Microsoft AI Business Professional (AB-730)

D is correct.

Explanation: Cowork takes AI further by acting across your Microsoft 365 apps on your behalf. It handles multi-step, autonomous work that spans multiple apps and carries out connected steps in sequence. Use Cowork when a request has grown beyond a single answer and you need Copilot to carry out several connected steps and hand back a finished result.

A is incorrect: Chat helps you finish an answer or draft in a single session. This request requires connected actions across several different applications.

B is incorrect: Code builds a reusable tool, such as a tracker or dashboard, from your description. The request is to complete tasks, not tools.

C is incorrect: Autopilot provides ongoing background progress on repeating work. This request is a one-time connected sequence that you want finished and handed back.

A is correct.

Explanation: Agents grounded in instructions or public websites are included at no additional cost where enabled. In Agent Builder, instructions define the response format, and Knowledge can attach public websites. Because the team has no Microsoft 365 Copilot license and no metered billing, they must avoid organizational sources. Administrator settings can still affect agent creation, so verify availability.

B is incorrect: Outlook and Teams are organizational sources. Agents using them need metered consumption billing or a Microsoft 365 Copilot license, which this specific team lacks.

C is incorrect: Cloud files under Work content are organizational sources. They require metered consumption billing or a Microsoft 365 Copilot license, so they do not fit here.

D is incorrect: Copilot connectors are Work content sources. Using them makes this an organizational-source agent, which needs metered consumption billing or a Microsoft 365 Copilot license.

A is correct.

Explanation: Copilot Memory saves preferences such as bullet points or a professional tone and applies them automatically to future responses, so you stop repeating instructions. You can also view or delete saved memories at any time from Settings > Personalization, which meets both requirements here without changing which data sources Chat uses to generate responses.

B is incorrect: Work IQ controls whether Chat reasons over organizational content such as files and emails. It does not store formatting preferences for responses.

C is incorrect: Copilot Pages turns a response into a shared, editable page. It does not save tone or format preferences for future conversations.

D is incorrect: A temporary chat keeps a conversation from being remembered. That is opposite to the goal of saving preferences across chats.

C is correct.

Explanation: By default, Chat uses Auto mode to select a model for each prompt. You can use the model selector to choose Think deeper for complex questions that benefit from extended reasoning. This matches the request because the question is complex and the user wants more reasoning than the default selection provides in this scenario.

A is incorrect: Auto is the default and selects a model for the prompt. Keeping it does not choose extended reasoning for this question.

B is incorrect: Quick response is intended for straightforward requests. The question here is complex and needs extended reasoning, so this mode does not match.

D is incorrect: A suggested prompt helps you get started or refine a response. It is not a response mode in the model selector.

A is correct.

Explanation: Generative AI relies on models trained on large collections of existing content to identify patterns. During generation, the model interprets your input and uses those learned patterns to build an output. Because a generative model creates rather than retrieves a fixed response, its output can vary even when prompts are similar. This explains the differing drafts.

B is incorrect: Classification models identify and categorize information. Generative models create new content based on learned patterns instead of returning a stored category response.

C is incorrect: Training is a separate step where the model learns patterns. The model does not update its training data after every single prompt.

D is incorrect: The model interprets your input and uses learned patterns to build an output. Variation comes from creating content, not ignoring the prompt.

C is correct.

Explanation: A Microsoft 365 Copilot license gives Chat permission-aware grounding across meetings, chats, and files. When the reference is available, # followed by a meeting name directs Chat to that meeting, so the user can ask for a to-do list without uploading or opening anything. Access also depends on organization settings, so contact your IT administrator if the option isn't available.

A is incorrect: Using @ with a person's name directs Chat to specific content about that person, such as their emails, chats, and files, rather than meetings.

B is incorrect: Using / with a file name directs Chat to a specific file. The task involves drafting a list from a meeting, not a file.

D is incorrect: Selecting + uploads a file for Chat to analyze. The user explicitly wants no upload, and provided content does not give broad organizational access.

C is correct.

