Free AB-730 practice test: 30 questions with answers

Microsoft 365 Copilot respects existing permissions, sensitivity labels, and tenant boundaries, so responses are grounded only in content the user can access. The other options describe access or data handling that Copilot does not perform.

Copilot grounds responses in the content it finds, so an older draft can be selected if it matches the request. Specifying the approved file or source corrects this, while the other explanations do not fit.

Copilot Chat handles quick questions and drafts grounded in web and work context. Cowork is for multi-step tasks, custom agents are for repeated workflows, and scheduled prompts run on a timetable.

Context such as the referenced files, the app in use, and Work IQ shapes which information Copilot uses. Screen brightness, name length, and unrelated mailboxes do not influence grounding.

Over-reliance means trusting AI output without verification, which lets errors reach decisions. The other options are not risks of accepting unverified content.

Checking citations and getting qualified human review matches the verification level that a client-facing legal document requires. Accepting self-assessment or removing citations removes the means of verification.

Entering sensitive data into unapproved tools can expose it outside governed controls. The other outcomes are not the central risk.

Prompt injection places instructions in data the assistant reads, such as a document or web page, to alter its behaviour. The other options describe speed, patching, or backups.

Copilot can use content from multiple Microsoft 365 apps to create related business content. The other options require manual steps or do not produce the presentation.

Connectors bring in content from external systems so Copilot can use it with permission controls. Screenshots, printed reports, and shared passwords are not secure or structured connections.

Copilot can gather context from related content so the attendee arrives prepared. Reading aloud, relying on memory, or printing mailboxes are inefficient and miss the point of preparation.

Comparing with the objective and checking numbers against source systems confirm that the content meets its purpose and is accurate. Accepting drafts unreviewed, publishing before review, and removing the objective each weaken validation.

Copilot Pages is a collaborative workspace where teams can build and refine content together with Copilot. A personal file, printed copies, and repeated attachments do not provide shared, live collaboration.

Intelligent recap summarizes meeting content, including decisions and action items, so people who missed it can catch up. Agendas, invitations, and voicemail greetings do not summarize the discussion.

Real-time catch-up provides a running summary during the meeting so a participant can rejoin the discussion quickly. Disabling chat, starting a new meeting, and reading agendas do not give that update.

Creating the chart from the source data keeps it accurate, and verification confirms the figures. Estimated values, unrelated charts, and a file name as the title do not meet the requirement.

Scheduling a prompt runs it automatically at the chosen time, which suits a recurring weekly report. Deleting chats, retyping from memory, and sharing chats do not run the prompt on schedule.

Iterative refinement with explicit length and focus instructions shapes the next response. Accepting the answer, restarting, or repeating the same prompt does not reliably change it.

Searching chat history by a distinctive keyword finds the previous conversation directly. Retyping content wastes effort, and printer queues and vendor emails do not locate chats.

Renaming a chat gives it a clear title that makes it easy to find later. Recreating chats daily, changing resolution, or disabling history do not organize the content.

Deleting the chat removes it from the chat list, which is the direct way to remove outdated or sensitive conversations. Sharing, renaming, or exporting without deletion keeps the content in place.

Notebooks in Microsoft 365 Copilot organize and share information from chats and sources in one workspace. Printed binders and screenshots are not structured for collaboration, and calendars do not hold project notes.

Instructions and memory help Copilot apply preferences consistently across interactions. A random seed, a printed guide that Copilot cannot read, and removing context all work against consistency.

Prebuilt agents are faster to adopt when they fit the process, and custom agents are reserved for needs that prebuilt ones do not meet. Building for every expense, manual review, and approval without rules are less effective.

Cowork plans and executes multi-step work by using skills and shows its plan, progress, and results for review. Single answers, replacing email, and storage alone do not describe it.

Reviewing the plan, progress, and results keeps the user accountable for the outcome. Accepting without review, ignoring the task, and granting unrestricted access are unsafe.

Scheduled tasks can be managed and their results reviewed, which keeps recurring work visible and controlled. Paper calendars, deleting schedules after each run, and assuming automatic success do not manage the task.

Prebuilt skills are faster for common tasks, while custom skills fit organization-specific processes. Choosing by cost alone or ignoring fit leads to poor results.

Least-privilege access and awareness of known limitations reduce the risk of exposing confidential content. Full rights, public sharing, and ignoring classification increase risk.

Prebuilt agents can be applied to processes from Copilot experiences, and chats can be filtered by agent. Agents still need human review, they do not require coding for prebuilt use, and they are designed for business tasks.