What it means in practice
When you open ChatGPT and type “Write me an email to a client who is late paying”, you have just written a prompt. It is your input instruction to the AI.
The biggest mistake we see in companies is that employees use AI like a Google search. They write one-word or very short prompts and are then disappointed with the result. The basic rule of working with AI is: bad input means bad output. Write a detailed prompt with context and the AI will generate exactly what you need.
Why companies need to know this
- Time saved. Employees who can write good prompts get a usable result on the first attempt. Those who cannot spend hours arguing with the AI and end up doing the work themselves.
- Output quality. A good prompt makes sure the output is in the right tone (professional rather than too casual, say) and contains the right information, which reduces the risk of hallucinations.
- Standardisation. You can build a library of proven prompts (a prompt library) for recurring tasks, so every employee gets equally good results.
How to write a good prompt
- Give the AI a role. Instead of “Write an article”, try “Act as a B2B marketing expert with ten years of experience. Write…”
- Add context. The AI does not know your company. Tell it who you are, who you are writing to, and what the goal is.
- Be specific about format. Tell the AI whether you want a table, bullet points, or continuous text of at most 300 words.
- Give an example. If you want the AI to write in your style, paste a sample of your earlier writing into the prompt.
- Iterate. The first prompt is rarely perfect. If the answer misses, add instructions and try again.
In our company AI workshops we teach teams not only how to write prompts, but above all how to build systems where employees no longer have to write prompts at all, because they are built into the tools they already work in.