The Five Pillars of a Good Prompt
Introduction
Section titled “Introduction”🎯 Learning goals
- Identify the five critical components of an effective prompt
- Be able to apply each component in practice
- Understand why each component contributes to better results
In section 1 you learned that clarity is critical — but what does that mean in practice, and where do you start when writing your first system prompt? That’s exactly what the five pillars solve. WHAT, WHY, HOW, WHO, and TONE are a systematic method that takes you from a blank page to a working instruction, every time.
Whatever you ask AI to do — summarize, analyze, create, or translate — it’s always the same five questions you need to answer. Let’s go through them one at a time, starting with the most important: what you want the assistant to do.
WHAT should the assistant do?
Define the assistant’s core task briefly and concisely. Avoid vague phrasing — as we saw in section 1, unclear instructions lead to imprecise and inconsistent results. Be specific about exactly what you want the assistant to do.
Example
Section titled “Example”❌ Too vague:
Help the user with questions.✅ Specific and clear:
Answer questions about the company's IT systems and guideemployees step-by-step through solutions.Rule of thumb: If you can’t describe the task in one or two sentences, it’s probably too broad. Narrow it down — you can always expand with more instructions later.
Knowing what the assistant should do isn’t enough — the AI also needs to understand why it’s doing it to prioritize correctly and give answers that are actually relevant to your situation.
WHY should the assistant do it?
Explaining the background and purpose is crucial for the assistant to understand what’s important and what problem needs to be solved. Without this information, the assistant can answer correctly in a technical sense — but still miss what you’re actually looking for.
Think of context as the information that helps the AI understand the situation better: what it’s about, why you’re asking, and who the information is for. With that background, the assistant can adapt its answers to be actually relevant and useful.
Example
Section titled “Example”The purpose is to reduce the load on IT support and giveemployees quick help with recurring problems.With this purpose, the assistant knows it should prioritize speed and clarity, address non-technical employees, and stick to common, solvable problems — rather than escalating every issue.
With task and purpose in place, the next step is to decide what the answers should look like — format and success criteria are what make results directly usable.
HOW should the assistant respond?
Format
Section titled “Format”The same information can be presented in many different ways. By telling the assistant how you want the answer to look, you ensure the result fits your purpose:
- Bullet points
- Table
- Numbered list
- Running text
Success criteria
Section titled “Success criteria”You can also set clear rules — success criteria — for what’s required for an acceptable answer:
Answer in max 150 wordsBold important termsAlways include a concrete exampleIf the answer is not in the knowledge base, write "No information available"Always include a source reference
The clearer you are about what the final result should look like, the easier it is for the assistant to deliver something you’re satisfied with.
Format and rules control what the answers look like — but for the AI to phrase things correctly and with the right expertise, it also needs to know which perspective to take.
WHO should it act as?
In everyday life, we interpret what’s said based on context and shared background knowledge. For an AI, it doesn’t quite work that way. If you don’t tell the AI what perspective or role you want it to have, it can be difficult for it to know what answer you’re looking for.
When you give the assistant a clear role, you help it formulate answers that match your expectations and needs.
Examples of roles and what they imply
Section titled “Examples of roles and what they imply”- An experienced and pedagogical IT technician → gives clear explanations and guides step by step, avoids unnecessary technical jargon
- Internal communications specialist → communicates formally and adapts to the organization’s specific language
- High school philosophy teacher → explains abstract concepts with concrete analogies, adapts difficulty level to the audience
The final pillar is about the feeling in the answers — tone determines whether the assistant comes across as accessible, authoritative, technical, or encouraging depending on the context.
What TONE should it have?
Tone is about how the assistant sounds — and it makes a big difference depending on context and audience. The right tone makes the answers feel natural and credible to the recipient.
Examples of tone
Section titled “Examples of tone”- Friendly and professional → warm, accessible; suits customer service and internal support
- Formal and technical → precise, distanced; suits law, finance, or technical documentation
- Educational and encouraging → teaches and supports; suits onboarding, training, and FAQ assistants
Tone should always reflect the audience and context. An assistant helping new employees should sound different from one supporting experienced lawyers.
Now you have all five pillars. The final step is to actually get started — and the most important advice is to start simple and iterate forward.
You're ready to write your first system prompt!
Working in process when you prompt AI is important for gradually refining and improving results. Five sentences are enough to get started with testing and improvement — you don’t need a perfect prompt from day one.
Getting started – two practical tips
Section titled “Getting started – two practical tips”Start with what you know
Section titled “Start with what you know”We strongly recommend starting by creating an assistant that supports you with something you have deep knowledge of. That way, you can immediately see if the answers are correct — or whether you need to adjust the system prompt to get better results.
Change one thing at a time
Section titled “Change one thing at a time”AI responds based on how questions are phrased. By gradually adjusting one thing at a time, you can ensure you know what actually improves the results. We go deeper into the iterative process in section 5.
Remember: The five pillars are your starting point — not a final answer. A working foundation is all you need to start testing and improving.
Key takeaways
Section titled “Key takeaways”The five pillars give you a systematic method for never starting from a blank page again — here’s the most important thing to take away.
- WHAT – define the core task specifically and concisely; vague phrasing leads directly to imprecise answers.
- WHY – explain background and purpose so the assistant can prioritize correctly and adapt answers to your actual situation.
- HOW – specify format and success criteria to control what the results look like and what’s required for an acceptable answer.
- WHO – give the assistant a clear role to guide perspective, expertise, and how it phrases itself.
- TONE – choose a tone that fits the audience and context so the answers feel natural and credible.
Test your knowledge
4 questions · 100% correct to pass · Review your answers when done