
Litware's sales team relies on a shared Agent Builder agent that a single employee owns. That owner moves to a new role in two months. The team lead wants a colleague to have full edit rights starting now, while the current owner keeps full access until the move. What should the AI Administrator do in the Microsoft 365 admin center?
Correct answer: A. Open the agent's Owners pane and use Add owner to add the colleague.
The governance and lifecycle actions page says Agent Builder agents support multiple owners, and all owners have the same rights to edit, share, manage, and maintain the agent, with no primary or secondary roles. Add owner leaves the current owner in place, so both keep full access during the transition. The departing owner can be removed later, since only the last remaining owner can't be removed.
Why the other options are wrong:
- B. Add the colleague as a specific user under Agents, then Settings, then Sharing. The Sharing setting controls who can share Agent Builder agents in the organization; it grants no one edit rights on a specific agent.
- C. Use Assign new owner on the agent and select the colleague as owner. Assign new owner gives the colleague full rights but removes all of the previous owner's access, including read, which breaks the transition requirement.
- D. Run the Reassign ownerless agents created with Agent Builder to manager rule. That rule acts only on ownerless agents and hands them to the previous owner's manager; this agent still has an owner.
Memory hook: Add owner keeps both people in; Assign new owner locks the old owner out.
Microsoft Learn: Agent actions
Litware turned on Restricted SharePoint Search in March 2026 with 60 reviewed sites on its allowed list. The Budget Planning site isn't on the list and isn't associated with any hub site, yet a planner who edited two files there last week gets Copilot answers that quote them. What explains this?
Correct answer: B. Restricted SharePoint Search still surfaces files a user recently viewed, edited, or created.
The Restricted SharePoint Search page says the feature isn't a security boundary and doesn't guarantee that only allowed sites appear. Results and Copilot responses still draw on files a user viewed, edited, or created, files shared directly with them, and their frequently visited sites, up to the last 2,000 such entities. The planner edited the files last week, so Copilot can use them. The feature is retiring, and new enablement has been blocked since July 31, 2026.
Why the other options are wrong:
- A. Allowed-list changes can take more than a week to reach Copilot for sites that hold many items. No list change happened here, and the week-plus delay for sites over 500,000 items belongs to Restricted Content Discovery; Restricted SharePoint Search goes into effect within an hour of being enabled.
- C. The allowed list governs org-wide search only, and Copilot answers don't honor it. As a temporary measure, the allowed list applies to Copilot Chat and agentic experiences as well as org-wide search.
- D. Budget Planning is included because it rides along with a hub site on the allowed list. Associated sites of an allowed hub site are included without counting toward the 100-site limit, but Budget Planning belongs to no hub.
Memory hook: Restricted SharePoint Search is an allowed list, not a wall: recent and shared files still surface.
Microsoft Learn: Restricted SharePoint search
Contoso licenses Microsoft 365 E5 for every user and has bought no add-ons. An administrator goes to Agents, then All agents, in the Microsoft 365 admin center and selects an agent on the Registry list. The agent's Security tab shows no details, and the list has no Risks column. What should Contoso buy so both appear?
Correct answer: B. Microsoft Agent 365 licenses, sold per user as an add-on for eligible plans that include E5
The agent details page in the Microsoft 365 admin center says a Microsoft E7 or Agent 365 license is needed to see the Risks column in the Agent Registry and the details on an agent's Security tab, and Contoso's E5 tenant has neither. The Entra licensing page adds that Agent 365 is included in Microsoft 365 E7 and is otherwise sold as an add-on for Microsoft E5, A5, or Business Premium, licensed per user, so buying Agent 365 licenses turns on both views.
Why the other options are wrong:
- A. Microsoft 365 Copilot licenses, the AI assistant component that Microsoft 365 E7 adds on top of E5 Copilot is part of the E7 bundle, which makes it tempting, but the Risks column and the Security tab check for E7 or Agent 365, and a Copilot license is neither.
