I remember my first time playing around with ChatGPT. I typed “write an intro for my newsletter” and got back three paragraphs of the most robotic, soulless text I had ever seen.
It started with “In today’s fast-paced digital landscape.” I deleted it immediately.
Everyone thinks AI is magic until they actually try to use it for real work. Then they realize it reads like a boring corporate brochure. Stop treating it like Google. It isn’t a search engine.
That is usually when you discover “prompt engineering.” It sounds like a highly technical degree you’d get from MIT. Really, it’s just the art of giving better instructions.
If you give a computer vague instructions, it gives you a vague, boring answer. Prompt engineering is how you fix that.
The hidden layers: System, User, and Assistant
Before we look at the actual tactics, you need to understand how these AI models listen to you. Think of it like directing a play.
There are three main roles happening behind the scenes.
The System Prompt is the director giving the actor their core identity. (“You are a cynical New York copywriter.”) The User Prompt is you, handing the actor their specific task for the scene. (“Write a 50-word ad for dark roast coffee.”) The Assistant Prompt is the AI actually talking back to you. In standard chat interfaces, this is just the answer it prints on the screen.
Most people only ever type in the User box. They ignore the System prompt entirely.
This is where things usually break. If you don’t tell the AI who it is supposed to be from the start, it defaults to a polite, incredibly boring digital assistant. You end up fighting the AI’s natural instinct to be overly helpful and wordy. It tries to please everyone and ends up sounding like nobody.
3 core techniques to get better answers
You don’t need a massive library of 500 prompt templates you bought on Twitter. You just need to change how you ask. Here are the three techniques I actually use every day.
1. Give it a specific persona
Don’t just ask the AI to explain a concept. Give it a job title first.
If I want to understand a complex coding issue, I don’t just ask “how does this database work.” I type: “Act as a grumpy senior software engineer explaining this to a junior developer. Keep it brief.”
The tone instantly changes. It stops sounding like a Wikipedia article and starts sounding like practical advice. The persona forces the AI to filter its knowledge through a specific lens. It restricts the vocabulary it is allowed to use.
2. Put it in a tight box
AI loves to ramble. Left alone, it will give you a ten-point list for a simple question. It hands you a 1,200-word essay when you really just needed a quick Slack update.
You have to constrain it. Tell it exactly what not to do. “Write a summary. Do not use the words ‘delve’, ‘landscape’, or ‘unlock’. Keep it strictly under three sentences.”
This looks simple. It usually isn’t. Figuring out the right constraints takes some trial and error. But forcing boundaries is the fastest way to make the output actually usable. If you don’t build the box, the AI will wander all over the place.
3. Show, don’t just tell
This part often gets ignored. We try to describe the format we want using fifty different adjectives. “Make it punchy, but professional, but also slightly casual.”
It rarely works. Data scientists call the fix for this “few-shot prompting.” I just call it giving the intern an example.
Instead of describing the vibe, just show it. If you want it to write a specific type of cold email, paste two of your actual best cold emails into the prompt first. Tell the machine: “Analyze my writing style in these two examples, then write a new email for this new product.”
It learns by seeing your examples much faster than it learns by reading your long list of rules. You wouldn’t ask a human designer for a logo without giving them a mood board. Treat the AI the exact same way.
Why this actually matters
AI is only as smart as the person typing the instructions, so if you want better answers, you have to ask better questions. The tool doesn’t do the work for you; it just amplifies how well you can explain what you want.
