One bad answer doesn't mean AI can't help you
Most people try AI once, get a weird or useless response, and quietly decide it's not for them. This is the most common mistake beginners make.
Getting a bad first answer isn't a sign that AI failed. It's a sign the conversation is just getting started.
Prompting is almost never a one-and-done situation. — Tina Huang
Think of it as a conversation, not a vending machine
A vending machine either gives you the snack or it doesn't. There's no back-and-forth.
AI is more like talking to a person. You ask something, they respond, you react to that response, and together you home in on what you actually need.
The first answer is a draft, not a final product. Your job is to react to it.
The E and I in TCRIE: Evaluate and Iterate
Tina Huang's TCRIE framework ends with two steps most beginners skip: Evaluate and Iterate.
Evaluate means: read the answer the AI gave you. Do you like it? Does it feel right? Is something off? Be honest about what's missing.
Iterate means: go back and refine. This isn't starting over — it's steering. You're telling the AI what to fix.
How to give useful feedback to AI
Don't just say "that's bad, try again." That's as unhelpful as the vague prompts we talked about before. Be specific about what's wrong.
"This is too formal — rewrite it in a casual, conversational tone." Or: "The second paragraph is great, but the intro is too long — cut it in half."
The more specific your feedback, the more targeted the fix.
Copy the part of the answer you liked, paste it back in, and say 'keep this part but change [specific thing].' AI responds well to precise direction.
Going back to your original prompt
Sometimes the answer is off because the original prompt was missing something. Ask yourself: did I give enough context? Was the task clear? Did I include an example?
Often a weak answer points to a missing ingredient in the prompt. Adding one thing — a persona, an example, a constraint — and running the prompt again can completely change the result.
Separating your prompt can help too
One of the quickest fixes is also one of the simplest: break a long, tangled prompt into shorter, cleaner sentences.
If you've crammed five different requests into one paragraph, the AI may only pick up on one or two of them.
Short, clear sentences are easier for AI to follow. Think of how you'd explain something step-by-step to a person — same principle.
The mindset shift that changes everything
The best AI users don't write a perfect prompt on the first try. They write a decent prompt, look at the answer, and improve from there.
Each iteration teaches you something — about how to phrase things, about what the AI responds to, about what you actually want.
You're not failing when the first answer is bad. You're just in the middle of the conversation.
That's not quite right. Here's specifically what's wrong: [describe the problem in one sentence] Try again, but [what to change]. Keep [what worked in the first answer], but fix [what didn't].