Writing good prompts for the AI process description

Modified on Tue, 09 Apr 2024 at 11:43 AM

The prompt is the instruction that the user gives an AI as input. It is crucial for the quality of the result to create good prompts. Large language models react sensitively to the formulation of instructions Small changes in the formulation can result in differences in the generated output. A good prompt should consist of the following elements: 


  1. Instruction (by Q.wiki)
  2. Input data (by the user)
  3. Context (by the user)
  4. Output indicator (by Q.wiki)



The Q.wiki AI already knows the instruction and the output indicator. Therefore, the user does not have to worry about this. The user in Q.wiki concentrates on the input data and the context that supports the input data. 


There are tips for the input data and the supporting context, following which will improve the result. The prompt should have the following properties: 


  • Structured: Break down large and complex tasks into subtasks. Describe the subtasks in separate paragraphs.
  • Specific: The more descriptive and detailed the prompt, the better the results.
  • Relevant: Too much unnecessary detail is still not a good approach in terms of character limitation and "attention" of the AI. The AI is trained by humans and processes the information at the beginning and end very well, as this is where attention is at its highest, just like in humans.
  • Human: AI is trained by humans and therefore understands the input better the more human-orientated it is described. Imagine you want to roughly explain the process to a new employee. Use these formulations for the prompt.
  • Iterative: You should be aware that "prompt engineering" is often an iterative process. If you are not satisfied with the results, you can discard the page, change and improve the prompt and then generate a new page.


The scope of the prompt can be used to consciously control the creativity of the AI. The more detailed the prompt describes the desired results, the closer the generated result remains to the input. If you want to give the AI more room for creativity, formulate the prompt more briefly and deliberately omit detailed information



Was this article helpful?

That’s Great!

Thank you for your feedback

Sorry! We couldn't be helpful

Thank you for your feedback

Let us know how can we improve this article!

Select atleast one of the reasons

Feedback sent

We appreciate your effort and will try to fix the article