Explanation: Copilot can generate responses that sound confident but contain incorrect facts, numbers, or citations, so treat every response as a draft. Before sharing it externally, open each cited source and confirm that it exists, is trustworthy, and supports the claim. The analyst cannot verify the figure from memory, so checking the source is the right first step before the partner receives the draft.

A is incorrect: Asking Copilot to confirm itself does not verify the figure against a trusted source, and Copilot can produce incorrect numbers and citations.

B is incorrect: A citation does not guarantee accuracy. Citations can be incorrect, so you must check that the source supports the claim.

D is incorrect: Removing the citation hides the problem and leaves the statistic unverified. You remain responsible for the final content you share.

C is correct.

Explanation: Changing topics abruptly is a common prompting mistake. The best practice is to finish one task before starting a new one. Here, the user should complete the budget request, refine it with feedback if needed, and then send the team event request as a separate prompt so Copilot Chat receives one clear task at a time.

A is incorrect: Slang and jargon are common mistakes. Informal language can cause low-quality responses, so it does not fix the problem of changing topics abruptly.

B is incorrect: Combining both requests with different tone rules adds conflicting information. Multiple or conflicting pieces of information in one prompt should be avoided.

D is incorrect: Feedback helps Copilot understand your needs. Withholding it removes a best practice and does not address the abrupt change of topic.

D is correct.

Explanation: People with access to a page can review and edit its content in real time. They do not gain access to the Chat conversation that created the page. Creating, editing, and sharing are core Pages workflows. A Microsoft 365 Copilot license is only required to use Copilot directly within a shared page for AI-assisted collaboration. This colleague can edit manually.

A is incorrect: Shared access covers the page content only. The colleague does not gain access to the Chat conversation, so the original conversation remains completely unavailable.

B is incorrect: Pages lets people with access edit in real time. Exporting to Word creates a standalone file and is not required for basic real-time collaboration.

C is incorrect: Creating, editing, and sharing pages are core Pages workflows. The license requirement applies to using Copilot within a shared page, not for manual editing.

A is correct.

Explanation: Read Aloud reads a Copilot response aloud. You select the Read Aloud option for the specific response you want to hear. This fits the need to listen to an existing response without changing how you enter prompts or starting a spoken conversation with Copilot. It also gives an alternative to reading the response on screen.

B is incorrect: Real-time voice starts a spoken conversation with Copilot and requires a Microsoft 365 Copilot license. The user wants only to hear an existing response.

C is incorrect: Dictation converts your spoken words into a written prompt. It handles input, not output, so it does not read an existing response to you.

D is incorrect: The bookmark icon saves a prompt for reuse in Prompt Lab. It does not read a response aloud or produce audio for any response.

C is correct.

Explanation: Use # followed by the meeting name to reference a recurring event in your prompt. The user has a Microsoft 365 Copilot license, which references to meetings require. Copilot Chat can then focus on that meeting when it identifies action items. This gives Chat a defined source instead of relying on a general description of the meeting.

A is incorrect: The / symbol references a file or folder. It does not reference a meeting or recurring event, which requires the # symbol instead.

B is incorrect: Plain text omits a reference symbol. The user wants to reference the meeting directly, which requires typing # before the meeting name.

D is incorrect: Mentioning the organizer with @ includes that person's perspective for collaboration. It does not retrieve the meeting by name or recurring event.

A is correct.

Explanation: Deleting a conversation does not remove memories saved from it. Open Personalization in Chat settings, view saved memories, and delete the specific memory you no longer want Chat to use. This removes only that detail and keeps other saved memories in place. Delete all memories would clear everything, which is broader than needed.

B is incorrect: The conversation is already deleted, and deleting a conversation does not remove memories saved from it. The memory must be deleted from personalization settings.

C is incorrect: Turning off saved memories stops Chat from adding new memories. It does not delete existing memories, so the saved detail would remain.

D is incorrect: A temporary chat avoids saved memories for that session only. It does not delete the memory, and the detail applies again in normal chats.

B is correct.

Explanation: Custom instructions are a one-time setup that tells Chat how to respond across all conversations. Select Custom instructions in Personalization settings and enter your role and preferences, such as concise responses with bullet points. After you save them, they apply automatically, so each prompt can focus on the task. You can update or clear them later when your role changes.

