Library › Agents & Automation AI › Finance and accounting

Brief an assistant to enter receipts without fudging the numbers

  • For Finance and accounting
  • You practice Giving the right context
  • Time about 5 minutes

About this exercise

You want an AI assistant to enter this trip's receipts into an expense report. Write its brief: skip the hotel (it was already paid on the company card), use only the categories Meals, Ground Transport and Supplies, keep each amount exactly as printed, flag any meal over $75 per person, ask about missing details instead of guessing, and never submit. Output exactly 7 numbered plain-text lines: the add_line lines, one flag line, one ASK_HUMAN line, then a DONE line with the total.

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."
  • Leaving out what only you know. The AI enters every receipt, so the hotel gets paid twice, and it invents its own category names. Try: Add: "Skip the hotel, it's on the company card. Categories: Meals, Ground Transport, Supplies."
  • Not giving the policy. The AI can't know your meal limit or that it's per person. Try: Add: "Flag meals over $75 per person. Divide by party size."
  • Allowing guesses or submitting. The AI fills a missing date with a likely one and calls the job done by submitting. Try: Add: "Ask me about missing details. Never call submit_report."

What a strong prompt includes

  • Skips the company-card hotel
  • Names the categories
  • Gives the meal limit
  • Rules out made-up data
  • Forbids submitting

Practice this now ↑

Related exercises