AB-730 Exam Results

This scenario is an example of fabrication, which is commonly referred to in generative AI contexts as a hallucination. Fabrication occurs when an AI system generates information that appears credible but is factually incorrect, invented, or unsupported by verifiable sources.

According to Microsoft AI Business Professional guidance, large language models predict text based on patterns learned during training. They do not “know” facts in the human sense. As a result, when asked to generate reports using web-based information, the model may produce plausible-sounding but fictional details if sufficient grounding or reliable sources are not provided.

Deepfake refers specifically to synthetic media such as manipulated images, audio, or video. Overreliance describes a human behavior risk where users trust AI outputs without verification. Prompt injection is a malicious technique designed to manipulate model behavior. Bias refers to systematic unfairness in outputs.

In this case, the presence of fictional information in the generated report directly aligns with fabrication, making option B the correct answer.

In the Microsoft 365 Copilot app, the Conversations list can display status indicators next to individual conversation titles. In the exhibit, the Standings conversation has a distinct indicator (a green badge) shown to the right of the conversation name. This badge is used to signal that the conversation includes a verified response.

A verified response indicator is intended to help users quickly recognize outputs that were produced with stronger grounding and traceability—for example, responses that include supporting citations or references that can be reviewed for accuracy. This aligns with Microsoft’s guidance on improving trust in AI output by making source-backed or verification-supported responses easier to identify and validate.

The other options do not match the indicator shown. Conversations that are shared or added to notebooks typically use different share/notebook iconography, and scheduled prompts are represented with scheduling-related indicators rather than a verification badge. “Anonymous data” is not something Copilot marks at the conversation level in this manner.

Therefore, the best inference is that the Standings conversation contains a verified response.

In Microsoft 365 Copilot, a notebook is a workspace intended to organize and ground related Copilot work. Microsoft guidance describes notebooks as a way to bring together multiple conversations and keep them aligned to the same set of reference materials—such as documents, notes, links, or other resources—so you don’t have to repeatedly attach or restate the same context. This improves consistency and prompt-grounding across a set of related tasks (for example, managing a project, developing a proposal, or maintaining a recurring report).

Option B correctly captures this purpose: a notebook provides a private, organized location where reference materials can be curated and reused across related Copilot conversations.

Option A is incorrect because notebooks are not primarily designed to generate email-ready transcripts of conversations. Option C is incorrect because generating a summary of interactions is a conversation-level function; notebooks are broader organizational containers for materials and related workstreams, not a “summary generator” of a single chat thread.

Therefore, the purpose of a notebook is to organize reference materials across related conversations.

All three statements are false because generative AI responses are not guaranteed to be identical even when the prompt is the same. First, when grounded in the web , results can vary due to changing web content, different retrieved sources, or differences in how information is summarized at run time. Second, when grounded in your organization’s data , responses can change based on updates to files, emails, meetings, permissions, or which specific items Copilot retrieves as the most relevant context at that moment. Third, even when relying only on the model’s general knowledge, large language models are probabilistic: they may choose different wording, structure, examples, or emphasis across runs, especially when temperature/decoding settings and internal routing differ. In business scenarios, this means Copilot outputs should be treated as drafts that may require validation, and repeatability should be improved by adding precise constraints (cite specific sources, use fixed formats, specify exact sections, and request verbatim quotes where appropriate).