Two people using the exact same AI tool can get completely different results, and the reason is almost never the tool: it's how they phrased the instruction (the "prompt"). Writing better prompts is a specific skill, and it's learned through practice.
Give context, not just the question
Asking "write a follow-up email" gives a generic result. Adding context (who it's for, what happened before, what tone it should have, what you want to achieve with that email) gives a far more useful result, much closer to what's actually needed.
Be specific about the output format
Explicitly stating the desired length, whether it should be a list or paragraphs, or whether it needs a formal or casual tone, avoids having to reformulate the answer several times afterward. AI doesn't guess these preferences unless told.
Ask for examples or provide your own
When the task is repetitive (writing several product descriptions in the same style, for example) giving one or two already-written reference examples noticeably improves the result's consistency compared to describing the desired style only in words.
Iterate instead of expecting the perfect answer first try
The first response is rarely exactly what's needed. Asking for specific adjustments to that first response ("shorter", "less formal", "focus more on X") usually reaches the desired result much faster than rewriting the whole prompt from scratch every time.
Frequently asked questions
Do I need to learn special technical language to write good prompts?
No, it's mostly about communicating clearly, the same way you'd explain a task to someone new on the team: context, goal and expected format, without assuming information the AI doesn't have.
Do long prompts work better than short ones?
Not necessarily because of their length, but because of the relevant information they include. A short prompt with just the right context can work better than a long one full of details irrelevant to the specific task.
Is it worth saving prompts that work well to reuse them?
Yes, having a small library of tested prompts for repetitive tasks (writing a certain type of email, summarising a certain type of document) saves a lot of time compared to reformulating each instruction from scratch every time.