A is incorrect: A saved prompt reuses one specific request from Prompt Lab. It does not apply role and format preferences automatically to every conversation.

C is incorrect: A temporary chat does not access your custom instructions or saved memories. It would remove the preferences the user wants applied.

D is incorrect: Turning off chat history stops Chat from using past conversations to personalize responses. It does not store a role or format preference.

C is correct.

Explanation: You cannot edit a saved prompt in place. Run or copy the prompt from Prompt Lab, make your changes in the chat input, and save it as a new prompt. Then delete the original to avoid duplicates. This leaves the updated version under the Your prompts tab. The user wants to revise the details without keeping two similar prompts.

A is incorrect: Saved prompts cannot be edited in place from Your prompts. Changes must be made in the chat input and saved as a new prompt.

B is incorrect: Sharing sends a prompt to a team or provides a direct link. It does not provide a way to edit the saved prompt in place.

D is incorrect: The documented process saves a new prompt after you enter a name. The original remains until you delete it, so duplicates would remain.

B is correct.

Explanation: By default, Chat draws on real-time web information. Grounding refers to the source data Copilot uses to shape its responses, so Chat does not automatically access your organization's files, emails, or meetings. It can draw on that work context only after you connect it, so last week's internal meeting is not available in the default configuration.

A is incorrect: Automatic access to meetings does not happen by default. Chat uses files, emails, and meetings only when connected to your work context.

C is incorrect: Chat does not switch modes on its own. Cowork is a separate mode for multi-step work, not a fallback for retrieving meetings.

D is incorrect: Frontier preview enrollment applies to Code and Autopilot. Chat does not require it and is available in browsers and Microsoft 365 apps.

D is correct.

Explanation: The prompt keeps the same goal, context, and expectations and changes only the source to organizational data. It names emails, meetings, chats, and files for Project Atlas, and it sets the audience and a five-bullet format. Organizational data requires a Microsoft 365 Copilot license, which this user has. The clear source and format make the request specific.

A is incorrect: This prompt is vague. It gives no source, audience, or format, so Copilot Chat has less direction on what to summarize for leadership.

B is incorrect: This prompt uses supplied content, an attached brief and report. The user has no attachments and wants to use organizational data instead.

C is incorrect: This prompt gives conflicting instructions by asking for both bullet points and an essay. Multiple or conflicting pieces of information should be avoided.

C is correct.

Explanation: Enterprise Data Protection applies only when you sign in to Chat with your work account. The shield icon next to the Copilot icon shows when you are signed in to Chat with enterprise data protection. Signing in with your work account at m365copilot.com is the correct action when the shield icon is missing and protection is required.

A is incorrect: Think deeper selects a response mode for complex questions. It does not change the sign-in account or the enterprise data protection status.

B is incorrect: The Work IQ button controls whether Chat can use work data. It does not sign you in with a work account.

D is incorrect: The Edge side pane also requires a work account sign-in for protection. Changing the access location does not fix the account problem.

B is correct.

Explanation: Chat can analyze files that you explicitly provide, including Excel workbooks. Select the + button or drag the file into the prompt box, then ask for insights. This works without broad organizational grounding. Uploaded files are stored in your OneDrive, and file upload limits apply. The user can delete the uploaded file at any time.

A is incorrect: Using / with a file name depends on broad organizational grounding from a Microsoft 365 Copilot license, which this user doesn't have.

C is incorrect: Using @ with a person's name directs Chat to specific organizational content. It needs the same broader, license-based grounding that this user currently lacks.

D is incorrect: Chat analyzes files that you explicitly provide. Without broad organizational grounding, it can't search all organizational files, so the workbook isn't supplied.

D is correct.

Explanation: A temporary chat does not access your saved memories or custom instructions. The conversation does not appear in chat history, and no memories are saved from the session. This gives a neutral response without changing your personalization settings. Temporary chat data may still be retained according to your organization's data retention policy. Use the Start a new temporary chat option.

A is incorrect: Deleting the conversation afterward happens after Chat has already used your memories and instructions. It does not remove the memories saved from that conversation.

