Fix Your Output
Find the Problem, Fix Your Prompt
When the output isn’t quite right, identify what’s wrong and update your prompt to get AI back on track.
-
Generic output?
Add context. What information or references does the AI need?
Examples
- “This is for new hires on their first day.”
- “The goal is to reassure the client.”
- “Use the attached report as your source.”
-
Wrong length or shape?
Add a constraint. How many words? What format? How much detail?
Examples
- “Keep it under 150 words.”
- “Format it as a numbered list.”
- “Cover only the timeline and next steps.”
-
Wrong tone or voice?
Set the style. What tone fits? Is there an example to match?
Examples
- “Warm and direct, like a note to a colleague.”
- “Use plain language, no jargon.”
- “Match the tone of the attached email.”
-
Unnecessary content?
Say what to leave out. What should the AI skip or avoid?
Examples
- “Skip the intro. Start with the recommendation.”
- “Avoid phrases like ‘it’s important to note.’”
- “Don’t define terms the reader already knows.”
-
Missed the mark?
Remove ambiguity. Could a word mean two things? Do any instructions conflict?
Examples
- “By ‘soon,’ I mean by end of day Friday.”
- “‘Review’ means proofread, not rewrite.”
- “Focus on cost, not features.”
Remember that AI can make mistakes. Be sure to verify its output, especially when the stakes are high.