A is correct: This prompt includes all four recommended elements: a clear goal (draft a summary), context (Q3 results for the leadership team), expectations (one page, formal tone, bullet points), and a source (/Q3-Review.pptx). Including these elements produces more targeted, relevant results. Reference: https://support.microsoft.com/en-us/topic/learn-about-copilot-prompts-f6c3b467-f07c-4db1-ae54-ffac96184dd5

B is incorrect: This prompt is overly broad and unfocused, covering every possible topic without specifying a source file. Effective prompts narrow the scope with a specific goal, a defined audience, referenced source files, and clear format expectations. Reference: https://support.microsoft.com/en-us/topic/get-better-results-with-copilot-prompting-77251d6c-e162-479d-b398-9e46cf73da55

C is incorrect: This prompt has a goal but lacks format expectations and references all SharePoint data instead of a specific file. Copilot does not automatically search all files in SharePoint. Effective prompts reference specific sources using the / feature. Reference: https://support.microsoft.com/en-us/topic/learn-about-copilot-prompts-f6c3b467-f07c-4db1-ae54-ffac96184dd5

D is incorrect: This prompt includes a goal and source but lacks audience context and delegates all formatting decisions to Copilot. Setting clear expectations for format, length, and tone produces more predictable and useful output. Reference: https://support.microsoft.com/en-us/topic/get-better-results-with-copilot-prompting-77251d6c-e162-479d-b398-9e46cf73da55

A is incorrect: Copilot does not automatically detect content from other open documents or browser tabs. You must explicitly reference files using the / command or the Add content button within the Draft with Copilot prompt. Reference: https://support.microsoft.com/en-us/office/draft-and-add-content-with-copilot-in-word-069c91f0-9e42-4c9a-bbce-fddf5d581541

B is incorrect: Pasting text provides some context but is inefficient and limited by prompt length constraints. Directly referencing files with the / command gives Copilot access to the complete source documents for more comprehensive grounding. Reference: https://support.microsoft.com/en-us/office/draft-and-add-content-with-copilot-in-word-069c91f0-9e42-4c9a-bbce-fddf5d581541

C is incorrect: Copilot does not automatically search your SharePoint environment for relevant files. You must explicitly reference specific files using the / command to direct Copilot to the correct sources for generating your draft. Reference: https://support.microsoft.com/en-us/office/draft-and-add-content-with-copilot-in-word-069c91f0-9e42-4c9a-bbce-fddf5d581541

D is correct: Using the / file reference tells Copilot exactly which files to draw from when drafting. When you provide references, Copilot delivers more relevant, accurate drafts grounded in your company's specific terminology, data, and KPIs, reducing generic responses. Reference: https://support.microsoft.com/en-us/office/draft-and-add-content-with-copilot-in-word-069c91f0-9e42-4c9a-bbce-fddf5d581541

A is incorrect: Public web content returns generic information, not your internal Confluence articles. Use a connector to ground the agent in that specific repository. Reference: https://learn.microsoft.com/en-us/microsoft-copilot-studio/knowledge-add-public-website

B is incorrect: Pasting content into each prompt is error-prone and inconsistent. A connector lets the agent draw directly from the Confluence repository. Reference: https://learn.microsoft.com/en-us/microsoft-365-copilot/extensibility/overview-business-applications

C is correct: When authoritative content lives in a non-Microsoft system, a connector lets the agent ground its answers in that external source without duplicating the data. This keeps the agent aligned with the existing system of record. Reference: https://learn.microsoft.com/en-us/microsoft-365-copilot/extensibility/overview-business-applications

D is incorrect: Manually exporting and re-uploading content is unsustainable and quickly becomes outdated. A connector grounds the agent in the live external source without duplication. Reference: https://learn.microsoft.com/en-us/microsoft-copilot-studio/knowledge-add-file-upload

A is incorrect: A subject line does not contain the negotiation details, so the summary would lack the agreed terms. Use Summarize on the thread itself to ground the response in the full message content. Reference: https://support.microsoft.com/en-us/office/summarize-an-email-thread-with-copilot-in-outlook

