Microsoft AB-730 Practice Questions with Explanations

Correct answer: C

When a prompt references a file that the user does not have access to, Copilot cannot retrieve or process that file content, resulting in a 'prompt not found' error. This is the most common cause of access failures for shared prompts. An outdated prompt would still be technically findable, prompts outside the organization may work if shared appropriately, scheduled prompts have different failure modes, and sensitivity labels affect output behavior differently than access denial.

Why the other options are wrong:

• A. An outdated prompt would still be locatable even if its referenced content has changed.
• B. Prompts outside the organization can function if properly shared through cross- organization permissions.
• D. Scheduled prompts fail at execution time with different error messages, not during link resolution.
• E. Sensitivity labels restrict output handling but do not cause 'prompt not found' errors.

Correct answer: D

When Copilot reports that a file appears empty, corrupted, or in an unprocessable format, the most likely cause is that the user does not have access to the file, preventing Copilot from reading its contents properly. Without access permissions, Copilot cannot verify the file's actual state. Scheduling is not relevant to the execution failure, the choice of app does not typically cause this error, and using the wrong agent would produce a different error message.

Why the other options are wrong:

• A. Not scheduling a prompt is unrelated to processing errors that occur during execution.
• B. Running from a web app versus desktop app does not typically cause file corruption detection errors.
• C. Using the wrong agent would produce a different error indicating agent incompatibility rather than file processing issues.

Correct answer: C

Content credentials are metadata that can verify whether an image was generated by AI tools like Microsoft 365 Copilot by embedding provenance information about the creation process and tool used. This is the standard verification method for AI-generated content. File names are user-assigned labels with no verification capability, watermarks can be added to or removed from any image, and file descriptions are user-created text without authentication properties.

Why the other options are wrong:

• A. File names are arbitrary user-assigned labels that provide no reliable verification of image origin.
• B. Watermarks can be added to or removed from images and do not provide verified authenticity.
• D. File descriptions are user-written metadata that cannot authenticate the actual generation method of an image.

Correct answer: B

Overreliance occurs when a user depends too heavily on AI output without verification or critical analysis. The scenario describes placing a large order based solely on Copilot suggestions without reviewing them, checking market trends, or applying human judgment. This is the definition of overreliance on AI recommendations.

Why the other options are wrong:

• A. Verification is the opposite of what occurred here-the user explicitly did NOT verify the suggestions.
• C. Fabrication refers to AI generating false information, not to a user's decision-making process based on AI output.
• D. Prompt injection is a security attack involving malicious input to manipulate AI behavior, not relevant to this scenario.

Correct answer: B

Copilot Pages are available in both Microsoft 365 Copilot and Microsoft 365 Copilot Chat, allowing users to create and manage collaborative documents across both versions. The Researcher agent and Copilot Notebooks are specific features not available in both versions.

Why the other options are wrong:

• A. The Researcher agent is not available in all versions of Microsoft 365 Copilot.
• C. Copilot Notebooks is a specific feature not uniformly available across both Copilot versions.

Correct answer: A

When using the Work scope, Microsoft 365 Copilot operates under the principle of least privilege, using only data the user can access combined with the general knowledge from its training data. This respects organizational security boundaries while providing context- aware assistance. Users cannot access organization-wide data beyond their permissions, and Copilot alone uses only accessible data.

Why the other options are wrong:

• B. Copilot cannot access all organizational data; it respects user permissions and access controls.
• C. Copilot also incorporates general knowledge from its training data, not just user- accessible data.
• D. Copilot uses both user-accessible data and general training knowledge, not training data alone.

Correct answer: E

In the Conversations section of the left sidebar, the 'Standings' conversation has a green dot indicator next to it. This green dot is the visual indicator in M365 Copilot that signifies a conversation has been shared. The other conversations listed (Auckland Time Zone, 2025 Canadian and U.S. National, PTO policies, and Overview of Corporate Procedures) do not have this green dot indicator, making Standings distinctly marked as a shared conversation.

Why the other options are wrong:

• A. There is no indication in the interface that the conversation contains only anonymous data; the green dot does not represent anonymity.
• B. A verified response would be indicated by a checkmark or verification badge within the conversation content, not by a green dot next to the conversation title.
• C. If a conversation were added to a notebook, it would typically show a notebook icon or be listed under the Notebooks section, not just indicated by a green dot.
• D. A scheduled prompt would require specific scheduling interface elements or indicators; a green dot does not represent scheduled prompts in this interface.

Correct answer: A

The Search feature in Microsoft 365 Copilot is designed to query existing content and data within the organization to find similar or related items. To check for existing training plans similar to a newly generated one, the user should use Search to locate comparable plans across the organizational workspace.

Why the other options are wrong:

• B. Designer is used for creating visual content, not for searching existing organizational documents.
• C. Apps refers to integrated applications and extensions, not a search function for locating similar documents.
• D. Pages is used for creating collaborative documents, not for searching existing organizational content.

Correct answer: A

Citations are the mechanism through which the Researcher agent provides transparency and verifiability. They link generated content back to the original sources used, allowing users to validate that information came from legitimate, traceable origins and to assess source quality.

Why the other options are wrong:

• B. Instructions relate to how to use Copilot, not to verifying the validity of sources used in reports.
• C. Memory stores user preferences and details but does not verify source validity.
• D. Capabilities describe what Copilot can do, not the mechanism for verifying sources in generated reports.

Correct answer: B

Fabrication is when an AI system generates false, made-up, or hallucinated information that appears plausible but is not factual. The scenario explicitly describes discovering that some information in the Copilot-generated report is fictional, which is the definition of fabrication.

Why the other options are wrong:

• A. Deepfake refers to synthetic media (video/audio) that impersonates real people, not false text information.
• C. Overreliance is about trusting AI output without verification; this describes the discovery of false content itself.
• D. Prompt injection is a security attack using malicious prompts, not relevant to false information generation.
• E. Bias refers to systematic favoritism or prejudice in AI output, not the generation of fictional information.