B is incorrect: Turning off saved memories stops new memories from being added. It does not delete existing memories, so they remain stored and affect the response.

C is incorrect: Clearing custom instructions changes your standing preferences for later conversations. It is a settings change, and the user wants a response without changing settings.

B is correct.

Explanation: The Work IQ button turns access to work data on or off without leaving the conversation. With Work IQ off, Chat does not use Work IQ to ground its response, but it can still use public web data and content you provide directly. Check the button before you enter the prompt so you know what Chat can use.

A is incorrect: Quick response is a response mode for straightforward requests. It does not control whether Chat uses Work IQ grounding for organizational content.

C is incorrect: A temporary chat keeps a conversation from being remembered. It does not turn off Work IQ access for the prompt you enter.

D is incorrect: Deleting a conversation removes it from your history and search results. It does not change which data sources Chat can use.

B is correct.

Explanation: After you edit suggested prompts for a shared agent, anyone with access must reinstall the agent to see the updated prompts. The agent is already shared, so each person with access should reinstall it instead of changing the sharing setup. Suggested prompts show others how to interact with the agent, such as asking to summarize a report.

A is incorrect: A Viewer can use the agent and an Editor can modify it. Changing a role does not update the suggested prompts that your colleagues see.

C is incorrect: Recreating the agent in Agent Builder is not needed. The required step is reinstalling the shared agent, and a new agent would remain completely private.

D is incorrect: Making the agent available to the entire organization changes who can use it. It does not refresh the suggested prompts for people with existing access.

B is correct.

Explanation: The prompt already includes context, a goal, and expectations. The missing element is the source. Define where Copilot Chat should look, such as web content, text or files you provide, or organizational data available to you. Specifying the source helps Copilot Chat focus its response. Organizational data requires a Microsoft 365 Copilot license. This keeps the other details unchanged.

A is incorrect: The prompt already states the audience and tone, which are expectations. Repeating them does not tell Copilot Chat where to find the information.

C is incorrect: The purpose of the meeting is context, and the prompt already includes it. Adding more context does not define the information source.

D is incorrect: The prompt already has a goal with a five-bullet summary output. Changing the format does not address the missing source element.

B is correct.

Explanation: Scheduled prompts require a Microsoft 365 Copilot license and connected experiences enabled for your account. After the prompt runs, select Schedule this prompt next to it, choose daily at 8 AM, and select Save. You can create up to 10 scheduled prompts. If your organization disables connected experiences, scheduling is not available. Scheduling lets Chat run the prompt without her being online.

A is incorrect: Saving and sharing store or distribute a prompt. They do not run it automatically at a set time while the user is offline.

C is incorrect: Run now executes the prompt immediately. It does not run the prompt automatically at 8 AM while the user is offline.

D is incorrect: Without a Microsoft 365 Copilot license, scheduled prompts are not available. Connected experiences alone do not meet the strict license requirement for this feature.

B is correct.

Explanation: Platform safeguards reduce risk, but you still decide what to share with Copilot. You must share only what the task needs, such as an excerpt instead of an entire document. You must also respect restrictions on files labeled confidential and handle content according to your organization's policies. Here, only the renewal terms are needed for the summary.

A is incorrect: Pasting the whole 60-page file shares more than the task needs. You should paste an excerpt instead of an entire document.

C is incorrect: Enterprise data protection applies controls, but you remain responsible for respecting sensitivity labels. Available controls vary by subscription and organizational configuration.

D is incorrect: You must follow restrictions shown on labeled files. Removing the label bypasses those restrictions and violates your organization's data handling policies.

D is correct.

Explanation: Deleting all activity history in the My Account Portal removes Copilot conversations across your devices, and the deletion cannot be restored. However, if you shared a conversation link with someone, that person can still access it even after you delete your history. The colleague's access is the documented behavior for shared links in this case.

A is incorrect: Usage reports are not affected by deleting history, but that concerns reporting. It does not explain why the shared link remains accessible.

B is incorrect: Deleted conversations cannot be restored. Access through a shared link is separate from restoring a conversation, so this contradicts the documented behavior.