B is correct: Copilot in Outlook can summarize an open email thread, using the messages in that conversation as the grounding source. This produces a concise recap of the agreed terms directly from the relevant emails. Reference: https://support.microsoft.com/en-us/office/summarize-an-email-thread-with-copilot-in-outlook

C is incorrect: Manually reassembling messages risks omitting context and introducing errors. Use Summarize on the actual thread so Copilot grounds the summary in the complete conversation. Reference: https://support.microsoft.com/en-us/office/summarize-an-email-thread-with-copilot-in-outlook

D is incorrect: Summarizing the whole inbox is unfocused and unlikely to surface the specific agreed terms. Summarize the specific thread so Copilot grounds its response in the relevant messages. Reference: https://support.microsoft.com/en-us/office/summarize-an-email-thread-with-copilot-in-outlook

A is incorrect: Manually issuing copies on request is inefficient and creates version drift. Agent builder's option to share with specific people or a security group gives the sales team consistent, ongoing access. Reference: https://learn.microsoft.com/en-us/microsoft-365/copilot/extensibility/agent-builder-share-manage-agents

B is incorrect: Agents are shared through agent builder's sharing options, not distributed as source files for individual rebuilding. Rebuilding per user causes inconsistency. Share with specific people or a group instead. Reference: https://learn.microsoft.com/en-us/microsoft-365/copilot/extensibility/agent-builder-share-manage-agents

C is incorrect: Sharing with the whole organization makes the agent available to everyone, not just the sales team. The requirement is to limit access, so sharing with specific people or a security group is the correct, controlled approach. Reference: https://learn.microsoft.com/en-us/microsoft-365/copilot/extensibility/agent-builder-share-manage-agents

D is correct: Agent builder lets you share an agent with specific people or groups. Choosing the sales team's members or a security group restricts access to that audience while still letting them use the agent in Word and other Copilot experiences. Reference: https://learn.microsoft.com/en-us/microsoft-365/copilot/extensibility/agent-builder-share-manage-agents

A is incorrect: Resubmitting the same prompt without additional context is unlikely to address the specificity issue. Copilot produces better results when you refine your prompts by adding details about the audience, purpose, and scope of the summary. Reference: https://support.microsoft.com/en-us/office/create-a-summary-of-your-document-with-copilot-in-word-79bb7a0a-3bf7-41fe-8c09-56f855b669bf

B is incorrect: The legacy AutoSummarize feature is not the recommended approach for generating management summaries in Microsoft 365. Copilot's iterative prompting capability, where you add context about audience and focus areas, is designed for this purpose. Reference: https://support.microsoft.com/en-us/office/create-a-summary-of-your-document-with-copilot-in-word-79bb7a0a-3bf7-41fe-8c09-56f855b669bf

C is incorrect: Manual deletion followed by a grammar check does not leverage Copilot's ability to generate a more focused summary. A follow-up prompt with specific context about the executive audience and desired focus areas produces a more relevant and complete result. Reference: https://support.microsoft.com/en-us/office/create-a-summary-of-your-document-with-copilot-in-word-79bb7a0a-3bf7-41fe-8c09-56f855b669bf

D is correct: Copilot works best with iterative back-and-forth conversation. Providing details about the audience, the purpose, and the specific focus areas in a follow-up prompt helps Copilot refine the summary to be more targeted and useful for management. Reference: https://support.microsoft.com/en-us/office/create-a-summary-of-your-document-with-copilot-in-word-79bb7a0a-3bf7-41fe-8c09-56f855b669bf

A is incorrect: Prompt injection manipulates the AI's behavior through processed content; it is not a credential-theft technique that hands over sign-in details or remote account control. Reference: https://www.microsoft.com/en-us/msrc/blog/2025/07/how-microsoft-defends-against-indirect-prompt-injection-attacks

B is correct: Indirect prompt injection is dangerous because the AI operates with the signed-in user's access; a successful attack can issue commands as if they were the user and exfiltrate data the user is permitted to see. Reference: https://www.microsoft.com/en-us/msrc/blog/2025/07/how-microsoft-defends-against-indirect-prompt-injection-attacks

