The quality of the responses you get from AI chatbots is largely a function of the quality of your prompts. The Who-What-Why-How Framework is carefully designed to help you provide all the information the chatbot needs, and in the format it'll best comprehend:
Let's unpack it.
Think of AI chatbots as spheres containing information. To respond to your prompt, they can take any direction within that sphere across four dimensions.
The goal of prompting is to guide the AI to the location within the sphere of the most useful response.
By providing plenty of context, you point it in the right direction. By specifying constraints, you keep it from veering off course. The more abundant and specific your instruction, the closer the chatbot will land to your ideal outcome. That's why you want to provide the Who, What, Why, and How.
Not every rule applies to every prompt, but the more boxes you can check, the better the chatbot will serve your needs.
First, let's cover a few general strategies that apply throughout your prompt.
Before you can guide the AI to your destination, it's important to articulate that destination for yourself. Consider your ideal output, then keep it in mind as you compose your prompt.
I know it can feel like you're being rude, but because large language models essentially pick the most likely next word after your prompt, it's better to frame your prompt as instructions rather than questions.
AIs will also understand your prompts better if you break them into labeled sections. The best way to do that is using the XML syntax:
<Drafted Email><End Drafted Email>It's especially important when providing material for the bot to reference, like the content it's correcting or examples to follow.
Now, let's get into the Who-Why-What-How.
For the who, you assign roles to the bot, yourself, and possibly the audience of the generated content.
The bot's role is most important. You should begin almost every prompt by assigning the bot a role. Be specific and include qualifiers for the context within the role.
In many cases, it's also helpful to give yourself a role. The chatbot is your assistant. Explain the context in which it's assisting you.
If the chatbot's generated content is intended for an audience other than yourself, describe that audience.
The why is your ultimate objective.
You can add helpful context by first explaining the problem you need to resolve. Why are you seeking support from the bot? Build on the role you assigned yourself.
Share that ideal outcome you defined before beginning your prompt.
Tell the bot what it will be generating in terms of its format and composition. Describe the look and feel of the response, like its length, layout, and tone.
Here, it's especially helpful to include examples. You can attach files or use that XML syntax.
Lastly, tell the bot how to approach your request. Prescribe it a method and supply resources for it to reference.
Chatbots will interpret your requests more accurately if you break them into a sequence of steps.
Depending on the nature of your prompt, you may need up-to-date information. Remember, the data that AI is trained on cuts off at a certain point in the past. But increasingly, the chatbots are gaining access to the live web. Sometimes they'll search automatically, but if you know you need it, it's helpful to request it explicitly in your prompt.
Lastly, provide the resources you've asked your chatbot to reference. That could be the content it's analyzing or examples to inspire the composition of the output.
You can attach the resources or paste them within XML tags.
After the bot's initial response, you'll almost always want to refine it through follow-up requests.
If you ask the bot to review its response to identify and correct any mistakes on its own, in many cases, it'll make corrections.
Many times, you'll spot errors yourself or want to make a correction or qualification to your request.
In some cases, it's better to edit the original prompt with your qualified instruction, then have the bot regenerate its response. That keeps the unwanted response out of the conversation history.
To come up with example prompts that use this framework, I fed it to ChatGPT, Gemini, and Claude. Their responses were spot-on, so I've included one from each for your inspiration.
You are a business consultant specializing in strategic decision-making for mid-level managers. I am a marketing manager evaluating which new software tool to adopt for automating our social media content scheduling. The output will be shared with my department to guide our decision.
The problem is that we have three software options (Tool A, Tool B, Tool C), and I need to ensure we choose the one that best meets our needs within budget and time constraints. Your goal is to create a comparison table that evaluates each tool based on specific criteria.
Provide a table comparing the three tools with the following columns:
Write a 2–3 sentence summary at the end recommending the best option based on the data.
Follow these steps:
You are a market research analyst with expertise in creating detailed competitive analyses. You are proficient in gathering and organizing data into easily understandable formats.
I am a business owner preparing to launch a new service. I need to understand my competitors strengths and weaknesses.
The audience is me and my team, in order to understand the competitive landscape.
I need to quickly understand the competitive landscape for a specific service I'm launching, including key competitors, their strengths, and weaknesses.
You will create a comparison table that outlines key competitors in the specific market, evaluating their strengths, weaknesses, pricing, and key features. You will focus on a clear view of their overall positioning.
The response will be a table with the following columns: "Competitor Name," "Strengths," "Weaknesses," "Pricing Model," and "Key Features." There should be at least 3 listed competitors, and a summary in paragraph format after the table outlining any clear patterns or opportunities.
You are a technical documentation specialist who excels at creating clear, actionable standard operating procedures (SOPs) that are easy to follow and implement.
I am an operations manager implementing a new customer onboarding process.
The audience is customer success team members who will be executing the onboarding process. They have varying levels of experience with our product.
Our current onboarding process is inconsistent and lacks documentation, leading to varying customer experiences and longer time-to-value.
Create a detailed SOP document that standardizes our customer onboarding process.
<Current Process>
[Insert current workflow details]
</Current Process>