Most people type a question into ChatGPT, get a mediocre answer, and decide AI isn’t that useful. I used to think the same.
The problem isn’t the AI. It’s the prompt.
Think of it like briefing a contractor. If you say “build me something”, you’ll get something – but probably not what you had in mind. The more specific and useful your brief, the better the result. AI works exactly the same way.
That’s where prompt engineering comes in. And specifically, the RACE? method – a five-step framework I keep coming back to in my own work. It gets consistently better outputs from any AI tool, and you don’t need a technical background to use it. If you can write a decent brief, you can do this.
What is prompt engineering?
Prompt engineering is simply the practice of writing better instructions for AI. Nothing fancy. No coding required.
The quality of what you get out is almost entirely determined by what you put in. Prompt engineering is just learning how to put in the right things. It’s the difference between a vague request and a well-written brief – and most people are still writing vague requests.
The RACE? method at a glance
R Role – Tell the AI who it should be
A Action – Say exactly what you want it to do
C Context – Give it everything it needs to know
E Execution – Specify the format and output you want
? Review – Ask it whether it did a good job, and if it can improve
I’d also add one more step I consider the most important of all. More on that towards the end.
Step 1
R - Role
Start by telling the AI who it should be. Not a vague “you are a helpful assistant” – give it something specific and relevant to the task you’re asking it to do.
- Don't
You are a helpful assistant
- Do
You are an experienced social media manager who works with small hospitality businesses in the UK.
AI models are trained on enormous amounts of content and can adopt very different voices and perspectives depending on how you frame the role. A general answer looks very different from one written from the perspective of a subject matter expert in your exact field.
The more specific the role, the more grounded and useful the response. Think about who would be best placed to answer your question in real life, then frame the AI as that person.
Step 2
A - Action
Tell the AI exactly what you want it to do. Be specific about the task itself.
- Don't
Write something about my business.
- Do
Write a 300-word About page for my website.
- Don't
Help me with my Instagram.
- Do
Write five Instagram captions for a family-run farm shop in Cheshire, promoting their spring produce range.
The more specific the action, the more useful the output. If you’re vague here, the AI will fill in the gaps with guesswork – and you probably won’t like what it guesses. A clear task gives it nowhere to stray.
Step 3
C - Context
This is where most people fall short. You know your business inside out. The AI knows absolutely nothing about it unless you tell it.
Context is where you put in everything relevant to the task:
- What your business does and who your customers are
- Your tone of voice, and any brand guidelines worth knowing
- Specific details that are important for this particular task
- Examples of existing content you like or want to match
- Constraints – length, platform, audience, deadline
The more context you give, the more the AI can tailor its response. Don’t worry about giving it too much – you’re far more likely to give it too little. If in doubt, add it in.
Worth doing: If you use AI regularly for your business, write a short “business context” paragraph you can paste into prompts. Cover your name, what you do, who your customers are, and your tone of voice. It saves you rewriting it every time and keeps your outputs consistent.
Step 4
E - Execution
Tell the AI exactly what to produce. Format, length, structure, platform – all of it. Don’t leave the execution to chance.
- Don't
Write me something I can post.
- Do
Write two versions of a 100-word Instagram caption, casual tone, no hashtags, with a soft call to action at the end.
If you’ve ever looked at an AI response and thought “that’s not what I meant”, it’s almost always because the execution wasn’t specified clearly enough. Tell it the format, the length, who it’s for, and how it should be structured. The more specific, the better.
Step 5
? - Did it do a good job?
After you get your first response, ask the AI to review its own work. You can ask:
- “Did you follow the brief accurately?”
- “What could you improve about that response?”
- “Were there any assumptions you made that you’d change?”
You’d be surprised how often the AI will identify its own gaps and correct them without you needing to start again. It won’t always find something worth changing, but it’s a quick check that often improves the output at no extra effort.
Think of it as a built-in quality check. You wouldn’t send client work out without reviewing it first. Apply the same logic here.
The step I consider the most important
There’s one extra line I add to almost every prompt I write, especially for anything complex or creative:
“Before you begin, ask me any clarifying questions you need to do a great job. Do not assume.”
This one changes everything.
When you invite the AI to ask questions before it starts, it will often surface things you hadn’t thought about. Missing context. Ambiguities in the brief. Decisions you’d assumed were obvious. Answering those questions before the AI starts means the output is much closer to what you actually need, first time around.
I find this especially useful for longer or more creative tasks – blog posts, strategy documents, email campaigns. The AI will frequently ask questions that make me realise I hadn’t been clear enough, even in my own head. That’s a useful thing to discover before you get back a response you have to redo from scratch.
Don’t skip this. It looks like an extra step, but it regularly saves you several more.
As an extra tip, add it to your AI rules for it to be a defualt!
What this looks like in practice
Here’s a RACE? prompt for writing an Instagram caption for a client. Compare this to typing “write me an Instagram caption for my café” and you’ll understand why the outputs are so different.
Example prompt: Instagram caption for a farm shop
R You are a social media copywriter who works with small UK businesses, writing in an approachable but professional tone.
A Write an Instagram caption for a family-run farm shop promoting their new spring café menu.
C The shop is called Green Barn Farm Shop, based in rural Cheshire. Key details: locally sourced ingredients, relaxed countryside atmosphere, welcoming for families. The caption should feel warm and authentic – not salesy.
E Write two versions of the caption. Each should be 80-100 words. Include a soft call to action. No hashtags.
? After writing both versions, review them against the brief. What would you change if you could do it again?
That’s a complete brief. The AI knows who it is, what it’s doing, who it’s writing for, what the output should look like, and it’s being asked to sense-check its own work. You’re not leaving anything important to chance.
Why this matters for your business
If you use AI tools in your business – whether for writing content, drafting emails, planning social posts, or anything else – this framework will save you time and cut down on frustration.
You don’t need a technical background to use it. You just need to slow down slightly before you type, and think about the brief you’d give a human. Because that’s really all it is. Writing a good prompt is just writing a good brief, and RACE? is a checklist to make sure you haven’t left anything important out.
Try it the next time you open ChatGPT or any other AI tool. Even working through two or three parts of the framework will make a noticeable difference to what comes back.
And if you want help building a marketing approach that actually gets results – with or without AI – get in touch. I work with small businesses across Cheshire and the North West, and I’m always happy to have an honest conversation about what would actually make a difference.
