Understand generative AI fundamentals — AB-730 Practice Quiz

Question 1

A project manager repeatedly performs the same multi-step routine each week: pulling status updates from a specific SharePoint site, checking a project tracker list, and drafting a formatted summary email for stakeholders using a fixed template. They want a reusable solution their whole team can trigger on demand without re-explaining the steps and data sources every time. Which approach best fits this need?

Accepted Answer

Create a custom Copilot agent scoped to the SharePoint site and tracker list with the reporting instructions and template built in

Answer

Use a standard Copilot chat session and re-enter the full prompt with all instructions each week

Answer

Ask Copilot in Word to summarize whatever document happens to be open at the time

Answer

Export the SharePoint data to a personal spreadsheet and manually write the email without Copilot

Question 2

A marketing manager describes two different Copilot interactions to a colleague. In the first, she types questions into Copilot Chat and receives immediate answers she reads and acts on herself. In the second, her team uses a solution that continuously monitors a shared mailbox, categorizes incoming partnership requests against company criteria, and drafts routing responses without someone prompting it each time. Which statement best captures the defining difference between these two experiences?

Accepted Answer

The second is an agent experience because it operates on a task autonomously against defined instructions, while the first is a chat experience driven by the user's individual prompts.

Answer

The first is an agent experience because it responds instantly, while the second is a chat experience because it involves email.

Answer

Both are chat experiences; the only difference is that the second one has access to more organizational data.

Answer

The first is an agent experience because it uses natural language, while the second is simply a mailbox rule with no AI involved.

Question 3

A project manager frequently asks Microsoft 365 Copilot Chat the same set of questions about the company's onboarding process, then formats the answers using a fixed template and internal policy documents. She wants a reusable solution that automatically applies this specific knowledge and behavior every time, without re-typing detailed instructions. Which approach best fits her need?

Accepted Answer

Create a custom agent (for example, using Copilot Studio) grounded in the onboarding documents and configured with the required instructions and output format

Answer

Continue using Copilot Chat and paste the full instructions and policy text into each prompt

Answer

Turn off web grounding in Copilot Chat so responses only use the onboarding documents

Answer

Export the onboarding documents to a shared folder and rely on Copilot to detect them automatically in every chat

Question 4

A marketing coordinator occasionally needs to ask ad-hoc questions like 'Summarize the key points from last week's campaign emails' or 'Draft a short update based on my recent documents.' These requests are one-off, vary each time, and require no predefined logic or specialized knowledge base. Which Microsoft 365 Copilot experience best fits this need?

Accepted Answer

The Copilot chat experience, where the user types natural-language prompts for on-demand help

Answer

A custom-built agent configured with specific instructions and a dedicated knowledge source

Answer

An autonomous agent that runs on a schedule to complete the tasks without user input

Answer

A custom agent published to the whole organization so all coordinators share the same workflow

Question 5

A marketing analyst asks Microsoft 365 Copilot Chat to summarize the latest industry adoption statistics for a product category. The analyst notices that when web content is enabled, Copilot returns very recent figures, but when it is disabled, Copilot only references older data found in internal SharePoint reports. Which statement best explains what is happening?

Accepted Answer

Enabling web content lets Copilot ground its response in current public information from the internet, while disabling it restricts grounding to organizational data such as SharePoint files.

Answer

Disabling web content causes Copilot to fabricate older statistics because it can no longer access any grounding source.

Answer

Web content changes the underlying language model to a newer version that contains more recent training data.

Answer

With web content disabled, Copilot uploads the internal SharePoint reports to the public web to retrieve comparable figures.

Question 6

A marketing analyst uses Microsoft 365 Copilot to draft a competitive summary based on several internal SharePoint documents. Copilot's response includes specific market-share percentages and inline citations pointing to source files. Before sharing the summary with executives, what is the most appropriate responsible AI verification step?

Accepted Answer

Open each cited source file to confirm the percentages actually appear there and match the summary

Answer

Re-run the same prompt several times and accept the figures if Copilot returns them consistently

Answer

Trust the figures because Copilot only draws from the organization's secured internal files

Answer

Ask Copilot whether the numbers it generated are accurate and rely on its confirmation

Question 7

A sales manager opens Copilot in Microsoft Word to draft a client proposal. She notices that when she uses Copilot in a blank document, the responses are generic, but when she works inside a document already containing the client's meeting notes and pricing details, the responses are far more tailored. A colleague asks her to explain why the quality differs. What is the most accurate explanation of how context affects Copilot responses?

