Universal Prompt Format: Training Your AI Intern for Better Responses
Ever tried using an LLM to create marketing content, only to feel like it’s missing the mark?
You’re not alone. Think of an LLM chatbot like this: you’ve just brought on a new intern who’s fresh out of school. They graduated with straight A’s, they know their stuff inside and out, and they’re super eager to help. But here’s the thing—they don’t know anything about your specific goals or how your business works.
To get the best out of this intern, you need to give them some direction. You’ve got to be clear about what you need them to do, why it’s important, and how it fits into the bigger picture. Once they’ve got that context, they can really run with it and deliver exactly what you’re looking for.
That’s pretty much how you should approach working with a large language model like ChatGPT. It’s smart and capable, but it needs your guidance to give you the answers that matter. In this article, I’ll walk you through a simple framework for creating prompts that’ll help you get the most relevant and useful responses.
Today's Smarter Models
With today's more advanced AI models, like ChatGPT 4.0, you have the ability to set up memories. This means that once you provide the AI with detailed context about who you are, what your business does, and your specific preferences, the AI can retain this information across sessions. This feature allows for more seamless and personalized interactions, as the AI can recall important details without needing you to repeat them in every conversation.
In this article, we focus on the framework for creating effective prompts. However, in a later article, I'll cover how to set up and use the memory features in ChatGPT 4.0 to further enhance your experience.
The Origin of the Prompt Ladder
The Prompt Ladder format was developed in collaboration with Ben Fisher , co-founder of Steady Demand , and Steve Wiideman , Founder and CEO of Wiideman Consulting Group . We created this framework while preparing for a webinar focused on various ways to Use AI to grow your business online. You can watch the recording here.
The goal was to design a prompt that would train the AI to act as an SEO expert and develop a comprehensive SEO strategy.
We found that a structured approach to prompting could significantly improve the relevance and quality of AI responses. The Prompt Ladder below emerged as a way to guide the LLM step-by-step, ensuring that the output aligns closely with the specific needs of the task at hand.
Also, if you’re looking to grow your business using cutting-edge SEO strategies, I highly recommend checking out the agencies led by Ben and Steve. They’re two of the top experts in the industry, and even better people.
Step 1: Role Definition
The first step in the prompt ladder is defining the role you want the AI to take on. Think of this as giving your eager intern a specific job title or role. By doing this, you’re telling the AI which "hat" to wear, helping it focus on the most relevant information.
When you define a role, you’re narrowing the scope of the AI’s responses. Instead of trying to pull from all its knowledge at once, it zeroes in on the specific area that you need help with. This leads to more targeted and relevant answers, making the interaction more efficient and useful.
Example:
Let’s say you’re the owner of a local landscaping business and you want to launch an email campaign promoting your new services. A less effective prompt might be something like:
"Write me an email about my business’s new services."
While this is a start, it’s very broad and doesn’t give the AI much direction. Instead, you should define the role more clearly by saying something like:
"Pretend you are an email marketing expert who specializes in increasing open rates and click-through rates for local businesses in the landscaping industry."
This way, the AI knows to focus on strategies and content that are not just about writing an email but about crafting one that will engage customers in the landscaping industry specifically. Let’s keep going.
Step 2: Task Specification
After defining the role, the next step is to clearly specify the task you want the AI to perform. This is where you tell your "eager intern" exactly what you need them to do. The more specific you are, the better the AI can tailor its response to meet your needs.
Task specification is crucial because it eliminates ambiguity. If you’re vague about what you want, the AI might deliver something that’s technically correct but not exactly what you had in mind. By being specific, you guide the AI to produce content or answers that are directly aligned with your goals.
Example:
Let’s say your business specializes in landscape architecture and design and you want to promote a new service in your email campaign. Let’s add that to the next step of our prompt ladder:
"Pretend you are an email marketing expert who specializes in increasing open rates and click-through rates for local businesses in the landscaping industry.
领英推荐
Write a promotional email for our landscaping business that highlights our new custom landscape design services, focusing on eco-friendly materials and low-water plants, and encourages homeowners to schedule a consultation."
This gives the AI a clear direction on what the email should include, which aspects of the service to emphasize, and what the call-to-action should be. With this, the AI’s response will be much closer to what you actually need, but there’s still more we can do to get the best possible result.