C is incorrect: The risk is unauthorized actions and data exposure, not performance degradation. Performance is not the defining concern of a prompt injection attack. Reference: https://www.microsoft.com/en-us/security/blog/2024/04/11/how-microsoft-discovers-and-mitigates-evolving-attacks-against-ai-guardrails/

D is incorrect: Injection manipulates a specific interaction using planted content; it does not retrain or permanently alter the shared foundation model for all users. Reference: https://learn.microsoft.com/en-us/microsoft-365/copilot/extensibility/declarative-agent-architecture

A is incorrect: Copilot does not selectively delete the largest values as a standard behavior; the omission here was caused by an instruction planted in the file, which indicates injection. Reference: https://learn.microsoft.com/en-us/copilot/microsoft-365/microsoft-365-copilot-transparency-note

B is correct: A directive planted in external content that tells the AI to suppress specific information is prompt injection; it manipulates Copilot's output to mislead the user. Recognizing the altered summary and the planted note is key. Reference: https://www.microsoft.com/en-us/security/blog/2024/04/11/how-microsoft-discovers-and-mitigates-evolving-attacks-against-ai-guardrails/

C is incorrect: Fabrication means inventing content that is not there. Here, real figures were intentionally omitted because of a hidden directive, which is prompt injection rather than a hallucination. Reference: https://learn.microsoft.com/en-us/copilot/microsoft-365/microsoft-365-copilot-transparency-note

D is incorrect: Data protection controls restrict access to labeled or sensitive content; they do not silently strip figures based on a note inside an external file. The cause is a planted instruction, not a protection policy. Reference: https://learn.microsoft.com/en-us/copilot/microsoft-365/microsoft-365-copilot-privacy

A is correct: When generating a presentation, describing the topic along with the number of slides and the tone guides Copilot to produce a deck that matches your length and style needs, reducing the editing required afterward. Reference: https://support.microsoft.com/en-us/office/create-a-new-presentation-with-copilot-in-powerpoint-3222ee03-f5a4-4d27-8642-9c387ab4854d

B is incorrect: A vague one-word prompt produces unfocused output and leaves all length and tone control to manual cleanup. Specifying the slide count and tone up front is more effective. Reference: https://support.microsoft.com/en-us/office/create-a-new-presentation-with-copilot-in-powerpoint-3222ee03-f5a4-4d27-8642-9c387ab4854d

C is incorrect: Requesting the longest possible deck works against your time constraint and produces an overly detailed result. Stating the number of slides and tone in the prompt gives you a deck closer to what you need. Reference: https://support.microsoft.com/en-us/office/create-a-new-presentation-with-copilot-in-powerpoint-3222ee03-f5a4-4d27-8642-9c387ab4854d

D is incorrect: Designer adjusts visual layout, not the tone of text or the number of content slides. Length and tone are best controlled by including them directly in your Copilot prompt. Reference: https://support.microsoft.com/en-us/office/create-a-new-presentation-with-copilot-in-powerpoint-3222ee03-f5a4-4d27-8642-9c387ab4854d

A is correct: Usernames and passwords are secrets that grant system access. Placing them into a prompt and generated content risks exposing credentials in retained or shared material, which could enable unauthorized access. Credentials should never be entered into prompts. Reference: https://learn.microsoft.com/en-us/microsoft-365/copilot/enterprise-data-protection

B is incorrect: Over-reliance is about accepting AI output without verification. The actual risk here is that sensitive authentication credentials were entered into a prompt and may be exposed, which is a data exposure concern. Reference: https://learn.microsoft.com/en-us/microsoft-365/copilot/microsoft-365-copilot-architecture-data-protection-auditing

C is incorrect: Fabrication refers to AI producing false content. This scenario involves a user supplying real credentials, creating a credential-exposure risk rather than a fabrication risk. Reference: https://learn.microsoft.com/en-us/microsoft-365/copilot/microsoft-365-copilot-privacy

D is incorrect: A password is an access credential, not intellectual property or proprietary software. The risk is the exposure of authentication secrets, not the loss of intellectual property. Reference: https://learn.microsoft.com/en-us/microsoft-365/copilot/security-microsoft-365-copilot