- C. Microsoft Purview Suite licenses, since each Security tab action opens a Microsoft Purview solution The Security tab does hand off to Purview, but the license it checks is E7 or Agent 365. The Purview Suite, paired with the Defender Suite, is only one eligible base for buying Agent 365.
- D. Microsoft Entra Suite licenses, the identity component that Microsoft 365 E7 adds on top of E5 Entra Suite also ships in E7, yet it doesn't unlock the registry's Risks column or the Security tab. Microsoft Entra Agent ID itself is available to every Microsoft Entra customer.
Memory hook: Risks column and Security tab: no E7 or Agent 365, no data.
Microsoft Learn: Agent details
Woodgrove Bank labels Finance workbooks in Excel for Windows with a label that encrypts them and grants the Finance group the Editor permission level, then saves them to a SharePoint site. Copilot in Excel for Windows summarizes an open workbook, but Copilot Chat never finds or cites these files, SharePoint search can't find words inside them, and Excel for the web can't open them. What should you do?
Correct answer: B. Run Set-SPOTenant -EnableAIPIntegration $true, then have the existing labeled files edited or uploaded again.
Until sensitivity labels for Office files in SharePoint and OneDrive are enabled, those services can't process encrypted files, so search, Office for the web, and similar features fail, and Copilot reaches such files only as data in use in Office apps on Windows. Turn it on with Set-SPOTenant -EnableAIPIntegration $true or Turn on now in the Purview portal. Files labeled earlier don't change until they're edited or uploaded again.
Why the other options are wrong:
- A. Run Set-SPOTenant -EnableSensitivityLabelforPDF $true so SharePoint and OneDrive can process the workbooks. That parameter adds PDF support for sensitivity labels in SharePoint and OneDrive; it isn't the setting that lets the services process encrypted Excel files.
- C. Grant the Finance group Full Control on the SharePoint site so Copilot can read the encrypted workbooks. Finance users can already open the files, and site permissions don't let SharePoint process content that a label encrypted.
- D. Change the label so the Finance group gets the Owner permission level, which adds the EXTRACT usage right. The Editor level already includes Copy, the EXTRACT usage right, which is why Copilot in Excel for Windows can summarize an open workbook.
Memory hook: Encrypted files invisible to search and Copilot Chat: enable labels for SharePoint and OneDrive, then re-save the files.
Microsoft Learn: Sensitivity labels SharePoint OneDrive files
In Microsoft Purview Data Security Posture Management, the weekly default data risk assessment flags Fabrikam's Deals site: many files contain sensitive information types but carry no sensitivity label. The deal team still needs Copilot to use the site's other content, and the flagged files must end up protected by labels. Which remediation from the site's Protect tab fits this finding?
Correct answer: D. Create an auto-labeling policy, which applies a sensitivity label to files with sensitive data.
The data risk assessments page lists four Protect tab actions: restrict access by label, restrict all items, create an auto-labeling policy, and create retention policies. The auto-labeling action is the one for sensitive information found in unlabeled files: Information Protection applies a sensitivity label automatically, so the files gain protection while the rest of the site stays available to Copilot.
Why the other options are wrong:
- A. Restrict access by label, which creates a DLP policy that stops Copilot using labeled files. The DLP policy behind restrict access by label matches selected sensitivity labels, and these files have no label, so it wouldn't act on them.
- B. Restrict all items, which uses Restricted Content Discovery to exempt the site from Copilot. Restrict all items applies Restricted Content Discovery to the whole site, so Copilot loses the other content the deal team still needs.
- C. Create retention policies, which delete site content that hasn't been accessed for three years. Retention policies from this action delete stale content; they don't label sensitive files that are still in use.
Memory hook: Sensitive but unlabeled in a data risk assessment: answer with an auto-labeling policy.
Microsoft Learn: Data security posture management oversharing
Two days ago, Tailspin Toys' AI Administrator added the executive team's Microsoft Entra ID group under Manage access settings for Microsoft Copilot Dashboard. The executives can open the dashboard in the Viva Insights web app, but the AI Administrator still can't. What should the admin do?
