Question I Stopped Asking AI—and the One That Replaced It

I used to ask AI the same question everyone starts with: “Can you do this for me?” It worked… but the results were hit‑or‑miss. Some answers were great, others felt generic, and I always ended up tweaking more than I expected.

The breakthrough came when I swapped that question for something far more powerful: “What do you need from me to do this well?” Suddenly the output improved across every task. Instead of guessing my intent, Copilot started telling me what context, constraints, or examples it needed to match the way I think. I stopped treating AI like a vending machine and started treating it like a collaborator — one that performs best when I give it the right fuel.

The insight: Better inputs always lead to sharper outputs, and the fastest way to get them is to ask the AI what it needs before telling it what to do.

• Start with the need‑finding question. Ask, “What would you need to create the best version of this?”
• Give context in small doses. The AI will tell you when it needs more — don’t assume.
• Share two examples you like. Examples anchor the AI’s tone and structure better than long instructions.
• Let it propose a plan first. Reviewing a plan takes seconds; correcting a bad output takes minutes.
• Iterate with reasoning, not rewrites. Explaining why you’re making a change helps the AI self‑correct next time.

When I stopped asking AI to “just do it” and started asking what it needed, everything got better — and faster.