AI Business Professional

Sample AB-730 practice questions

5 of the 61 original questions in our AB-730 bank, written to match the real exam's style and difficulty. Every question includes a full explanation — expand each answer to check your reasoning.

1. Question 1easy

   A user asks Microsoft 365 Copilot to summarize a confidential document stored in their OneDrive. Which statement about how Copilot accesses that content is correct?

   • ACopilot can only access public web data, not the user’s files.
   • BCopilot honors the existing Microsoft 365 permissions of the user — it can only use content the user already has access to.
   • CCopilot shares the document with everyone in the organization to generate a summary.
   • DCopilot stores the document permanently to train the underlying model.
   Show answer & explanation

   Correct: B

   Microsoft 365 Copilot operates within the user’s existing permission boundary — it surfaces and reasons over only the content that user is already entitled to see, enforced by the Microsoft Graph and existing access controls. It does not broaden sharing, and customer content is not used to train the foundation models. Understanding this permissions-respecting, tenant-isolated model is a core responsible-AI fundamental on AB-730.

2. Question 2medium

   Which prompt is most likely to produce a useful first draft from Copilot in Word?

   • A"Write something about our product."
   • B"Write a 200-word product announcement email for our new analytics dashboard, aimed at existing customers, in a friendly but professional tone."
   • C"Product."
   • D"Make it good."
   Show answer & explanation

   Correct: B

   Effective prompts specify the goal, format, length, audience, and tone. Option B gives Copilot all four — deliverable (announcement email), length (200 words), audience (existing customers), and tone (friendly but professional) — so the output needs far less rework. The other options are too vague for the model to infer intent. A common AB-730 theme is that prompt quality, not Copilot itself, is usually the limiting factor in output quality.

3. Question 3easy

   What is a large language model (LLM) best described as?

   • AA spreadsheet formula engine
   • BA model trained on large text datasets that predicts and generates human-like language
   • CA database index
   • DA network firewall
   Show answer & explanation

   Correct: B

   An LLM is trained on large text datasets and predicts/generates human-like language.

4. Question 4easy

   In generative AI, what is a prompt?

   • AThe model's training dataset
   • BThe instruction or input you give the model to produce a response
   • CA security role
   • DA type of license
   Show answer & explanation

   Correct: B

   A prompt is the instruction/input you give the model to produce output.

5. Question 5easy

   A colleague says Copilot "made up" a citation that doesn't exist. This is an example of:

   • AA hallucination / fabrication
   • BA connector error
   • CA license issue
   • DA network timeout
   Show answer & explanation

   Correct: A

   Inventing non-existent facts/citations is a hallucination (fabrication) — a known limitation requiring verification.