This article started as a two-minute ramble into speech-to-text.
No prompt. No framework. Just me talking to my AI assistant the way I’d talk to a colleague, with all the extra words, half-finished thoughts and misspelled names that come with dictation.
It understood exactly what I meant.
I bring that up because I see the opposite all the time. Someone spots a prompt on a slide or in a LinkedIn post and copies it word for word, like it’s a magic spell. Change one word and maybe it stops working. So they don’t change anything. And when there’s no prompt to copy, they freeze.
That hesitation is what holds most people back. Not the AI. The fear of talking to it wrong.
I see it in companies rolling out Copilot. Employees open that empty prompt box and just stare at it. The usual fix is to hand them a library of 200 prompts. That helps a little, but it doesn’t fix the freeze. It gives them something to copy instead of showing them they can simply ask.
What I’ve learned is simple. AI cares about your intent far more than your wording. If you can explain what you want to a smart person, you can explain it to AI. Use your own words. Ramble if you need to. Give it the background you’d give a new hire. That’s it.
Now, I’m not saying frameworks are useless. I use one myself. I call it CRISP: Context, Role, Instruction, Structure, Parameters. It’s a really good checklist, especially when the stakes are high or the results aren’t landing. But it’s a checklist, not a script. You don’t need all five every time. Use what makes sense and what gets you good results. Same with prompt libraries. They’re training wheels. Great for getting started, not something you ride forever.
The other piece people miss is that you can train your AI the same way you’d train a person.
My assistant’s name is Albert. Over months, Albert has learned how I write, the tone I want, the words I never use and how I like things delivered. When something comes back wrong, I don’t rewrite my prompt from scratch. I tell Albert what was off, the same way I’d give feedback to someone on my team. Next time it’s better.
To be fair, that didn’t happen on day one. It took the same thing it takes with a new hire. Repetition and feedback. And it depends on the tool. Some AI tools remember your preferences across conversations, others barely remember anything. If yours doesn’t, keep your preferences somewhere you can paste them in, and keep correcting it. You’ll still see the difference.
The people who get the most out of AI aren’t the ones with the best prompts. They’re the ones who stopped worrying about prompts and started having a conversation.
So stop copying. Start talking.
I’m curious. What’s the first thing you’d tell a nervous employee staring at an empty Copilot box?