Library › Customer Support & Success AI › Customer support teams

Write an engineering escalation that sticks to what you reproduced

  • For Customer support teams
  • You practice Keeping the AI to your sources
  • Time about 5 minutes
  • Level Advanced

About this exercise

You're escalating a bug to engineering. Fill exactly these six labels in this order: Summary:, Steps:, Expected:, Actual:, Impact:, Tried:. Engineers only trust what support reproduced, so stick to your own test results, not the customers' guesses. Impact must give the number of tickets and accounts and the platforms affected. Plain text, under 190 words.

Common mistakes when you ask AI for this

  • Using its own sections. The AI writes a bug report in its own layout, so engineers have to hunt for steps and impact. Try: List the labels: "Exactly: Summary:, Steps:, Expected:, Actual:, Impact:, Tried:"
  • Counting loosely. The AI says "multiple customers" or counts the two Brightwater tickets as two accounts. Try: Ask for it: "Impact: number of tickets, number of accounts, platforms."
  • Repeating customer guesses. The AI passes on "all users" and "server is down" as if they were findings. Try: Add: "Only what I reproduced. Ignore customer guesses."
  • Formatting for a doc. AI bolds every label, which turns into asterisks in the bug tracker. Try: Say "plain text, labels followed by a colon, no bold".
  • Trusting the AI to count words. AI is bad at counting and usually runs long, so a vague "keep it short" lands well over 190 words. Try: Cap each part instead of the total: "each line under 15 words".

What a strong prompt includes

  • Lists the labels
  • Asks for the impact count
  • Sticks to reproduced facts
  • Asks for plain text
  • Asks for plain text

Practice this now ↑

Related exercises