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

Question 44 — AB-620 Case Study Set

Based on the scenario, Paideffort is preparing the planned release that combines the regulatory API, a SQL-backed policy service, and a Power Platform connector while applying regional business rules. Which integration design is most appropriate? (Select one answer)

  1. Use a Power Automate flow to aggregate, filter, rank, and return the formatted regulatory table to the topic.

  2. Add three independent HTTP request nodes and pass all returned records directly into one Custom data variable.

  3. Invoke the Foundry compliance agent separately for each source and concatenate the generated answers afterward.

  4. Create separate generative answers nodes for every source and allow each one to answer the same compliance question.

Correct answers

  1. Use a Power Automate flow to aggregate, filter, rank, and return the formatted regulatory table to the topic.

Explanations

Correct Answer:

Use a Power Automate flow to aggregate, filter, rank, and return the formatted regulatory table to the topic.

Explanation:

Paideffort's future design introduces multiple systems, a Power Platform connector, regional rules, removal of superseded records, and ranking logic. This exceeds the simple single-call transformation used in the current phase. A Power Automate flow provides a more maintainable place for the complex assembly logic and can return the final table to Copilot Studio.

Wrong Answers:

Add three independent HTTP request nodes and pass all returned records directly into one Custom data variable.

This would push complex integration and ranking logic into the topic and could produce an oversized, poorly ordered result set. Paideffort specifically wants centralized multi-source processing and compact relevant grounding.

Invoke the Foundry compliance agent separately for each source and concatenate the generated answers afterward.

Paideffort needs one grounded answer based on curated source records. Multiple model-generated intermediary answers would add latency and make source ranking and governance more difficult.

Create separate generative answers nodes for every source and allow each one to answer the same compliance question.

This duplicates generation logic and doesn't provide the centralized filtering and ranking required across the combined sources.