it took me 2 years to figure out this 1 powerful AI trick

You're spending more time prompting AI than it would take to just do the work yourself.

And the outputs are still rubbish.

The mistake is trying to control AI like a micromanager controls an employee.

Breathing down it’s neck, it’ll tell you what you want to hear, not what you need to hear.

A massive prompt with 37 instructions, examples and every tiny detail accounted for?

ChatGPT will nod along like a yes man and still generate something that’s not what you wanted.

With this approach, you're doing the thinking FOR the AI.

But AI doesn't need you to think for it.

It needs you to think WITH it.

There's a single sentence you can add to any prompt that changes this dynamic completely.

It turns AI from a tool you micromanage into a co-pilot that pulls the expertise out of YOUR head instead of guessing.

I'll show you exactly what it is, why it works, and how to use it in the next 10 minutes.

By the end, you'll know how to get higher quality outputs in a fraction of the time.

Spoon Feeding

Here’s what everyone tells you about prompt engineering:

"Be specific. Give examples. Write detailed instructions. The more context you provide, the better the output."

I don’t disagree with this. I use a similar process in all of my custom instructions.

So why do they still feel like they miss the mark?

You give it your company background, your target audience, your tone of voice, 3 examples of what good looks like, and a 5-step process to follow.

All this time to get back something that’s technically correct, but with gaps you spend even more time filling.

You've now spent an hour trying to automate a 30-minute task.

This is because you’re removing yourself from the process - you know, that person with 16 years of experience in your given field?

I did this for 2 years.

Building client proposals? I'd write prompts longer than the actual proposal.

Writing anything that required my actual expertise? I'd end up rewriting 80% of it anyway.

The problem wasn't that my prompts weren't detailed enough.

The problem was that I was treating AI like an idiot who needed to be spoon-fed every instruction.

When really, I should have been treating it like a smart assistant.

Reversing the Power Dynamic

Stop telling AI what to do.

Start letting it ask YOU what it needs.

At the end of any prompt where you need quality output, add this:

That's it.

I call it ‘The Q & A Prompt’ (save it to wherever you store your prompts)

Think about it like this:

When you work with someone intelligent on a project, you don't hand them a 47-step instruction manual and say “Do this”.

They’re not following a Paul Hollywood bread recipe, deeply craving one of his signature handshakes (Google/ChatGPT the ‘Paul Hollywood handshake).

No, you give them a goal, a bit of context, and you let them ask questions.

"What's the budget?"
"Who's the audience?"
"What's worked before?"

Smart people don’t need you to think of everything.

They pull the expertise out of your head by asking the right questions.

AI works the same way.

Plus, reasoning models (like Claude Sonnet 4.5, GPT-5.1, and Gemini 3 Pro) are built for this.

They're designed to think through problems.

So, we need to start thinking in terms of AI Autonomy.

1. Enough guidance to point it in the right direction.

2. Front loaded Q & A to set it on it’s way.

This is similar to when I sold big software projects.

We’d front load the project with a large amount of workshops and solution design.

Because we knew that more detail and understanding up front…

Would lead to more time savings, and more accuracy down the line.

Here’s some examples of how I’ve used this recently.

Examples

Note - All of the screenshots below are Claude used in dark mode. I use ALL of my apps in dark mode. As a long term migraine sufferer (and someone who gets eye strain looking at screens for 8 hours a day) this is essential for me. Give it a go, your head might thank you for it - even if you’re not prone to headaches.

1. The Pricing Calculator

I’m not very good with spreadsheets.

But as soon as I heard Claude was getting good at this I set it a task.

1. My initial instruction (based on a very basic example spreadsheet):

2. Claude’s questions:

3. The rest of Claude’s questions plus my answers:

2. The Newsletter Drafter

I even used it for the newsletter you’re reading right now ;)

The Bottom Line

Stop micromanaging AI.

Start co-piloting with it.

Micromanaging: "Here's exactly what to do, step by step, don't deviate."

Co-piloting: "Here's what I need. What else do you need from me to get this right?"

Here’s the prompt modifier again.

"Ask me any questions you need before we start, to maximise your chances of success with this task."

Let AI pull the expertise out of YOUR head instead of guessing.

You'll get higher quality outputs in less time.

And you'll feel like AI is working WITH you, not against you.

Try it right now.

You'll know it's working when the questions AI asks make you think "wow, I’d not thought of asking that."

Until next time.

Adam

P.S. - What did you think of this week’s edition? Hit reply and let me know (I read all replies)

P.P.S - In exchange for referring 5 people to my newsletter, I will give you 30 mins FREE consulting (normally $99) on anything you're struggling with that can be solved with AI & Automation. From speeding up your delivery, to working less hours, or winning more clients, we can tackle anything you like. Share using the link below.