Most disappointing AI output comes from a small set of avoidable mistakes.
Being Vague
"Write something about our product" invites generic filler. Fix: specify audience, purpose, length and key points.
Missing Context
The model doesn't know your company, customers or constraints. Fix: include the relevant background.
Asking for Everything at Once
A single prompt to research, analyse, write and format a report produces mediocre results on each part. Fix: break it into steps.
No Output Format
Unstructured requests yield unstructured answers. Fix: describe or show the format.
Not Providing Examples
Describing a style is hard; showing it is easy. Fix: include one or two examples.
Trusting Specific Facts
Numbers, citations and dates from memory may be invented. Fix: provide sources or verify.
Contradictory Instructions
"Be comprehensive but under 100 words." Fix: prioritise.
Hiding Instructions in Data
Instructions mixed into pasted documents get confused with content. Fix: separate them with clear delimiters.
Giving Up After One Try
The first answer is a draft. Fix: give specific feedback and iterate.
Never Testing
Prompts that work on one example may fail on others. Fix: test on varied inputs before relying on them.