7 min read

Iteration: First Draft, Then Refine

Treat the first prompt as the draft, not the deliverable

Most people write a prompt, get a result, and either accept it or scrap it and start over. Both are wrong moves.

AI conversations are conversations. The fastest way to a great answer is rarely a great first prompt — it's a decent first prompt plus three small adjustments. Knowing how to iterate well is the difference between getting 70% of the way there and getting all the way.

Three iteration patterns

Use the right one for the situation:

Refine — "Good, but make it shorter / tighter / more direct." Use when the output is on-track but the form is off.

Focus — "Good, but expand on point 3 / drop the section about X / lead with the recommendation." Use when AI got the substance right but distributed attention wrong.

Reset — "Forget that. Let's start over with a different angle." Use when the output is fundamentally wrong — AI misunderstood the task or you misframed it. Don't try to refine a wrong direction into a right one.

The one-iteration test: if a single follow-up message can get you what you want, refine or focus. If you need three or more, reset and rewrite your prompt with what you learned from the bad output. Most second prompts are far better than the first because you've seen what doesn't work.

Iteration is also feedback to yourself

When AI's output is wrong, the most useful thing you can do is ask yourself: what was missing from my prompt that would have made this right on the first try?

Usually it's one of the four RICE elements — most often Expectations (what format / length / tone you wanted) or Context (a specific constraint you assumed AI would know).

Over time, this turns iteration from a fix-the-bad-output activity into a write-better-prompts-from-the-start skill.

Quick Check

You asked AI to summarize a long report and the output is structured well but reads as too corporate and dry for your audience. What's the right iteration move?