C is incorrect: Delete all activity history removes conversations across your devices, not only the current device, so a device limit does not explain access.

C is correct.

Explanation: In the Copilot pane in Outlook, switching from Default mode to Allow actions lets Copilot take direct actions on your calendar, such as creating the event itself rather than only drafting text. Copilot then proposes the full invite with a title, attendees, and agenda. You review the details and select Approve to add it to your calendar.

A is incorrect: Default mode does not let Copilot take direct actions on your calendar, so copying drafted text leaves you creating the event yourself.

B is incorrect: Starting a recording generates the transcript Copilot needs during a Teams meeting to summarize questions; it does not create calendar events beforehand.

D is incorrect: Summarizing the thread and creating an agenda only helps with meeting preparation; it does not create the calendar event directly for you.

D is correct.

Explanation: Using / followed by the meeting name in Copilot Chat grounds the response in that meeting's content rather than general knowledge. You can then ask for the open blockers and follow-up tasks with owners and due dates, which fits the follow-up email you need to send. This keeps the draft tied to the actual meeting.

A is incorrect: A prompt without the meeting reference is not grounded in that meeting's content, so the draft could rely on general knowledge.

B is incorrect: @Facilitator is typed in the meeting chat and shares the interaction with all participants. It does not ground your follow-up email draft.

C is incorrect: The Recap tab shows the AI summary, but rewriting it manually skips using Copilot Chat with the specific meeting reference for grounding.

C is correct.

Explanation: Files Cowork creates appear in the Output folder of the Details pane, each with Download and Preview options. Many file types preview in the side panel without downloading. When Cowork produces multiple output files, Download all files at the top of the list downloads everything as a single zip archive. Files are saved to your OneDrive.

A is incorrect: The Input folder holds files you provided as context. It does not contain files Cowork created, so it is the wrong location.

B is incorrect: The Scheduled tab lists scheduled prompts with next run time and frequency. It is not where you download files.

D is incorrect: The Skills list shows which skills Cowork loaded during the conversation. It does not provide downloads of the newly created files.

A is correct.

Explanation: Agents are designed for specific types of tasks, topics, or workflows. Benefits and policy questions are a defined area with a repeatable pattern, and an HR agent is a documented example. Agents work within the boundaries of what they were built or configured to do. Choose this when the task fits the agent's purpose rather than an open-ended goal.

B is incorrect: Cowork is an assignment for multi-task goals across apps. A single repeatable policy question does not need multi-step coordination or multiple deliverables.

C is incorrect: Copilot Chat suits single-task work where you steer each turn. It is not the purpose-built option for a defined, repeatable workflow.

D is incorrect: Scheduling a Cowork prompt repeats a delegated assignment on a timetable. It does not provide a helper scoped to answering policy questions.

B is correct.

Explanation: Plan mode makes Copilot propose a plan before it makes any changes, so you can review the proposed work first. This fits the requirement because the workbook stays unchanged until you review the proposal, while Copilot still prepares the multi-step work you requested. The mode selector near the bottom of the pane controls this behavior.

A is incorrect: Allow editing is the default mode that edits the workbook directly, adding sheets and formulas, so changes would be applied without review.

C is incorrect: Chat only responds in the chat pane without touching the workbook, so it would not propose a plan of changes for review.

D is incorrect: The model selector chooses which AI model powers responses. It does not control whether Copilot edits the workbook or proposes a plan.

C is correct.

Explanation: Get Coaching in the Copilot pane reviews your email draft and offers specific suggestions on tone, clarity, and reader sentiment. You can apply each suggestion individually, such as adjusting one sentence, a phrase, or the overall tone, until the message lands as intended. This fits a draft that is already written and needs refinement before you send it to stakeholders.

A is incorrect: A meeting request prompt generates a completely new draft. Your email draft already exists, and you need feedback on it, not another message.

B is incorrect: Chat only mode is a Word feature that keeps responses in the chat pane without changing the document. It is not the Outlook coaching feature.

D is incorrect: The Work and Web toggle controls where Copilot Chat draws from in Outlook. It does not review tone, clarity, or sentiment in your draft.

