ExamSprintLab

Free AB-730 practice test: 20 questions

Updated October 9, 2026

Twenty questions across the four AB-730 skill areas, weighted roughly as the real exam is: six on generative AI fundamentals, four on prompts and chats, five on business content and collaboration, and five on agents. These are original scenario questions, not questions from the real exam. Read the constraint in each question, such as how often, who, or which risk, before choosing an answer.

Generative AI fundamentals

Question 1. A user asks Copilot for the sales figures in a spreadsheet on a SharePoint site she cannot open. What happens?

  • A. Copilot returns the figures because it can read every file
  • B. Copilot grants her access to the site
  • C. Copilot does not use that file, because she lacks access
  • D. Copilot invents figures to fill the gap
Show answer

Answer: C

Copilot works within the user’s existing permissions, so it cannot use content she cannot access. It does not read everything by default, it does not grant access, and it does not return fabricated figures on purpose.

Want more questions like this? Full AB-730 practice tests →

Question 2. An answer from Copilot looks professional but includes a legal citation that the user cannot verify. What is the best action?

  • A. Use it, because it reads confidently
  • B. Verify the citation in a trusted source and have a person review it
  • C. Ask Copilot whether it is sure and accept its reply
  • D. Delete the chat and move on
Show answer

Answer: B

Fluent text can still be wrong, so check the citation against a trusted source and involve human review for anything legal. Trusting the confident tone is over-reliance. Asking Copilot if it is sure is not independent verification. Deleting the chat does not fix the question.

Want more questions like this? Full AB-730 practice tests →

Question 3. A user asks the same question inside a Word document and in a general chat and gets different answers. What most likely explains this?

  • A. The context available in each place is different
  • B. Copilot is broken in one of them
  • C. The user’s permissions changed between asking
  • D. Word always returns shorter answers
Show answer

Answer: A

The context available to Copilot differs: inside a document it is grounded in that document, while a general chat can draw on different sources. Different answers do not mean either is broken or that the prompts are secretly stored differently.

Want more questions like this? Full AB-730 practice tests →

Question 4. An employee wants to delegate a long, multi-step task and review the plan and results. Which option is designed for this?

  • A. A single chat prompt
  • B. A saved prompt
  • C. Intelligent recap
  • D. Microsoft Copilot Cowork
Show answer

Answer: D

Copilot Cowork is the capability for delegating multi-step tasks and reviewing its plan, progress and results. A single chat answers one question at a time. A saved prompt stores wording. A meeting recap summarises a meeting.

Want more questions like this? Full AB-730 practice tests →

Question 5. Which statement about verification best reflects good practice?

  • A. AI output never needs checking
  • B. Choose a check that fits the task and its risk
  • C. Never use AI for any business task
  • D. Let the AI grade its own answers
Show answer

Answer: B

The check should match the task: a casual idea needs little, while a figure for a board paper needs a source check and a customer message needs a human read. Always trusting AI or always refusing it are both extremes. Letting the AI mark its own work is not independent.

Want more questions like this? Full AB-730 practice tests →

Question 6. A user pastes customer records marked confidential into a tool not approved by her company. Which risk does this create?

  • A. Inaccuracy
  • B. Prompt injection
  • C. Exposure of sensitive data
  • D. A weak summary
Show answer

Answer: C

Putting confidential content where it should not go creates a sensitive data exposure risk. Inaccuracy concerns wrong answers. Prompt injection concerns hidden instructions. A weak summary is a quality issue, not a data risk.

Want more questions like this? Full AB-730 practice tests →

Prompts and chats

Question 7. A manager wants Copilot to summarise open risks every Friday without asking each time. What should she use?

  • A. Schedule the prompt
  • B. Save the prompt only
  • C. Share the prompt
  • D. Rename the chat
Show answer

Answer: A

Scheduling a prompt makes it run at a regular time. Saving keeps wording but does not run it. Sharing gives others the prompt. Renaming a chat only changes its label.

Want more questions like this? Full AB-730 practice tests →

Question 8. An answer is too long for a busy executive. What is the best refinement?

  • A. Submit the prompt again unchanged
  • B. Delete the chat
  • C. Add several more sources
  • D. Ask for a five-bullet version for an executive audience
Show answer

Answer: D

A specific follow-up that states the audience and the length improves the answer. Repeating the same prompt gives the same kind of result. Deleting the chat discards the work. Adding more sources would likely lengthen it.

Want more questions like this? Full AB-730 practice tests →

Question 9. A user keeps retyping the same preference about tone and formatting at the start of each chat. What feature reduces this effort?

  • A. A notebook
  • B. Memory and instructions
  • C. Scheduling the prompt
  • D. Deleting chats
Show answer

Answer: B

Memory and instructions let Copilot apply persistent preferences so you do not repeat them. Notebooks collect material, scheduling repeats a prompt, and deleting chats does not store preferences.

Want more questions like this? Full AB-730 practice tests →

Question 10. A team wants to gather chats, files and notes about one project in a single place and share it. What should they use?

