Prompting isn’t just about typing a question and hoping for the best. It’s a skill. Whether you’re drafting emails, generating ideas, or writing user stories, the quality of your prompt shapes the quality of the output. Stef – our AI specialist – walks you through the do’s and don’ts of prompting, to make sure your results are accurate, relevant, and actually useful.
Do
- Treat ChatGPT like a high-potential intern.
Give clear instructions, provide background information, and specify the expected outcome. The more specific your prompt, the more accurate – and useful – the outcome.
- Break complex tasks into smaller steps.
Divide larger assignments into manageable prompts and have ChatGPT assist you step by step. It makes the output more accurate and allows you to adjust and guide the process step by step.
- State the goal of your prompt clearly.
Explain what you want to do with the output or who the output is intended for. This way, the response will be tailored to your needs. Without context, you may get generic or less relevant results.
- Assign a role to ChatGPT, if helpful.
The context helps guide its tone, language, and approach to better match your needs. Add for example:- “You are a project manager” – The focus will be on planning, prioritization and coordination.
- “You are a client with limited technical knowledge” – The answer should be jargon-free and easy to understand.
- “You are a legal advisor” – You’ll get a response based on legal information, in a formal tone.
- Use examples to guide the output.
Include a sample of the tone, structure, or content you’re aiming for. By showing what you expect, you help ChatGPT match your style and avoid guesswork.- If you want a formal email, you might say: “Write it like this: ‘Dear [Name], I hope this message finds you well…’”
- For a casual LinkedIn post, you could add: “Here’s the vibe I’m going for: ‘Big news! We’re thrilled to announce…’”
- If you want bullet points, you can say: “Format it like this: • Quick fact • Another detail.”
- Think of prompting as an iterative process.
Give feedback, adjust, and refine until the result is useful. You can ask ChatGPT to revise or clarify its response if it’s not quite right. Iterating helps improve the outcome and narrows in on exactly what you need.
Don’t
- Don’t start without context.
Avoid empty or vague prompts. Always provide meaningful input and direction. If you give vague prompts like “Write something for my project,” ChatGPT won’t know what you need. Make sure you add background info, goals, and any specific requirements to get relevant and useful results. - Don’t send empty prompts.
Avoid prompts like “Thanks for your answer”. Non-functional prompts like this use up system resources without adding value to the conversation. If you want to continue or refine the output, ask a follow-up question or give new input. - Don’t expect AI to “read between the lines”.
ChatGPT cannot guess your intentions. If you don’t specify the audience, tone, or purpose, the result may not suit your needs. Be explicit about the topic, audience, tone, and purpose. - Don’t try to cram a complex task into one massive prompt.
Work clearly and in stages. Breaking down a big task into smaller steps helps you guide the process and get better results at each stage. Huge, unclear prompts can overwhelm the model and lead to messy or incomplete output. - Don’t leave tone or style up to chance.
If you don’t say whether you want formal, casual, technical, or playful writing, ChatGPT will guess — and it might not match your expectations. Ask explicitly for the tone, writing style, or structure you need.
Let’s prove we’re right
The right way
The prompt
You are an experienced functional analyst working on a software project. You receive a vague requirement from the business: “The system should make it easier for customers to update their profile.”
Use your expertise to convert this requirement into a clear user story with well-defined acceptance criteria.
Work according to the INVEST principle and provide your output in two separate sections:
- User Story: Write a concrete user story in the format: “As a [type of user], I want to [goal/action], so that [value/result].”
- Acceptance Criteria: Use the Given / When / Then format. Include at least three relevant scenarios that describe when the story can be considered successful. Present the acceptance criteria in a table format.
Make sure that:
- The user story and acceptance criteria are functionally testable
- No technical implementation details are included
- The language is simple and understandable for all stakeholders
The output
The wrong way
The prompt
Could you write a user story for me about updating customer profiles?
The output
The do’s & don’ts of prompting in a nutshell
To get the best results with ChatGPT, be clear and specific in your prompts, provide context, and break complex tasks into smaller steps. Use follow-up questions to refine results and always review outputs critically for accuracy and relevance. Avoid vague instructions, don’t expect perfect answers on the first try, and don’t rely on AI for confidential or sensitive decisions without human oversight.
Ready to start your AI journey? Whether it’s for your day to day work or on a larger business scale, we’re happy to help!