Microsoft GH-700: Enhancing Operational Efficiency with Agentic AI

Last checked October 22, 2026 against the official Microsoft exam page.

1. Question 1Business processes and efficiency

   A company wants to measure whether an agentic process has improved operational efficiency. Which metric is most useful?

   1. AThe number of agent messages sent per day to the operations team channel
   2. BThe colour scheme and font size of the monitoring dashboard layout
   3. CCycle time from request to completion, compared before and after the change
   4. DThe number of GitHub stars and forks gained by the repository
   Show answer and explanationHide answer

   Correct answer: C

   Cycle time before and after the change shows whether work is completed faster. Message counts, dashboard colours and repository stars do not measure operational efficiency.

2. Question 2Governance and responsible use

   A company's agent accesses customer records. The governance team requires that each agent has a clearly defined scope of what it may access and do. Which practice meets this?

   1. AFull administrator access for every agent, including tenant-wide rights
   2. BNo scope definition, so the agent can decide what to access
   3. CLeast-privilege permissions scoped to the agent's specific tasks
   4. DShared credentials reused across each agent in the pipeline
   Show answer and explanationHide answer

   Correct answer: C

   Least-privilege scoping limits each agent to its defined tasks, which reduces risk and supports accountability. Administrator access, shared credentials and undefined scope all expand what an agent can do.

3. Question 3Agentic AI fundamentals

   A business leader asks what distinguishes an AI agent from a basic chatbot. Which statement is most accurate?

   1. AAn agent can plan steps and use tools to act toward a goal, with oversight
   2. BAn agent only answers one question and cannot use any tools
   3. CA chatbot always takes actions in external systems, while an agent never does
   4. DThere is no difference between the two
   Show answer and explanationHide answer

   Correct answer: A

   An agent plans multi-step work and uses tools to act toward a goal, usually under human oversight. A single-question responder is closer to a basic chatbot, and the statements about taking actions reverse the distinction.

4. Question 4Agentic AI in software delivery

   A development team wants an agent to propose code changes for an issue, which a developer then reviews through a pull request. Why is the pull request step important?

   1. AIt removes the need for human review of the diff, since the agent's output is already trusted
   2. BIt lets the agent merge its changes directly into the main branch without a reviewer seeing them
   3. CIt keeps a human reviewer in control of what is merged, with tests and checks in place
   4. DIt exists mainly to document the change history for auditors and release notes
   Show answer and explanationHide answer

   Correct answer: C

   A pull request keeps human review, tests and checks in front of merges, which is essential for agent-proposed code. The other options remove review or misdescribe the step.

5. Question 5Business processes and efficiency

   A process owner notices that an agent completes tasks quickly but many outputs must be corrected by staff. What is the most useful measure to add?

   1. AThe number of tasks completed per minute by the agent pool
   2. BThe rate of human correction needed per completed task
   3. CThe number of agents assigned to the team this month
   4. DThe length of the agent's display name in bytes
   Show answer and explanationHide answer

   Correct answer: B

   The correction rate shows the quality of the outputs and the hidden effort they create. Speed alone and counts of agents or names do not reveal quality problems.