AB-730 Course Notes

Microsoft 365 Copilot answers are grounded in the user's work context instead of only in the language model. For the exam, grounding means Copilot can use Microsoft Graph content such as documents, email, chats, meetings, calendar items, and contacts that the signed-in user is allowed to access. The practical point is permission trimming: Copilot should not expose a file, message, or site that the user could not already view through Microsoft 365.

Effective Copilot prompts usually have four working parts: a goal, context, referenced content, and expectations. The goal says what the user wants done, context explains the business situation or audience, referenced content points Copilot to files or information to use, and expectations define the format, tone, length, or decision criteria. Exam questions often hide the right answer in the missing part of that prompt structure.

Advanced prompting for AB-730 is mostly about turning repeatable prompt behavior into an agent. Agent Builder templates give users a starting point for a declarative agent, but the agent still needs a clear name, description, and purpose so users can recognize when to use it. Identity details are not cosmetic on the exam; a weak description makes the agent hard to find and easy to misuse.

Copilot content creation starts with a prompt and the context Copilot is allowed to use. In Word, users can draft or rewrite content and ask questions about a document; in PowerPoint, they can create a presentation from a prompt or a Word file; in Excel, they can ask for formulas, charts, and insights; in Outlook, they can summarize threads or draft replies; and in Teams, they can summarize chat or meeting context when the required source content is available.