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

Question 43 — AB-620 Case Study Set

Given that the system uses only a limited number of records from Custom data and the regulatory API can rank results by the officer's question, which two changes best improve grounding quality? (Select two answers)

  1. Include Topic.ComplianceQuestion in the API's query parameter so the service ranks results for the current question.

  2. Ensure the three strongest results appear first before the formatted table reaches the generative answers node.

  3. Return all available regulations and rely on the generative answers node to search the complete table for the best records.

  4. Sort results only by lastUpdated because the newest regulations will always be the most relevant to the officer's question.

Correct answers

  1. Include Topic.ComplianceQuestion in the API's query parameter so the service ranks results for the current question.

  2. Ensure the three strongest results appear first before the formatted table reaches the generative answers node.

Explanations

Correct Answer:

Include Topic.ComplianceQuestion in the API's query parameter so the service ranks results for the current question.

Ensure the three strongest results appear first before the formatted table reaches the generative answers node.

Explanation:

Paideffort's pilot returned eight records, but the directly applicable prohibition appeared fourth. Because only the first three Custom data records contribute to answer generation, ranking must occur before the table reaches the node. Passing the officer's question to the API allows relevance-based retrieval, and placing the strongest three records first ensures those records are actually used.

Wrong Answers:

Return all available regulations and rely on the generative answers node to search the complete table for the best records.

The node doesn't perform relevance selection across an unlimited Custom data table. Paideffort's result ordering must already place the strongest records in the positions that will be consumed.

Sort results only by lastUpdated because the newest regulations will always be the most relevant to the officer's question.

The pilot demonstrates that recency and relevance can differ. The fourth record was more directly applicable to the projected-return question than the first three recently updated records.