  • A. A saved prompt
  • B. Real-time catchup
  • C. A notebook
  • D. A renamed chat
Show answer

Answer: C

A notebook organises related information and can be shared. A saved prompt holds one instruction. Real-time catchup serves a meeting attendee. Renaming a chat helps one chat be found.

Want more questions like this? Full AB-730 practice tests →

Business content and collaboration

Question 11. A user needs a chart and insights from a table of monthly sales, then a short narrative. Which sequence is most sensible?

  • A. Analyse and chart in Excel, then draft the narrative in Word and review it
  • B. Write the narrative first, then invent the numbers
  • C. Use Outlook for the analysis
  • D. Skip the review since Copilot did the work
Show answer

Answer: A

Analysing numbers is a strength of Excel, and the findings can then be carried into a Word draft, which uses the right app for each step and a check at the end. Starting in Outlook, or skipping analysis, does not use the right tools.

Want more questions like this? Full AB-730 practice tests →

Question 12. A report must include data from a system outside Microsoft 365. What helps Copilot use that data?

  • A. Saving the prompt
  • B. Sharing the prompt
  • C. Intelligent recap
  • D. A connector
Show answer

Answer: D

Connectors bring in sources beyond Microsoft 365. Saving or sharing a prompt does not add data, and a recap summarises a meeting.

Want more questions like this? Full AB-730 practice tests →

Question 13. An executive joins a meeting late and wants to know what was covered so far. Which feature fits?

  • A. Intelligent recap
  • B. Real-time catchup
  • C. Meeting preparation
  • D. Copilot Pages
Show answer

Answer: B

Real-time catchup is meant for someone who joins late. Intelligent recap is a summary after the meeting. Meeting preparation happens before. Pages is for co-creating content.

Want more questions like this? Full AB-730 practice tests →

Question 14. A generated slide deck looks polished, but one slide contradicts the project brief. What should the user do?

  • A. Present it, because it looks polished
  • B. Add a better chart to the slide
  • C. Fix the slide to match the brief and recheck the rest
  • D. Delete the deck and abandon the task
Show answer

Answer: C

Content must be validated against business goals, so correct the slide against the brief. Presenting it ignores the error, and a better visual does not fix a wrong message. Deleting the whole deck is unnecessary.

Want more questions like this? Full AB-730 practice tests →

Question 15. A group wants to build a project brief together in one shared space. Which feature fits best?

  • A. Copilot Pages
  • B. A saved prompt
  • C. Intelligent recap
  • D. A renamed chat
Show answer

Answer: A

Copilot Pages is a shared space for collaborating on content. A saved prompt stores wording. Intelligent recap summarises a meeting. A renamed chat does not enable co-creation.

Want more questions like this? Full AB-730 practice tests →

Agents

Question 16. A task is common and repeatable, and a prebuilt agent covers it. What is the best choice?

  • A. Build a custom agent
  • B. Do the work manually
  • C. Wait for a new feature
  • D. Use the prebuilt agent
Show answer

Answer: D

Use the prebuilt agent when it fits the need. Custom building adds effort without a reason, manual work forgoes the benefit, and waiting does not solve the task.

Want more questions like this? Full AB-730 practice tests →

Question 17. A process depends on a company-specific rulebook that no prebuilt agent covers. What is the best approach to consider?

  • A. Force a prebuilt agent that does not fit
  • B. Consider a custom agent
  • C. Avoid agents completely
  • D. Ignore the rulebook
Show answer

Answer: B

When the need is specific to the organisation and no prebuilt agent covers it, a custom agent is worth considering. A prebuilt agent that does not fit will give weak results, and avoiding agents or ignoring the rulebook does not meet the need.

Want more questions like this? Full AB-730 practice tests →

Question 18. After Cowork finishes a delegated task, what should the user do before relying on the result?

  • A. Review the plan, progress and results
  • B. Forward it unchecked
  • C. Delete it
  • D. Ask Cowork to certify its own accuracy
Show answer

Answer: A

Reviewing the plan, progress and results is part of delegating, and accountability remains with the user. Forwarding unchecked, deleting the work, or asking Cowork to certify itself are not appropriate.

Want more questions like this? Full AB-730 practice tests →

Question 19. A user wants to find only the chats she had with one particular agent. What should she do?

  • A. Rename all chats
  • B. Schedule a prompt
  • C. Filter chats by agent
  • D. Delete all other chats
Show answer

Answer: C

Chats can be filtered by agent. Renaming all chats is slow, scheduling creates future runs, and deleting chats removes what she wants.

Want more questions like this? Full AB-730 practice tests →

Question 20. A user must choose between a prebuilt skill and a custom skill in Cowork for a standard task. The prebuilt skill fits. Which is the better choice?

  • A. Always build a custom skill
  • B. Use the prebuilt skill
  • C. Skip skills altogether
  • D. Do the task by hand
Show answer

Answer: B

Use the prebuilt skill when it fits the standard task. A custom skill is better when the work is specific to the user or organisation. Skipping skills or doing it by hand loses the benefit.

Want more questions like this? Full AB-730 practice tests →

Scoring

Count your correct answers by area. Below 80% overall, return to the domain guides: fundamentals, prompts and chats, content and collaboration and agents and Cowork. See the exam tips for pacing.