Accepted Answer

Copilot uses the content of the open document, referenced work files, and the app in use as context, so richer relevant content produces more tailored responses.

Answer

Copilot only produces better responses when connected to the public web, so the blank document lacked an internet connection.

Answer

Copilot randomly varies response quality and the difference is unrelated to the document content.

Answer

Copilot stores previous prompts from all users in the tenant and reuses the most popular answers, which improves quality over time.

Question 8

A project manager asks Microsoft 365 Copilot in Business Chat to 'summarize the latest budget decisions across our department.' A colleague in the same department has a budget spreadsheet stored in their private OneDrive that has never been shared with the project manager. When Copilot generates its response, what should the project manager expect regarding that spreadsheet?

Accepted Answer

Copilot will not use the spreadsheet because the project manager does not have permission to access it.

Answer

Copilot will include the spreadsheet because both users belong to the same department.

Answer

Copilot will include the spreadsheet but redact the numbers to protect sensitive data.

Answer

Copilot will request temporary access to the spreadsheet on the project manager's behalf.

Question 9

A marketing coordinator at Contoso uses Microsoft 365 Copilot to ask, 'Summarize the latest details from the confidential 2025 acquisition plan.' The coordinator has no permission to open that SharePoint document. What will Copilot most likely do, and why?

Accepted Answer

Copilot will exclude that document from its response because it only returns content the user is already permitted to access.

Answer

Copilot will summarize the document because Copilot bypasses individual file permissions to give complete answers.

Answer

Copilot will return the summary but automatically notify the document owner that the coordinator viewed it.

Answer

Copilot will summarize a cached public version of the file stored in its training data.

Question 10

A compliance analyst at a healthcare company asks Microsoft 365 Copilot to summarize all patient case documents stored in a SharePoint site. The analyst notices that several documents labeled with a strict sensitivity label are excluded from the summary, even though they exist in the site. What is the most likely reason Copilot omitted those documents?

Accepted Answer

The sensitivity label's encryption and usage rights prevent Copilot from accessing content the user isn't authorized to view or extract.

Answer

Copilot randomly samples documents to reduce processing time, so some files are always skipped.

Answer

Copilot only reads files that were created within the last 30 days and ignores older documents.

Answer

The documents were excluded because they contained images, which Copilot cannot process.

Question 11

A financial analyst is working in Microsoft Excel on a workbook that contains quarterly sales figures. She opens Copilot in Excel and asks it to identify trends and suggest a formula to forecast next quarter's revenue. Which factor most directly determines the relevance of Copilot's response in this situation?

Accepted Answer

The data and structure within the open workbook that Copilot uses as context for the request

Answer

The number of other users in the organization who have recently used Copilot in Excel

Answer

The total amount of public web data Microsoft has indexed about revenue forecasting

Answer

The subscription tier of the analyst's personal Microsoft account

Question 12

A marketing analyst asks Microsoft 365 Copilot to summarize a competitor's pricing trends and requests supporting figures. Copilot returns a confident summary with specific percentage growth numbers, but when the analyst clicks the provided citations, the linked documents do not actually contain those percentages. What is the MOST accurate way to characterize what has occurred, and what should the analyst do?

Accepted Answer

This is a fabrication (the model generated plausible-sounding but unsupported figures); the analyst should not use the numbers until they are verified against the actual source data.

Answer

This is a prompt injection attack embedded in the competitor documents; the analyst should report the documents to security and delete them.

Answer

This is a data protection restriction that hid the real numbers; the analyst should request elevated permissions to reveal the correct figures.

Answer

This is expected behavior because Copilot cannot cite sources; the analyst should accept the figures as reliable since Copilot is grounded in organizational data.

Question 13

A marketing manager asks Microsoft 365 Copilot to draft a blog post about industry trends. Copilot produces polished text that includes a specific statistic: '78% of enterprises adopted this technology in 2023.' The manager cannot find this figure in any of the source files Copilot referenced, and no citation is attached to the claim. What is the most appropriate action before publishing?

Accepted Answer

Treat the statistic as a possible fabrication and verify it against an authoritative source before including it

Answer

Publish the post as written, since Copilot draws on trusted organizational data and rarely produces incorrect numbers

Answer

Ask Copilot to confirm whether the statistic is correct and accept its answer as validation

Answer

Remove all statistics from the post because Copilot cannot generate any accurate numeric data

Question 14

A financial analyst wants to use Microsoft 365 Copilot to turn a set of raw quarterly sales figures into a formatted chart and a summary table with calculated totals. Which Microsoft 365 app's Copilot experience is best suited to this specific task?