Step 3: Guidelines and Guardrails
Once you’ve defined the role and specified the task, the next step is to set any guidelines or guardrails for the AI to follow. This is where you lay out any rules, limitations, or preferences that the AI should keep in mind while generating its response.
Guidelines and guardrails help keep the AI’s output on track and ensure it aligns with your brand’s voice, tone, or specific requirements. Without these, the AI might produce content that, while relevant, doesn’t fully meet your expectations or could go off in a direction that’s not helpful.
Example:
Continuing with the landscaping business example, suppose you want the email to maintain a professional but friendly tone, avoid technical jargon, be accessible to a wide audience, and capture their attention by highlighting a potential pain point. Here’s how we can add it to our prompt ladder:
"Pretend you are an email marketing expert who specializes in increasing open rates and click-through rates for local businesses in the landscaping industry.
Write a promotional email for our landscaping business that highlights our new custom landscape design services, focusing on eco-friendly materials and low-water plants, and encourages homeowners to schedule a consultation.
Use a professional yet friendly tone, avoid technical language, keep the email under 110 words, write at a 9th-grade reading level, and be sure not to use emojis in the email. Let’s take a problem/solution approach that highlights a pain point and shows how our services are the solution they need."
By setting these specific guidelines—including what the AI should and should not do—you ensure that the content is well-suited to your brand’s style and audience. The AI now has clear instructions on the tone, length, complexity, and any restrictions, leading to a response that’s not just relevant but also perfectly aligned with your needs. Let’s move to the final step in the ladder.
Step 4: Desired Outcome
The last step is to clearly articulate the desired outcome you want from the AI’s response. This is where you tell the AI what the end goal is, so it can shape its response accordingly.
By specifying the outcome, you help the AI focus on what truly matters to you. It’s like giving your eager intern a clear idea of what success looks like. Without this direction, the AI might provide useful information, but it might not hit the mark in terms of what you actually need.
Example:
For the landscaping business, suppose your goal is to drive more consultations for your new sustainable landscape design service. Here’s how we could add it to our prompt ladder:
"Pretend you are an email marketing expert who specializes in increasing open rates and click-through rates for local businesses in the landscaping industry.
Write a promotional email for our landscaping business that highlights our new custom landscape design services, focusing on eco-friendly materials and low-water plants, and encourages homeowners to schedule a consultation.
Use a professional yet friendly tone, avoid technical language, keep the email under 110 words, write at a 9th-grade reading level, and be sure not to use emojis in the email. Let’s take a problem/solution approach that highlights a pain point and shows how our services is the solution they need.
The goal of this email is to encourage homeowners to schedule a consultation for our new sustainable landscape design services."
This makes it clear to the AI that the focus should be on crafting an email that not only informs but also persuades the reader to take a specific action—scheduling a consultation. The AI can then tailor its response to prioritize that outcome, ensuring the final product aligns with your business objectives.
Let’s see what the model comes back with.
Bringing It All Together
By following this prompt ladder, you’re setting yourself up for success when interacting with large language models like ChatGPT. Remember, the AI is like an eager intern: incredibly smart and full of potential, but it needs your guidance to truly shine.
Let’s quickly recap the steps:
By taking the time to craft your prompts using this framework, you’ll find that the AI delivers responses that are not just relevant but also highly tailored to your specific needs. Whether you’re working on email marketing for a landscaping business or any other project, these steps will help you get the most out of your AI interactions.
So next time you sit down to work with a large language model, remember this prompt ladder. It’s your tool for turning an eager, knowledgeable AI into a valuable partner that helps you achieve your goals.
Author’s note: I used ChatGPT’s Voice mode to help brainstorm ideas and format this article. If you’re curious about the prompts I used or how to leverage AI tools like this in your own content creation, feel free to message me on LinkedIn!
(Image Prompt: A photograph of a person standing next to a robot that looks like an AI intern, pointing at a whiteboard with instructions. The robot is attentive and eager, ready to learn. The whiteboard has simple diagrams or flowcharts representing tasks or instructions. The scene is in a modern, minimalistic office setting with a collaborative and professional atmosphere. The colors should be soft and inviting, with a focus on the interaction between the human and the AI. --v 6.1 --ar 4:3)