The Art of the Prompt

When I first started using ChatGPT, I did what most people do. I played with it. Asked it silly questions. Marvelled at the fact that it could hold a conversation. Then the novelty wore off and I realised something: I was spending more time going back and forth trying to get a useful answer than I would have just doing the thing myself.

That was the moment I started taking prompting seriously.

The way we prompt has changed

A couple of years ago, prompting felt like a dark art. The difference between a good prompt and a bad one was enormous. Models were less capable, more literal and easily thrown off by vague instructions. People were writing elaborate prompts with specific formatting rules, temperature settings and carefully structured inputs just to get consistent results.

That gap has narrowed significantly. The models have got smarter. They're better at inferring intent, filling in gaps and understanding context. You can have a fairly natural conversation with Claude or ChatGPT today and get something genuinely useful without having engineered every word.

But here's the thing - prompting still matters. Especially when you're using AI for anything beyond a quick question. The better your prompt, the less back and forth you need, the more consistent your results and the more useful your AI actually becomes.

I learned this properly when I started building AI automations for other businesses. A vague prompt in a chatbot that's supposed to handle customer enquiries isn't just annoying - it's a business problem. That's when I developed a framework I've been using ever since.

The ICI Framework

ICI stands for Identity, Capability and Interaction. It's a three-layer approach to building prompts that work - not just once, but consistently, across different users and different situations.

Identity - who is the AI?

The first layer defines who the AI is and what it's there to do. Not just "you are a customer support assistant" but something with real context: what the business does, who it serves, what the AI's purpose is and how it should approach its role.

The difference matters more than you'd think. A vague identity produces vague responses. A clear identity - one that gives the AI a genuine sense of purpose and expertise - produces responses that feel considered and on-brand.

Capability - what can it do?

The second layer defines the boundaries. What should the AI help with? What should it refer elsewhere? What sources should it draw from and what decisions should it never make on its own?

This is the layer most people skip. They define who the AI is and how it should talk, but they don't tell it what to actually do when things get complex. Without a capability layer, your AI will either overpromise or go silent when it hits something it's not sure about. Neither is good.

Interaction - how does it communicate?

The third layer covers tone, style and conversation flow. How does it open? How does it handle ambiguity? What does it do when someone is upset or confused? How does it wrap things up?

This is what makes an AI feel human rather than robotic. The capability layer makes it useful. The interaction layer makes it trustworthy.

Why this still matters even as AI gets easier

You might be thinking - if the models are getting smarter, do I still need a framework like this?

For a quick task, probably not. Ask Claude to summarise something and you'll get a good summary. No framework required.

But the moment you're building something - a chatbot, an automation, an AI assistant that represents your business - the ICI Framework is the difference between something that works reliably and something that works most of the time. Most of the time isn't good enough when real customers are on the other end.

And even for everyday use, thinking in these three layers - who am I talking to, what do I want it to do, how do I want it to respond - will make you a faster, more effective AI user. You'll get better results with less back and forth. Which was the whole point.

Try it yourself

I've built a free tool called Prompt Architect that walks you through the ICI Framework and helps you build your own prompts. Just enter your name and email for access.

Access Prompt Architect here

If you want to go deeper on building with AI - tools, workflows and the stuff that actually works - come and join us at Vibe Coding Lab.