Accepted Answer

Copilot in Excel, which can analyze data, generate formulas, and create charts from the values in a worksheet

Answer

Copilot in Outlook, which can draft and summarize email messages

Answer

Copilot in PowerPoint, which can generate presentation slides from a document

Answer

Copilot in Word, which can draft and rewrite long-form narrative documents

Question 15

A project manager asks Microsoft 365 Copilot in Teams, "What did our client say about the delivery deadline?" The relevant details were only discussed verbally in a meeting that was never recorded and were never written into any file, email, or chat. What is the most likely outcome, and why?

Accepted Answer

Copilot may be unable to answer or could fabricate a response, because it can only ground on organizational content it can actually access, such as files, emails, chats, and recorded meetings.

Answer

Copilot returns an accurate answer because it can access all past conversations within the organization automatically.

Answer

Copilot searches the public web to reconstruct what the client likely said about the deadline.

Answer

Copilot listens to the ambient audio history on the user's device to recover the verbal discussion.

Question 16

A marketing coordinator uses Microsoft 365 Copilot in Teams chat to ask, "Summarize the key points from the executive budget planning documents for next quarter." The coordinator has never been granted access to the finance team's SharePoint folder where those documents are stored. What will Copilot most likely do?

Accepted Answer

Only reference content the coordinator already has permission to access, so the restricted budget documents are excluded from the response

Answer

Return a summary of the budget documents because Copilot has elevated access to all organizational content

Answer

Grant the coordinator temporary access to the finance folder so the summary can be generated

Answer

Send the request to the finance team's manager for approval before responding

Question 17

A paralegal uses Microsoft 365 Copilot in Word to draft standard contract clauses for a client agreement. Copilot produces well-formatted language that references a specific state statute. Before the paralegal submits the draft to the supervising attorney, what is the most important responsible AI practice to apply?

Accepted Answer

Have a qualified human review and validate the generated clauses and any legal citations before the draft is relied upon

Answer

Regenerate the response several times and submit whichever version is longest and most detailed

Answer

Assume the clauses are accurate because Copilot only draws from the organization's approved templates

Answer

Disable web grounding so that Copilot cannot introduce any incorrect legal references

Question 18

A product team uses Microsoft 365 Copilot across several apps. A program manager is drafting a shared status update in a Microsoft Loop workspace where teammates are co-editing components in real time. She wants to understand what makes the Copilot experience in Loop distinct from asking the same question in the standalone Copilot chat app. Which statement best describes how Copilot behaves differently in this Loop context?

Accepted Answer

Copilot in Loop is grounded in the shared, collaboratively edited content of the workspace, so it can summarize and build on what the team is actively co-authoring, whereas standalone chat has no awareness of that live page.

Answer

Copilot in Loop can access files that individual teammates do not have permission to view, because collaborative workspaces override standard file permissions.

Answer

Copilot in Loop automatically sends the workspace content to the public web to improve its answers, while standalone chat keeps everything internal.

Answer

Copilot in Loop and standalone chat are functionally identical; the only difference is the color of the interface.

Question 19

A marketing manager uses Microsoft 365 Copilot to brainstorm campaign taglines for a new product launch. Copilot returns ten creative options within seconds. Before presenting the top three to the leadership team, what is the most appropriate responsible AI practice for the manager to follow?

Accepted Answer

Apply human judgment to review the taglines for brand fit, accuracy, and any unintended trademark or messaging issues before sharing

Answer

Present the taglines exactly as generated, since Copilot draws only from trusted organizational data and cannot produce inappropriate content

Answer

Disable Copilot's web grounding so the taglines are guaranteed to be original and free from external influence

Answer

Ask Copilot to confirm that the taglines are legally cleared and free of trademark conflicts, then rely on that confirmation

Question 20

A marketing analyst uses Microsoft 365 Copilot to draft a competitive analysis that will be presented to the executive leadership team next week. Copilot generates a polished summary that includes specific market share percentages, competitor revenue figures, and a claim about an upcoming competitor product launch. The analyst is impressed by how confident and detailed the output looks. Which action best reflects a responsible AI practice for reducing over-reliance before the report is finalized?

Accepted Answer

Independently verify the specific figures and the product-launch claim against authoritative sources and apply human review before presenting

Answer

Submit the report as-is, since Copilot only uses trusted organizational data and cannot produce inaccurate figures

Answer

Ask Copilot to rate its own confidence in the numbers and trust the report if the confidence score is high

Answer

Re-run the same prompt several times and accept the figures if Copilot returns consistent numbers each time

More AB-730 practice