Library › Data AI › Finance and accounting

Check an expense report line by line against policy limits

Practice: Keeping the AI to your sources. About 5 minutes.

About this exercise

You're reviewing a sales rep's conference expense report against travel policy. Write exactly 6 numbered lines, one per expense, each as Item: $amount vs $limit = OK or OVER by $X. Apply the approvals and policy details in the notes, then one unnumbered last line: Total over policy: $X. Plain text, under 110 words, nothing else.

Common mistakes when you ask AI for this

  • Ignoring the approval. The AI flags both $289 hotel nights, though one was pre-approved. Try: Add: "The Mar 4 hotel was pre-approved, so it's OK. Mar 3 was not."
  • Using the wrong limit. The AI checks the client dinner against $75, not $75 per person. Try: Add: "Meal limits are per person; the client dinner had 2 people."
  • Not showing the limit. Without the limit next to the amount, you can't check the verdict. Try: Add a line to copy: "1. Hotel Mar 3: $289 vs $250 = OVER by $39"
  • Skipping the total. The AI flags items but never adds up what to recover. Try: Add: "End with: Total over policy: $X."
  • Trusting the AI to count words. AI is bad at counting. Ask for a total word limit and it often lands over it. Try: Cap each line: "each line under 14 words."
  • Forgetting where it goes. AI loves markdown tables and bold, which show up as clutter in an expense system comment. Try: Add: "Plain text, no bold, headings or tables."

What a strong prompt includes

  • Applies the approval
  • Handles per-person limits
  • Asks for a verdict per line
  • Sets a word limit
  • Asks for plain text

Practice this now ↑

Related exercises