Library › Agents & Automation AI › Office and admin

Brief an assistant to triage your inbox without sending anything

  • For Office and admin
  • You practice Setting limits and rules
  • Time about 5 minutes

About this exercise

You're boarding a flight and handing your Monday inbox to an AI assistant. Write its brief: give every email exactly one label (Urgent, Reply or FYI), draft replies but never send them, never pay anything, and send any money or security question to you with ASK_HUMAN. It must output a plain-text numbered action log of at most 12 lines, ending with a DONE line that counts how many emails got each label.

Common mistakes when you ask AI for this

  • Not showing the log format. Left alone, the AI writes a friendly report with headings instead of a checkable action log. Try: Show it literally: "Format every line like 1. ACTION: tool(args), then ASK_HUMAN: and DONE: lines. Plain text."
  • Trusting the AI to count words. AI is bad at counting and runs long, especially when it explains each step. Try: Cap each line instead: "each line under 20 words, no explanations."
  • Letting it send or pay. Told to "handle" the inbox, the AI may reply or pay on your behalf, including a likely scam. Try: Add: "Draft replies only. Never use send_reply or pay_invoice."
  • No rule for money and security. A changed bank account looks routine to the AI unless you tell it these always come to you. Try: Add: "Any payment, bank change or password email goes to ASK_HUMAN."
  • No finish line. Without a done rule the AI just stops, and you can't see what it did at a glance. Try: Say: "End with DONE: counts of Urgent, Reply and FYI."

What a strong prompt includes

  • Names the labels
  • Says draft, never send
  • Makes it ask before anything risky
  • Shows the action-log format
  • Says how to finish

Practice this now ↑

Related exercises