I see a lot of people frustrated with AI tools who are, essentially, using them like a Google search. They type a vague question, get a vague answer, and decide the tool is overhyped.
It’s not. They’re just not talking to it right.
This is the thing nobody wants to say out loud: getting good output from AI requires skill. That skill is called prompt engineering. It’s not hard to learn, but it does require you to actually try.
Here’s what I’ve learned from a year of doing this daily.
The Core Problem: Vagueness In, Vagueness Out
If you type “write me a marketing email,” you will get a generic marketing email. That output will be correct and useless.
If you type:
“Write a marketing email for a $79/month productivity app aimed at solo consultants who are overwhelmed by email and meetings. The tone should feel like a smart friend who’s been there, not a corporate brand. The CTA is a 7-day free trial. Keep it under 200 words.”
You will get something usable.
The difference is specificity. The AI isn’t trying to read your mind — it’s completing what you started. The more precisely you start, the more precisely it finishes.
The Four Elements of a Good Prompt
I use a simple framework:
1. Role — What kind of expert is the AI playing?
“Act as a senior UX designer with experience in SaaS onboarding…”
2. Context — What’s the situation?
“I’m building a landing page for a tool that helps remote teams run async standups…”
3. Task — What exactly do you want?
“Write 5 headline options for the hero section…”
4. Constraints — What limits apply?
“Each headline should be under 10 words, avoid jargon, and start with a benefit rather than a feature.”
You don’t need all four every time. But when your output is underwhelming, it’s usually because one of these is missing.
Specific Prompts That Work for Me
For generating ideas:
“Give me 10 unusual angles for a piece about [topic]. Not the obvious ones — I want approaches that would make someone who’s seen everything stop and think.”
For editing my own writing:
“Here’s something I wrote. Tell me where the logic is weak, where I’m being vague, and where I’m repeating myself. Be direct.”
For research:
“I’m going to make a claim that [X]. Steelman the counterargument — what’s the strongest case against this?”
For decision-making:
“I’m deciding between [option A] and [option B]. Here’s the context: [paste context]. What are the 3 most important things I should consider that I might be overlooking?”
The Iteration Mindset
The best output rarely comes from the first prompt. I think of prompting as a conversation:
- Send the first prompt
- Read the output critically
- Tell the AI what’s wrong or what you want more of
- Refine
“Good start, but the tone is too formal. Make it feel more like a text message from a knowledgeable friend.”
“The second option is closest. Expand on that angle in 3 different ways.”
You’re not just prompting — you’re collaborating. The people who get the most out of AI are the ones who treat it like a conversation, not a vending machine.
One Thing That Changed Everything for Me
Start your prompts with what you don’t want.
“Don’t use corporate buzzwords. Don’t start with ‘In today’s fast-paced world.’ Don’t give me generic advice.”
Most of what makes AI output sound like AI output comes from the patterns it’s trained on. Explicitly excluding those patterns is one of the fastest ways to get better results.
Try it on your next prompt and see what happens.
— Deco