Correct answer: C. Add their own account in the same access settings and allow up to 24 hours.
The article on managing Copilot Dashboard settings notes that the AI Administrator doesn't have automatic access to the Viva Insights web app and must add themselves through the same Manage access settings page. Changes to that access list take effect within 24 hours. The executives got in because their group was added, and the admin's own account wasn't part of it.
Why the other options are wrong:
- A. Ask a Global Administrator for the Reports Reader role, which opens the dashboard. Reports Reader opens the admin center usage reports. It isn't one of the ways a user gains Copilot Dashboard access.
- B. Assign themselves a Microsoft Copilot license, which dashboard viewers must hold. Neither a paid Viva Insights license nor a Copilot license is required to view the Copilot Dashboard.
- D. Turn the old Copilot Dashboard feature access control back on with PowerShell. The earlier Copilot Dashboard controls in the admin center and PowerShell are no longer available. Access now runs through the Viva Insights web app.
Memory hook: The AI Administrator isn't auto-enabled: add your own account to see the Copilot Dashboard.
Microsoft Learn: Manage settings Copilot dashboard
Fabrikam runs customer-facing Copilot Studio agents on its public website and forecasts about 1.5 million Copilot Credits over the next 12 months, most of it in two seasonal peaks. Finance wants a discount for committing up front, wants credits unused in a quiet month to stay usable later in the year, and wants the agents to keep answering if demand beats the forecast. Which purchasing approach meets all three goals?
Correct answer: C. Buy a Copilot Credit pre-purchase plan sized to the forecast, and keep a pay-as-you-go billing policy linked to cover usage beyond it.
The usage-based billing comparison page describes Copilot Credit pre-purchase plans as discounted commit units that cover eligible usage across a one-year term or until the units run out, so a quiet month's credits stay available for the peaks. The same page says usage draws down the pre-purchase balance first and can then bill through pay-as-you-go when it's enabled, which keeps the agents answering past the forecast.
Why the other options are wrong:
- A. Link the environments to an Azure subscription through a pay-as-you-go billing policy only, and count the metered rate as the discount. Pay-as-you-go has no upfront commitment and bills usage at the applicable rate. It keeps the agents running but gives Finance no commitment discount.
- B. Buy prepaid Copilot Credit capacity packs sized to one twelfth of the forecast each month, and let quiet-month credits carry into the peaks. Copilot Studio enforces purchased capacity monthly, and unused Copilot Credits don't carry over to the next month, so quiet-month credits are lost rather than saved for the peaks.
- D. Assign Microsoft 365 Copilot licenses to the customer-service staff, so the agents' conversations with website visitors are zero-rated. The no-charge inclusion covers employee-facing use by people licensed for Microsoft 365 Copilot. Website visitors aren't licensed users in Fabrikam's tenant, so their conversations still consume credits.
Memory hook: Pre-purchase plan: a discounted year-long pool. Capacity packs: use them each month or lose them. Pay-as-you-go: the overflow.
Microsoft Learn: Usage based billing compare dashboard views
Fabrikam's Confidential label encrypts files and assigns the Co-Author permission level to Add all users and groups in your organization. Labels are enabled for Office files in SharePoint and OneDrive, and the files are explicitly shared with an Agent 365 agent instance (preview). The agent instance summarizes unlabeled files but returns nothing from Confidential ones. What should the administrator change?
Correct answer: C. Add the agent instance to the label's encryption settings by name, with at least VIEW and EXTRACT
Agent instances, agents with their own identity, are available through the Frontier preview program. The Purview page for Agent 365 sets two conditions for an agent instance to use labeled files: the files must be explicitly shared with it, and label encryption must explicitly grant the agent instance the VIEW and EXTRACT usage rights. It adds that a label configured for Add all users and groups in your organization or Add any authenticated users isn't sufficient. Sharing and label enablement are already in place, so the explicit grant is the missing piece.
Why the other options are wrong:
- A. Grant the agent instance the Files.Read.All Microsoft Graph permission on its Permissions tab An API permission doesn't satisfy a label's encryption. The encryption itself has to grant the agent instance the VIEW and EXTRACT usage rights.
