6.4 Explain Human-In-The-Loop (HITL) Strategies AI-TECHNICAL-PRACTITIONER Practice Questions (Page 3)

Questions 11–15

1. 11expert · hard

   An e-commerce company deploys an agentic AI that manages inventory and automatically places orders with suppliers. The company wants to avoid stockouts but also wants to prevent over-ordering that ties up capital. They decide to implement HITL for order placement. Which design best balances efficiency and oversight?

   • AHave a human manually place all orders based on the AI's recommendations.
   • BAllow the AI to place orders automatically for low-value items, but require human approval for high-value or bulk orders.
   • CLet the AI place all orders automatically and review the orders monthly.
   • DRequire human approval for every order the AI suggests.
   Reveal Solution Discussion Select an answer first
2. 12foundation · easy

   In agentic AI systems, how is human-in-the-loop (HITL) typically applied to ensure responsible operation?

   • ABy using humans only to train the agent initially, with no ongoing oversight
   • BBy having the AI agent review human decisions to improve its own performance
   • CBy allowing the AI agent to operate autonomously without any human checks
   • DBy requiring human approval for high-impact actions taken by the agent
   Reveal Solution Discussion Select an answer first
4. 13application · medium

   A manufacturing company uses an agentic AI to control robotic arms on an assembly line. The AI occasionally makes errors that could damage products. The company wants to implement HITL to improve safety. Which benefit of HITL is most directly achieved by having a human supervisor monitor the AI's actions?

   • ALower costs, because the human replaces the need for expensive sensors.
   • BImproved accuracy, because the human can correct errors in real time.
   • CIncreased scalability, because the human allows the AI to handle more tasks.
   • DReduced latency, because the human speeds up the AI's decision-making.
   Reveal Solution Discussion Select an answer first
5. 14application · medium

   A marketing agency uses an agentic AI to generate social media posts. The agency wants to ensure the posts align with brand voice and comply with advertising regulations. Which HITL approach is most appropriate?

   • ALet the AI publish posts automatically, but have a human review the analytics later.
   • BUse the AI to generate drafts, and have a human select which drafts to publish.
   • CHave a human editor review and approve each AI-generated post before it is published.
   • DTrain the AI on past brand-approved posts and rely on its learning.
   Reveal Solution Discussion Select an answer first
6. 15expert · hard

   A startup is deploying an agentic AI to handle customer inquiries. The startup has a small team and limited budget. They want to incorporate HITL to ensure quality but are concerned about the cost of human review. Which approach is most cost-effective while still providing meaningful oversight?

   • ARely on customer feedback to identify issues and correct the AI.
   • BDeploy the AI without any human review to save costs.
   • CUse a sampling strategy where a percentage of AI responses are reviewed by humans.
   • DHire a large team of human reviewers to handle all AI outputs.
   Reveal Solution Discussion Select an answer first