Skip to main content
  1. Public practice questions
  2. AB-620 Practice Exams | AI Agent Builder Associate
  3. AB-620 Practice Exam 4

Question 13 — AB-620 Practice Exam 4

A support solution must answer customer questions by retrieving and synthesizing information from an approved knowledge corpus. The same solution must also classify documents submitted by customers according to defined criteria. How should these two AI tasks be implemented in Microsoft Copilot Studio? (Select one answer)

  1. Use custom prompts for both knowledge retrieval and document classification

  2. Use generative answers for both knowledge retrieval and document classification

  3. Use a generative answers node for knowledge retrieval and a custom prompt for document classification

  4. Use a custom prompt for knowledge retrieval and a generative answers node for document classification

Correct answers

  1. Use a generative answers node for knowledge retrieval and a custom prompt for document classification

Explanations

Correct:

Use a generative answers node for knowledge retrieval and a custom prompt for document classification

A generative answers node retrieves information from configured knowledge sources and synthesizes a grounded response. A custom prompt directs the model to perform a defined reasoning task, such as classifying a submitted document.

Wrong:

Use custom prompts for both knowledge retrieval and document classification

Custom prompts can perform instructed reasoning tasks, but they are not the mechanism designed to retrieve and synthesize grounded answers from configured knowledge sources.

Use generative answers for both knowledge retrieval and document classification

Generative answers are designed for knowledge retrieval and synthesis. Document classification is a reasoning task that is more appropriately implemented with a custom prompt.

Use a custom prompt for knowledge retrieval and a generative answers node for document classification

This reverses the intended responsibilities. Generative answers handle source-grounded retrieval, while custom prompts perform instructed reasoning tasks such as classification.