Library › Finance AI › Accountants

Screen manual journal entries against red-flag criteria

  • For Accountants
  • You practice Setting limits and rules
  • Time about 5 minutes
  • Level Advanced

About this exercise

You're helping the controller screen October's manual journal entries. Write a prompt for plain text with one line per entry in the form "JE-####: flags" using the letters R, W, S, L, D (or "none"), then a last line starting "Follow-up:" listing entries that need the controller's attention. Flags are prompts for review, not conclusions about anyone's intent. Criteria: R = round amount (multiple of $1,000 and at least $10,000); W = posted on a Saturday or Sunday; S = preparer is also the approver; L = posted after the Nov 2 close date; D = description under 4 words. Under 80 words.

Common mistakes when you ask AI for this

  • Missing the date tests. The AI skips weekends and the close-date test. Try: Define each flag letter in the prompt and say "check all five tests".
  • Rounding rule drift. The AI flags $18,750 or $9,999 as round. Try: Say: "R = multiple of $1,000 and at least $10,000."
  • Accusing language. The AI calls entries suspicious or fraudulent. Try: Add: "Flags are prompts for review, not conclusions about intent."
  • Markdown extras. The AI adds bold labels, headings and a closing note. Try: Say: "Plain text, no bold. Output only the lines asked for."
  • Trusting the AI to count words. AI is bad at counting and runs past word limits. Try: Ask for fewer words than the real cap and keep each line short.

What a strong prompt includes

  • Defines the flags
  • Says to check all tests
  • Frames flags as prompts
  • Asks for a follow-up line
  • Caps length

Practice this now ↑

Related exercises