The prompt audit that will change how you use AI

Try this.

Write a prompt. Put it into ChatGPT or Claude. Take the output, don't read it yet, and drop it into a second AI tool. Ask it one question: what prompt produced this?

Now read what comes back.

That's your reverse prompt audit. And if you've never done it before, what you see next will probably make you wince.

The output reveals everything you didn't say. The context you assumed. The constraints you forgot to mention. The bias you baked in without realising. It's like sending a brief to a designer and getting the work back, and only then understanding how vague your brief actually was.

The tools are fine. The prompts are the problem.

What most designers are actually doing

They open a blank prompt, type a half-formed thought, get something generic back, and either use it anyway or close the tab and go back to doing it themselves.

This isn't laziness. It's what happens when nobody shows you a better way.

When you give AI no context, it answers for everyone. It doesn't know your product, your users, your constraints, or what the output actually needs to do. So it gives you the median answer. Technically fine. Completely useless.

The designers getting real value from AI aren't more technical. They treat prompts the way they treat briefs. Context. Constraints. A clear ask. That's a design skill. It just hasn't been taught yet.

This isn't about AI replacing you

Let's be clear about something. This isn't about AI generating your designs or doing your thinking for you. It's not about prototyping tools or image generation or any of the stuff that tends to dominate the conversation.

It's about the parts of your workflow that take time and don't need to. Synthesis. Documentation. Framing. Sense-checking. Getting a first draft of something so you can spend your energy making it better rather than starting from scratch.

Prompt thinking is about augmenting how you work, not replacing why you're there.

What changes when you get it right

The moment you start treating a prompt like a brief, with a role, a situation, a constraint, and a clear output, everything shifts. The responses get specific. They start to sound like they were written for your problem, not a generic version of it.

The reverse prompt audit is one way to get there. But it's a starting point. The bigger shift is building a set of frameworks you can reach for consistently, so you're not reinventing the prompt every time you need something.

That's what separates designers who get compound value from AI over time from those who dip in, get burned, and give up.

Why I wrote the book

I've been using AI inside live product work for two years. Across research synthesis, design critique, stakeholder documentation, concept exploration. What I kept finding was that the quality of the output was almost entirely determined by the quality of the setup.

So I wrote it down.

Designing with AI: Prompt Thinking for Product Designers is a practical guide built around one idea: prompting is a design skill, and you can get better at it.

It covers the mental models, the frameworks for structuring prompts that actually work, and 33 real prompts I've used in my own workflow. Not theoretical examples. The actual prompts, mapped to the stages of the design process where I've found them most useful.

If you've been getting underwhelming results from AI, this is probably why. And it's fixable.

____________________________________________________________________________________

Ebook £7.99 on Amazon: https://www.amazon.co.uk/Designing-AI-Thinking-Product-Designers-ebook/dp/B0GTZVQBNQ/