- B. Change the label's assignment to Add any authenticated users, so accounts beyond the org list qualify The same page names Add any authenticated users as insufficient, just like Add all users and groups in your organization, so the agent instance still gets no usage rights.
- D. Share the Confidential files with the agent instance a second time, this time with edit permission The files are already explicitly shared, which meets the sharing condition. Sharing them again changes nothing about the rights that the label's encryption grants.
Memory hook: Encrypted file and an agent instance: share it, and name the agent in the label with VIEW and EXTRACT.
Microsoft Learn: AI agent 365
Woodgrove Bank already licenses Microsoft 365 E5 for every user. A 25-person automation team builds and owns about 400 agents across Copilot Studio and Microsoft Foundry, and the bank wants the generally available Microsoft Agent 365 to observe, govern, and secure them. Which licensing action enables Agent 365 and covers those agents?
Correct answer: A. Assign Agent 365 licenses to the 25 team members, since agents that a licensed user owns or manages need no license of their own.
The Agent 365 overview says the product has been generally available for the Commercial segment since May 1, 2026, is licensed per user, and needs at least one user with a qualifying Agent 365 license before it can be enabled. The Frontier program FAQ adds that under general availability every agent managed or owned by a licensed user is covered by that user's Agent 365 or Microsoft 365 E7 license. Licensing the 25 owners covers the 400 agents.
Why the other options are wrong:
- B. Assign Microsoft 365 Copilot licenses to the 25 team members, since the Copilot add-on includes Agent 365 for agents its users build. Agent 365 is its own subscription, included in Microsoft 365 E7 but not in the Copilot add-on, so Copilot licenses don't enable it.
- C. Buy about 400 Agent 365 licenses and assign one to each agent instance before those agents can be governed. Per-instance licensing is how the Frontier preview works. Under general availability Agent 365 is licensed per user, and agents don't require their own Agent 365 license.
- D. Enroll the tenant in the Frontier program, since Frontier turns on Agent 365 for Microsoft 365 E5 tenants without separate licenses. Frontier previews Agent 365 features and requires at least one E7 license in the tenant. It isn't a route to the generally available product for an E5 tenant without Agent 365 licenses.
Memory hook: Agent 365 licenses people, not agents: license the owners and their agents are covered.
Microsoft Learn: Microsoft agent 365: Frontier
Contoso's Copilot program lead holds the AI Administrator role. In Copilot, then Cost Management, she can change the limits and alerts on the existing spending policy, but she can't create the new Cowork spending policy for the Legal group. Which role should she also be given so she can create it?
Correct answer: C. Billing Administrator
The usage-based billing role requirements separate editing from creating. AI Administrator and License Administrator can edit spending policies and manage limits and alerts, but they can't create spending policies. Global Administrator and Billing Administrator are the roles that run the initial setup, which activates the default spending policy, and that set billing methods in policies, so Billing Administrator lets her add the Legal policy without Global Administrator.
Why the other options are wrong:
- A. License Administrator License Administrator sounds like it should govern consumption, but it has the same boundary as AI Administrator here: it can edit spending policies and limits, not create them.
- B. Global Reader Global Reader is a reader-based role that can view consumption dashboards and reports. It can't create or change spending policies.
- D. Office Apps Administrator Office Apps Administrator creates Cloud Policy configurations for the Microsoft Copilot app. It isn't one of the roles that manage spending policies.
Memory hook: Spending policies: AI and License admins edit, Global and Billing admins create.
Microsoft Learn: Usage based billing manage Copilot credits
Fabrikam runs production agents in Amazon Bedrock and Google Vertex AI. The AI administrator wants those agents listed in the agent registry in the Microsoft 365 admin center for central visibility and governance, without asking the agent teams to do any development work. What should the administrator do?
Correct answer: B. From the All agents page, open Connected platforms, connect each environment with its credentials, then select Sync agents.
