Most people type one sentence into AI, get an average result and decide the tool is not useful. Usually, the problem is not the tool. It is the instruction.
A good prompt has a few clear parts, and your team can learn them quickly.
The recipe
Imagine you are giving a task to a capable colleague who does not know your context. The better you explain it, the better the output.
1. Role - who should AI act as?
Start with the perspective. "You are an experienced salesperson." "You are an accountant." This helps AI choose the right angle.
2. Task - what exactly do you want?
Be specific. Not "write an e-mail", but "write a short follow-up e-mail to a client who has not replied for two weeks".
3. Context - what does AI need to know?
Add the goal, audience, tone and any examples. Context prevents generic answers.
4. Format - what should the output look like?
"Three bullets." "Maximum 100 words." "A table." If you do not define the format, AI will guess.
Example
You are an experienced salesperson. Write a short, friendly follow-up e-mail to a client who booked training last year and has not been in touch for six months. The goal is to reconnect and offer a follow-up course. Maximum 120 words, informal tone, no pressure.
Try this and compare it with a one-line instruction. The difference is huge.
We train this on real e-mails and documents from your company. Get in touch.