A is correct.

Explanation: Copilot needs a transcript to provide full meeting insights. Start recording first, but if recording isn't available in your organization, use Transcribe instead. Then select the Copilot icon in the meeting toolbar. Without a transcript, Copilot is limited to chat-based data. This meets the constraint that recording is unavailable and keeps real-time summaries possible.

B is incorrect: Facilitator must be enabled by an IT administrator. The scenario asks for Copilot insights during the meeting, which require a transcript.

C is incorrect: Outlook prompts are for meeting preparation. They do not start the transcript that Copilot in Teams needs for full meeting insights.

D is incorrect: The Recap tab and AI summary are used after the meeting ends. They do not provide real-time help during the active meeting.

A is correct.

Explanation: Always allow is available in the dropdown next to the approve button. It allows the action and skips the prompt for similar actions in the rest of the conversation. The approve button alone lets the action proceed only this one time, so it would not meet the requirement. Always check recipients and content before you approve.

B is incorrect: Selecting the approve button, such as Send, allows the action to proceed this one time. It does not skip later prompts.

C is incorrect: Cancel stops the action. Cowork skips it and continues with the rest of your request, so the email is not sent.

D is incorrect: Pause finishes the current step and then pauses the task, and Resume continues from where it stopped. Neither control changes approval prompt behavior.

A is correct.

Explanation: The exercise requires a Microsoft 365 subscription with a Microsoft 365 Copilot license. Access method is not the limiting factor, because you can use Excel and Copilot through the desktop apps or on the web. Without the license, either access method does not meet the stated requirement, so you must obtain the license.

B is incorrect: The web is a supported way to access Excel and Copilot, but the license requirement still applies. Access alone is insufficient.

C is incorrect: The desktop apps can access Excel and Copilot, but the license requirement still applies. Installing the applications alone is insufficient.

D is incorrect: The requirement is a Microsoft 365 Copilot license, not an Excel version. Changing the application version does not meet the requirement.

C is correct.

Explanation: Cowork acts as you, using your Microsoft Entra ID sign-in and access scope. It can only access data and services your Microsoft 365 account is authorized to use. If you can't see a file, neither can Cowork. Sensitivity labels and DLP policies also apply to all Cowork interactions. Cowork does not get separate or broader access to organizational data.

A is incorrect: Cowork operates inside the same security boundaries as Microsoft Copilot, so it has no separate access to files you cannot open.

B is incorrect: Approval applies to sensitive actions such as sending, sharing, and scheduling. It does not give Cowork permissions your account lacks.

D is incorrect: DLP policies and sensitivity labels apply to Cowork interactions, but they do not grant access. Access follows your account permissions.

B is correct.

Explanation: You can describe both the task and the timing in natural language, such as every Friday at 3 PM. Cowork sets it up, confirms the schedule, and asks for your approval before activating it. You can manage scheduled prompts later from the Scheduled tab in the Tasks view or the Schedule section in the Cowork side panel.

A is incorrect: Sending a new request each week repeats manual work. The Always allow option only skips prompts for similar actions in one conversation.

C is incorrect: Administrators configure the credit limit for the organization. Scheduling a prompt is something you describe to Cowork in natural language.

D is incorrect: Requests can include timing. You describe both the task and the schedule in natural language, and Cowork sets it up.

A is correct.

Explanation: The Analyst agent reads across all attached files and figures out the relevant sheet or column from your prompt. You only need to state a clear goal and scope, such as sales trends by region and quarter. More context in the prompt gives more focused output, and you can use follow-up prompts to drill into specific areas.

B is incorrect: You do not need to specify which sheet or column to use. The agent finds them automatically from your prompt.

C is incorrect: The agent reads across multiple attached files at once, so you do not need to submit them one at a time.

D is incorrect: Less context produces less focused output. A one-word prompt gives the agent no goal or scope for trends by region and quarter.

D is correct.

Explanation: Cowork usage is measured in Copilot Credits. Typing /cost in a conversation shows the credits used by the current task so far, for example 610.6 credits. Reviewing this periodically helps you learn which task types require little work and which require more. This gives a direct view of the task you are running without changing any limits or settings.