Connected platforms is the registry's route for agents that live on external AI platforms, and Amazon Bedrock and Google Vertex AI are both supported. From its web part on the All agents page, the administrator selects Manage, adds a connection per environment with its platform, region, and credentials, validates and saves it, and then selects Sync agents to pull those agents into the registry. Nothing changes in the agents' code, so the teams do no development work.
Why the other options are wrong:
- A. Integrate each agent with the Agent 365 SDK so that it gets a Microsoft Entra agent identity and shows up in the registry. The Agent 365 SDK is a real way to bring outside agents under the control plane, but it's development work for every agent team, which the stem rules out.
- C. Export each agent as a .zip package from its platform, then upload the packages with Add agent on the All agents page. Add agent uploads the .zip agent package that Copilot Studio and Agent Builder export; Connected platforms is the documented route for Bedrock and Vertex AI agents.
- D. Register each agent's endpoint as a bring-your-own MCP server with the Agent 365 CLI, then approve it on the Tools Requests tab. A bring-your-own MCP server registers a tool that agents call, so the Bedrock and Vertex AI agents themselves still wouldn't appear in the registry.
Memory hook: Agents running in another cloud: connect the platform, then Sync agents.
Microsoft Learn: Connected platforms
At Woodgrove Bank, User access under Agents then Settings is set to Specific users/groups with only the AI-Pilot group selected. A loan officer who isn't in AI-Pilot can't use an agent whose Available to setting is All users in the organization. The bank wants this officer to use agents without opening agents to every employee. What should the administrator do?
Correct answer: D. Add the loan officer, or a security group that contains the officer, to the User access setting.
User access under Agents then Settings is the tenant-wide gate on using agents. With Specific users/groups selected, the Agent settings page says only the users or groups picked there can use agents, even if they have permission to install and use agents from the registry. Available to is already All users, so the officer's block comes from User access; adding the officer, or a group containing the officer, fixes it for that person without opening agents to everyone.
Why the other options are wrong:
- A. Change the agent's Available to option to Specific users or groups, and select the loan officer. Available to already includes the officer as All users, and narrowing it doesn't lift the tenant-wide User access gate.
- B. Set Sharing under Agents then Settings to All users so the maker can share the agent with the officer. Sharing controls how Agent Builder makers share their agents. It grants nobody the right to use agents that User access leaves out.
- C. Set the agent's Installed for option to include the loan officer so that it's preinstalled for them. Preinstalling looks like a way to hand the officer the agent, but Installed for only controls who gets it preinstalled, and User access still refuses use.
Memory hook: Available to scopes one agent; User access is the tenant-wide gate on using any agent.
Microsoft Learn: Agent settings
Litware's on-call team wants Service health email only when a Copilot problem makes the service or a major function unavailable. It doesn't want mail about problems that affect some users and usually come with a workaround. What should the admin configure?
Correct answer: B. Customize, then Email in Service health, with incidents selected and advisories cleared
The service health article defines an incident as a critical issue in which the service or a major function is unavailable, and an advisory as a problem affecting some users while the service stays available, often with a workaround. Under Customize, then Email, each admin chooses incidents or advisories, the services to cover, and up to two addresses, so incidents only for Copilot matches the on-call rule.
Why the other options are wrong:
- A. Customize, then Email in Service health, with advisories selected and incidents cleared This is the reverse. Advisories are the limited-impact problems with workarounds that the team wants to skip.
- C. Customize, then Custom view in Service health, with only Copilot selected Custom view only filters which services the dashboard shows. It sends no email.
- D. Message center preferences that email a weekly digest of Copilot posts A weekly digest summarizes Message center announcements once a week. It can't alert the team when the service or a major function goes down.
Memory hook: Incident: down. Advisory: degraded, often with a workaround. Choose under Customize, then Email.
Microsoft Learn: View service health
Fabrikam creates new claims-processing agent identities from one agent identity blueprint every week. Security wants a single Conditional Access policy that blocks every claims agent, current and future, from a finance API registered in Microsoft Entra ID. No step may be required per new agent, and agents built from other blueprints must stay out of scope. How should the policy's agent assignment be set?
