Have you ever sat there, fingers hovering over the keyboard, trying to craft the perfect AI prompt?
Maybe you’ve downloaded a 500 Best ChatGPT Prompts guide, hoping it’ll offer some magical level of AI mastery. But let’s be honest with ourselves – how often have those cookie-cutter prompts actually given us the results you need? (I know they usually miss the mark for me…)
The truth is, AI isn’t a vending machine where you insert a perfect prompt and instantly get gold. It’s more like a human junior assistant – eager and capable, but needing clear direction and back-and-forth instructions to get things just right.
You need to think of AI prompting as a dialogue, not a one-shot instruction. So let’s ditch the prompt perfectionism and focus on what really matters – having a conversation.
Shift your mindset – GenAI is your junior assistant
Somewhere along the way, you’ve been convinced that AI only works if you crack the code – that if you can get the prompt exactly right, it’ll deliver flawless results. But AI isn’t some mythical oracle. It’s a tool. A smart, fast, and sometimes frustrating tool that needs human guidance to be useful.
Think about how you’d work with a real junior assistant. You wouldn’t expect them to read your mind or get everything right the first time. You’d provide clear, simple instructions. You’d check their work, give feedback, and refine along the way.
So why treat AI any differently? Instead of getting stuck in the perfect prompt trap, trust what you already know. You’re the expert in your field. Your knowledge, experience, and intuition are far more valuable than some generic, AI-generated prompt guide.
And here’s the thing – simplicity beats complexity. You don’t need a five-page prompt covering every possible detail. A simple instruction gets the ball rolling. If the output isn’t quite right, tweak it. If it’s missing something, clarify. AI is designed for conversation, not perfection on the first try.
The power of conversation loops
A rigid, over-engineered prompt tries to predict every possible need upfront. But in reality, you’ll get better results by iterating. Instead of spending ages crafting your prompts, try this approach:
1) Give a clear starting point. Keep it concise and direct.
✅ Example: “Write a short email announcing that Sarah has been promoted to Head of Operations. Keep it warm, appreciative, and under 150 words.”
This provides enough direction without overloading AI with unnecessary details.
2) Assess the output. Is it on the right track? What’s missing?
❌ Example AI Output (Needs Refinement): “We are pleased to announce that Sarah has been promoted to Head of Operations. She has been a dedicated team member and we look forward to her continued success.”
Assessment: The message is too generic. It should highlight Sarah’s achievements and express enthusiasm more naturally.
3) Provide feedback. Adjust, rephrase, or ask AI to improve specific areas.
✅ Example: “Could you please make it more personal? Mention her biggest contribution this year as per the attached report and how she has positively impacted the team. End with an inspiring note about the future.”
This process mirrors how we guide people – because AI, like a human assistant, improves with direction. The collaborative exchange will help you achieve the right outcome faster.
Real-world example: This article itself
If you’ve been following along, you’ve already seen the conversation loop in action – because this article is a real-world example of how iterative AI prompting works.
Here’s roughly how it played out:
1) Give a clear starting point
✅ My initial prompt: “Please draft an article outline for my next LinkedIn newsletter. The topic is ‘Stop focusing on perfecting AI prompts – Focus on clarifying conversations’.
The idea is that people are so caught up in knowing the exact prompts they need, rather than trusting what they know based on experience and expertise and treating GenAI like a human junior assistant.
What instructions would they provide their human assistant? How would they explain things? No need to download massive prompt guides – because they are usually generic and populated via AI not human subject experts. It’s not about creating a massive 5-page prompt, like I’ve heard some people do.
You’re better off starting with some basic instructions and then reviewing the output, providing more details or feedback as part of the conversation loop, than perfecting the initial prompt. Because that’s what you’d do with a human.”
This prompt included my initial ideas around the topic, providing enough information to get things rolling without overloading the request with excessive details.
2) Assess the output
✅ My feedback on the first article outline: “Please tie section 4 in with section 1.”
✅ Next feedback: “Move the topic ‘Your existing knowledge and experience are far more valuable than a massive, complicated prompt.’ to section 2.”
Each round of feedback shaped the content outline, making it sharper and better aligned with the message.
3) Provide feedback and refine
✅ My next request: “Please refer to all previous resources provided to draft the introduction. Ensure it’s in the HumanEdge brand voice and you feature rhetorical/hypothetical questions to create engagement from the very beginning. It needs to be a relatable topic.”
✅ Introduction drafted and refined further: “Here’s my edited version – please draft the next section while avoiding repetition.”
✅ More refinement: “Still some repetition here. What can you cut out to streamline it?”
✅ Further iteration: “Please combine these sections instead of adding a new one.”
✅ Even more refinement (Referring to the conversation loop): “Could you please provide two more examples?”
At no point did we aim for perfection in a single step. Instead, we worked through a conversation loop – starting with a basic draft, reviewing, adjusting, and refining until it reached the best possible version.
And guess what? This is exactly how you should approach AI in your own work.
Final takeaway – Stop over-engineering your prompts
Next time you find yourself agonising over getting your prompt perfect, stop yourself. Start with simple instructions, review the output, and refine as you go. That’s how you’ll get results – without the stress.
Want to learn more about this process and understand how to utilise your GenAI tools more effectively? We’re now offering one-hour training sessions (online or face-to-face if you’re in Sydney), for only AU$500+GST.
I look forward to experiencing those ah-ha moments with you!
P.S. This article was written in collaboration between ChatGPT and my human brain.
