The prompt 'engineering' myth: what actually gets better answers from AI

There's a small industry now selling "prompt engineering" — courses promising secret phrases that unlock the AI's hidden power. Most of it is nonsense dressed up as expertise. After a lot of real use, the honest truth is that getting good answers from AI has almost nothing to do with magic words and almost everything to do with three unglamorous habits.

Habit 1: Give context, because the AI can't see your life

The single biggest mistake people make is treating an AI like a search engine — throwing three words at it and expecting it to read their mind. It can't. It knows nothing about you, your business, or your situation unless you tell it.

Compare these two:

• "Write a message to a customer."
• "Write a short, polite WhatsApp message to a customer who ordered a laptop, paid a deposit, and whose delivery is delayed by two days because of a supplier problem. Friendly but professional."

The first gives you generic filler. The second gives you something you can almost send as-is. Same AI, same "skill" — the only difference is that you did the AI the courtesy of explaining the situation. Context is 80% of the game, and it isn't a trick. It's just telling the whole story.

Habit 2: Show an example of what "good" looks like

The second habit feels almost too simple: if you want the AI to produce something in a certain style, show it one example of that style. Paste in an invoice you like and ask it to make a new one in the same format. Paste a sentence you wrote and ask it to continue in your voice.

This works because AI is fundamentally a pattern-matcher. Give it a pattern to match and it will; leave it guessing and it defaults to the bland average of everything it has ever read. One good example beats a paragraph of instructions describing what you want.

Habit 3: Treat the first answer as a draft, never the final

Here's where most people quit too early. They ask once, get a mediocre answer, and conclude "AI isn't that good." But nobody expects a human assistant to be perfect on the first try either. The power is in the back-and-forth.

"Make it shorter." "That's too formal, relax it." "Good, but remove the last paragraph." "Now give me three different versions." Each reply steers the AI closer to what you actually wanted. The people who get remarkable results aren't using secret prompts — they're just having a conversation instead of demanding a miracle in one shot.

What this means for you

If you want to actually use AI well — for your business writing, your studies, your side projects — skip the paid "prompt engineering" course. Instead, practise these three things on a real task this week:

1. Explain the full situation before you ask.
2. Show one example of what good looks like.
3. Push back on the first answer until it's right.

That's it. That's the whole skill. AI rewards people who communicate clearly and iterate patiently — which, conveniently, are the same habits that make you better at working with humans.

If you're a student wanting to go deeper on using AI to actually learn rather than cheat, we wrote a whole guide on that over on the blog — and if you're curious about how the software behind these tools is built, our free code playground is a good place to start experimenting.