Correct answer: C. Include the claims blueprint so every agent identity derived from it is covered.
The Conditional Access for agents page says a policy applied at the agent identity blueprint level automatically covers all agent identities derived from it, including new ones added in the future. In the policy builder, the blueprint is picked on the Agent blueprint principals tab of the agent selection. That meets both constraints: no action per new agent, and agents from other blueprints are never in scope.
Why the other options are wrong:
- A. Include agents by a custom security attribute that is set on each new claims agent. As written, this needs the attribute set on every new claims agent, which is the per-agent step the requirement forbids; targeting the blueprint covers new agents with no extra step.
- B. Include each existing claims agent identity, then add new ones every week. Picking identities one by one never covers agents created after the policy is saved, so every new agent needs a manual edit.
- D. Include All agent identities so every current and future agent is covered. All agent identities also blocks agents built from every other blueprint, which the requirement rules out.
Memory hook: Target the blueprint to cover today's agents and tomorrow's from the same template.
Microsoft Learn: Agent ID
Woodgrove Bank's change board wants most users to receive deferred-capable Copilot updates later so it can validate them first. A 40-member IT pilot security group must get the same updates as soon as they reach general availability. The admin opens Copilot release preferences: General Availability in Copilot settings. What should she configure?
Correct answer: A. Select Deferred Release, then add the IT pilot group as an exception so it stays on Standard Release.
Copilot release preferences offer Standard Release, the default, and Deferred Release, which holds major deferred-capable updates for 30 days after the Standard rollout begins. You choose the option most users should get, then add up to 100 exceptions to the other one, and each member of an excepted group counts toward that limit. Deferred Release with the 40-member pilot group excepted to Standard gives the pilot general availability timing while everyone else waits.
Why the other options are wrong:
- B. Select Deferred Release, then set Copilot Frontier to specific user groups and add the IT pilot group. Frontier gives early access to preview features before they're generally available. It doesn't place the pilot group on the Standard Release timing for GA updates.
- C. Select Standard Release, then add the IT pilot group as an exception so it moves to Deferred Release. This mirrors the right design but inverts it: everyone stays on Standard Release and only the IT pilot group is delayed, the opposite of what the change board asked for.
- D. Select Standard Release, then give the IT pilot group targeted release under Organization profile. Leaving the tenant on Standard Release delays nobody, and targeted release under Organization profile belongs to the older model that doesn't govern deferred-capable Copilot updates.
Memory hook: Pick the ring most users ride, then except the rest: up to 100 users, each group member counted.
Microsoft Learn: Configure release options
A Reports Reader at Litware exports the Copilot usage report and finds anonymized values where user names should be. The privacy office approves showing names so managers can follow up with inactive users. What change reveals the names?
Correct answer: B. Have a Global Administrator clear the concealment option in Org settings, then Services, then Reports.
The usage reports overview says all reports hide usernames, display names, groups, and sites by default, and Global administrators can reveal them by clearing Conceal user, group, and site names in all reports under Settings, then Org Settings, then Services, then Reports. The setting is tenant-wide: it also governs the Microsoft Graph reports API, Power BI, and the Teams admin center reports, and showing identifiable user information is a logged event in the Purview audit log.
Why the other options are wrong:
- A. Have an AI Administrator turn off name concealment under Copilot, then Settings, then Data access. Concealment isn't a Copilot setting. It's one org-wide setting under Org settings that Global administrators change.
- C. In the Copilot usage report, open Choose columns and add the User name and Display name columns. Choose columns adds or removes columns, but the names in them stay concealed while the org setting is on.
- D. Pull the same data from the Microsoft Graph reports API, which the admin center setting doesn't cover. The same setting also applies to the Microsoft 365 usage reports in Microsoft Graph and Power BI, so the API returns concealed values too.
Memory hook: Names are hidden in every report by default, and one Global Admin setting in Org settings reveals them everywhere.
Microsoft Learn: Activity reports