A is incorrect: The Details pane shows progress, the step log, and input and output files. The documented way to view task credits is /cost.

B is incorrect: Requesting credits applies when you reach the organization limit, and your administrator decides on approval. It does not report task usage.

C is incorrect: Auto selects a model to balance capability and usage. It affects the model choice but does not show the credits used.

D is correct.

Explanation: An effective Cowork request describes the result you want and names the deliverables. This request gives the customer, the context to pull, and the output to produce, so Cowork can plan the steps itself. A long list of steps is often a signal to reframe the request as an outcome. You can then review the draft briefing.

A is incorrect: This request states an outcome but names no deliverable or context, so Cowork has less information to plan from.

B is incorrect: A step-by-step list tells Cowork how to work instead of what you need. Leading with the outcome lets Cowork plan.

C is incorrect: "Organize everything" is too broad and can lead to unfocused results. It defines no clear boundaries or deliverables for Cowork.

D is correct.

Explanation: Copilot can read a text column, identify sentiment, and insert a new labeled column with its findings. Prompt D names the column, defines the three labels, and asks for a new column, which matches the requirement. Because the prompt is specific, the output is more targeted, and you can aggregate the labels afterward with a follow-up prompt.

A is incorrect: This prompt returns recurring themes with brief descriptions. It does not label each row by sentiment in a new column as required.

B is incorrect: Grouping and counting is a follow-up after Copilot adds a labeled column. No labels exist yet to group in this table.

C is incorrect: This prompt creates a revenue chart by region. It analyzes revenue numbers, not the CustomerFeedback text column for sentiment as required.

B is correct.

Explanation: Copilot connectors bring data from external systems, such as Salesforce, into Microsoft Graph. Copilot can then search and summarize that content alongside your Microsoft 365 files and messages. Because administrators configure connectors, the practical step is to ask your IT team to set this up, rather than changing a prompt. The result includes a summary with a link back to the source record.

A is incorrect: A /filename reference points to files, emails, or meetings in Word or PowerPoint. It does not make external Salesforce records available to Copilot.

C is incorrect: The model selector chooses Auto, Claude, or GPT for a request. It does not connect Copilot to Salesforce data.

D is incorrect: Work IQ grounds Copilot in your role, relationships, and working patterns. It does not replace administrator-configured connectors for external systems.

C is correct.

Explanation: The work spans email, files, and meetings and needs more than one deliverable. Cowork accepts the goal as an assignment, plans the steps, works across your apps, and checks in when it needs input, so you can work on other tasks meanwhile. This matches the multi-task shape described for Cowork, where you describe the outcome instead of managing each step yourself.

A is incorrect: Copilot Chat keeps you in the loop for one prompt and one result at a time, so you would manage each task.

B is incorrect: An agent fits a scoped, repeatable task in a defined domain. It isn't necessarily meant to handle a broad goal across apps.

D is incorrect: Drafting each file separately in Word keeps you steering each turn instead of handing off the goal and moving on.

A is correct.

Explanation: In PowerPoint, you can start a presentation from content that already exists by typing / followed by the filename. A single prompt can reference up to five files at once, including a Word document, an Excel file, and a PDF. This uses your existing content as the source for the draft, ensuring the presentation includes all your Q3 results without needing to describe the content again.

B is incorrect: A prompt without a source document is for starting fresh. Your content already exists in three files, so this ignores them.

C is incorrect: Effort: Faster prioritizes speed for straightforward requests. It does not attach files, so it cannot make Copilot read your three sources.

D is incorrect: Try a new design updates the visual theme of the entire deck in one step. It does not add content from source files.

A is correct.

Explanation: Edit in Pages copies the Copilot response into a new page that you can edit directly or with Copilot. You can prompt Copilot to expand it into a project hub with sections. Pages are saved indefinitely and are accessible from the Pages tab in the Microsoft Copilot app, Teams, and Outlook. This fits the shared reference need.

B is incorrect: A notebook starts from references you add manually. It does not copy this specific Copilot response into a new shared, editable page.

