Introduction

Before we get into creating a useful prompt a quick intro. I’ve just started a graduate certificate in AI for business. I see AI as almost a manifestation of a socio-technical system, what I’ve seen for far though is a focus on technical details, models, implementation etc. I’ve chosen to study this to balance the conversation, so that people and organisational needs aren’t left out of conversations regarding AI adoption.

The CREATE Framework

In our weekly seminar, we were introduced to the CREATE framework for creating prompts for AI. The framework is:

• C – Character Define ChatGPT’s role or expertise.
• R – Request Clearly state what you want ChatGPT to do, with specific details and context.
• E – Examples Provide examples to guide tone, style, or structure (optional but helpful).
• A – Adjustments Refine your prompt with tweaks to improve responses.
• T – Type of Output Specify format and length.
• E – Extras Add instructions for better control: e.g Explain your reasoning 

I don’t use this approach for every prompt, but if I am using AI to create a specific output (for example, looking at how it can help improve efficiency in my Wife’s business without impacting her creative agency) then it is a useful way to add more structure to a prompt. The result being that the output should have less likelihood of hallucinations or crazy answers. 

This can be quite an arduous process, so I enlist AI to help write the prompt, which I will then give to it to complete the task. This is my typical approach. 

Step 1: tell AI the about the framework

We prepare the AI and let it know we are using the CREATE framework.

I’m going to get you to help create a prompt using the CREATE prompt engineering framework. Are you familiar with it? C – Character Define ChatGPT’s role or expertise. R – Request Clearly states what you want ChatGPT to do, with specific details and context. E – Examples Provide examples to guide tone, style, or structure (optional but helpful). A – Adjustments Refine your prompt with tweaks to improve responses. T – Type of Output Specify format and length.E – Extras Add instructions for better control: e.g Explain your reasoning

Step 2: provide some context

Then I give it some context for my scenario, in this example I am using my Wife’s business as a thought experiment:

The next scenario is one I’m not familiar with, so I will need your help to fill in the gaps – you will need to become a creative design agency expert. The example is about identifying AI use cases to aid in improving operational efficiency in a small design agency. Can you help with that?

Step 3: Get the prompt

The AI may then ask some clarifying questions, once this is done, I ask it to create the prompt based on the CREATE framework. In this case it didn’t and returned this output (remember, the AI has created a prompt for itself):

C – Character You are an experienced creative design agency consultant with expertise in operational efficiency, workflow optimisation, and the practical use of generative AI in small agencies. R – Request Identify and analyse practical AI use cases that could improve operational efficiency in a small creative design agency (5–15 staff).Focus on real agency workflows such as:Client onboarding and briefingConcept developmentDesign productionFeedback and revision cyclesProject management and deliveryFor each use case, explain:What the use case isWhere it fits in the workflowThe problem it solvesThe expected efficiency gain (time, cost, or quality) E – Examples (optional but included for guidance) Example use case format: Use Case: AI-assisted client brief generation Workflow Stage: Onboarding Problem: Incomplete or unclear client briefs leading to rework Solution: AI helps structure and clarify briefs using guided prompts Impact: Reduces back-and-forth and improves first-pass design quality A – Adjustments Prioritise high-impact, low-complexity use cases suitable for immediate adoptionAvoid overly technical or enterprise-scale solutionsConsider constraints of small teams (limited budget, time, and AI maturity)Highlight any risks or trade-offs (e.g. quality, originality, client perception) T – Type of Output Provide:A table with at least 6–8 AI use casesColumns: Use Case | Workflow Stage | Problem | Solution | ImpactA short summary (2–3 paragraphs) highlighting the most valuable opportunities E – Extras State any assumptions about the agency’s maturity or toolingWhere relevant, briefly explain why a use case is realistic or commonly adoptedDo not assume full automation — focus on augmentation of human work  Ask me five clarifying questions to help your focus the response.

See that final line, about clarifying questions? I added that into the AI generated prompt. This is gold, it gets the AI to prompt you to think about things you may otherwise not have. In my example it asked me:

1. Service mix: What types of services does the agency primarily deliver?
   (e.g. branding, web design, social media content, marketing campaigns, UX/UI)
2. Client profile: Are clients mostly small businesses, corporate, government, or a mix?
   (This affects expectations, approval cycles, and risk tolerance for AI use)
3. Current tooling maturity: What tools are already in use?
   (e.g. Adobe Creative Cloud, Figma, Canva, Notion, Asana, Monday, etc.)
4. Biggest pain point: Where does the team currently lose the most time?
   (e.g. revisions, client communication, admin, ideation, production bottlenecks)
5. AI posture / appetite: Is the agency:
   1. experimenting casually
   2. actively adopting AI
   3. or cautious due to quality/client concerns

Step 4: Review the prompt 

Review the prompt it has provided you, check that it meets your requirements and modify as needed. In a real world example, I would actually talk to my wife and add details to the ‘Request’ part of the prompt.

Step 5: give the prompt to AI

Now, we have a prompt ready to give to AI. Sometimes I find that AI will offer to run the prompt for me, sometimes it won’t. Either way, I paste the prompt it has created into AI.

The AI will now process your prompt and provide you with the output. Now, you may be thinking, why did I just get AI to write a prompt to itself? The answer is, you don’t have to, you could just write the prompt yourself. But getting AI to help,  particularly if it is in relation to an area we are not overly familiar with, helps us by drawing on its knowledge-base and adding that detail to refine the prompt.