본문으로 건너뛰기
D.Hub
KOEN
← Workshop overview
Chapter 6 of 7
5 min

Check the refund-accountability boundary

Before the final explanation, review the input batch, risk threshold, evidence tools, and human-accountability boundary in four questions.

Question for this chapter

Can you separate the AI recommendation, human decision, and actual business record?

Why this matters now

You enriched order evidence, ordered reviews by risk, and ran policy retrieval and HITL confirmation. Before the final explanation, distinguish what automation assists, what a person owns, and what the current demo does not persist. This quiz has no passing score and does not lock the next chapter.

Core decision check

Select an answer to see the correct choice and its evidence immediately.

0/4 questions reviewed

  1. Question 1 Which statement correctly describes the observed scope of the current batch risk assessment?
  2. Question 2 How is an order with a risk score of exactly 0.600 classified?
  3. Question 3 Which statement correctly separates policy RAG from the order lookup tool?
  4. Question 4 Which statement best explains the accountability boundary in this Workshop?

Connect the evidence in one statement

Use risk and policy retrieval to support review, keep the final decision human, and do not overstate a downstream Actor result as a completed refund, email, or persistent audit record.

Next decision

You have separated recommendation, decision, and record. In the final chapter, distinguish the batch results, Agent demo, and existing dashboard when explaining the accountable refund-review flow.