C is incorrect: An audio overview is a spoken summary of notebook content. It is not a shared, editable canvas for tracking ongoing project decisions.

D is incorrect: Selecting Create in a notebook generates a Word document or Excel spreadsheet from references. It does not create a persistent Pages hub.

B is correct.

Explanation: Copilot grounds notebook responses in the notebook's references, but only the first 100 references are used for grounding. With 120 references, the last 20 are outside the grounding set. You must reduce the number of references so the files you need are within the first 100 to ensure answers are directly relevant to your project.

A is incorrect: Supported reference types include Word, Excel, PDFs, images, text files, Teams chats, and emails, so the file type is not the cause.

C is incorrect: Custom instructions guide tone, formatting, language, and focus area across the notebook. They do not define which references Copilot uses for grounding.

D is incorrect: Sharing the notebook with edit access lets members add references, but the access level does not change the 100-reference grounding limit.

D is correct.

Explanation: In Word, you control what Copilot references through the wording of the prompt. Describing an internal source, such as your documents and emails, draws from organizational content. This keeps the request grounded in internal insights, and Copilot can automatically include citations when it draws from your work files. You do not change a model setting or mode for this requirement.

A is incorrect: The model selector routes a request to Auto, Claude, or GPT. It does not choose between internal organizational content or external web content.

B is incorrect: Wording such as publicly available data directs Copilot to the web. Your requirement is to avoid web content and use internal insights.

C is incorrect: Chat only mode makes Copilot respond in the chat pane without making any changes to the document. It does not set the source.

C is correct.

Explanation: Copilot in Excel works only with files saved on OneDrive, and the data must be formatted as a table or a supported range. Your workbook fails both requirements, so you must fix both. After that, Copilot can use the context of your prompt to identify the right table or range without a manual selection.

A is incorrect: You do not need to select data before prompting, because Copilot uses the prompt context. Selection does not fix the file location.

B is incorrect: The model selector chooses which AI model answers. Changing it does not move the file to OneDrive or format the data.

D is incorrect: Chat only keeps Copilot responses in the chat pane without touching the workbook. It does not fix file location or data format.

B is correct.

Explanation: More capable models handle complex reasoning but use more per request, and a lighter model often reaches a good result for simpler tasks. Auto selects an appropriate model based on the nature of the task, which helps balance capability and usage automatically for most work. Simple tasks such as summarizing one document might not need that level of capability.

A is incorrect: A more capable model uses more credits per request. For simple tasks, a lighter model can reach a good result efficiently.

C is incorrect: Requesting credits happens after you reach the limit, and your administrator decides approval. It does not address the model choice driving usage.

D is incorrect: Generating images often requires more processing than text alone. Adding it would increase usage without benefiting a simple task.

A is correct.

Explanation: Cowork can't access files stored on your local device. It works with files in OneDrive, SharePoint, and other connected cloud services. Moving the spreadsheet to one of those locations lets you reference it in the request, and you can attach files from OneDrive, SharePoint, or Teams to give Cowork context. Review the results before you act on them.

B is incorrect: Cowork can't access files on a local device, so asking it to read the laptop file directly fails in this scenario.

C is incorrect: Cowork can't delete files or folders in OneDrive or SharePoint. Deleting copies also would not give access to the laptop file.

D is incorrect: Custom skills aren't validated by Microsoft, and quality depends on how they were written. Nothing documented adds local file access.

B is correct.

Explanation: The Researcher agent conducts multi-step research across internal work content and public web sources, then returns a structured report with citations and links. Including your role, the purpose, and the format you need makes the output more targeted. Follow-up prompts can then reformat the same findings, such as summarizing them into three bullet points for a leadership slide, without re-running the query.

A is incorrect: Chat only mode in Word keeps Copilot responses in the chat pane. It is not the agent built for multi-step research with citations across internal and external sources.

C is incorrect: Drafting an email with Copilot in Outlook produces a message. Researcher is optimized for discovery and synthesis, which this market analysis requires.

D is incorrect: Creating a deck from a prompt drafts slides. Researcher is the agent that conducts multi-step research and returns a cited report first.