Appendix C — Instructor debrief questions

• Which app or agent produced the most immediately useful output for your department, and why?
• Where did the Analyst or Researcher agent need tighter scoping or better sources?
• Which validation step caught the most errors across the labs?
• How did specifying an output contract change reliability and reuse?
• When should you chain Analyst then Researcher, versus using one agent?
• When is a simple Agent Builder agent enough, and when do you need Copilot Studio?
• How did industry sensitivity change the way you wrote a prompt or handled an output?
• Which red-team test was hardest for your agent to pass, and how did you fix it?