Iteration: the skill nobody teaches
When to push a draft further, and when to start the thread again.
Here's a pattern you've probably seen: someone tries AI, gets a mediocre first output, decides AI doesn't work, and gives up.
What they actually did was ask once, accept the first answer, and walk away. They wouldn't do that with a junior colleague, or with a search engine. The skill that separates the people who get a lot out of AI from the people who don't is iteration.
What iteration actually looks like
Iteration is the back and forth that moves a generic first draft to something you'd actually send. It is not asking the AI to do it again. It is telling it what was wrong with the first attempt, precisely enough that the second one can be better.
A useful mental model: you're working with a very fast junior assistant who reads quickly, drafts confidently, and has no idea what your business does. Most weak first outputs are not a failure of the tool. They're missing something only you knew, and nobody told you it was your job to supply it.
Vague feedback
"Try again"
"Make it better"
"Be more punchy"
"I don't like it"
Asking the same question a second time
Feedback that works
"Too long. Cut it by half, keep the pricing paragraph"
"The audience is agency owners, not consumers"
"Closer, but more like this: [paste an example]"
"You used three of the words I asked you to avoid"
"Cut the last paragraph. End on the question"
Treat the first answer as the start of the conversation, not the end of it.
What good iteration sounds like
A real one. The task: a short LinkedIn post announcing a new service.
Round one comes back with "excited to announce" and three hashtags nobody clicks. Round two: "Don't use 'excited to announce'. No hashtags. The audience is fellow agency owners, not consumers, and they're tired of LinkedIn fluff. Lead with the problem the service solves, not the service." Much better, still slightly too long. Round three: "Cut the last paragraph. End on the question. Under 100 words." Usable.
Three rounds, maybe two minutes. That is normal. The people who get the most out of AI aren't getting better first drafts. They're better at the next two.
Three moves, then start again
Nearly every fix is one of three moves, and knowing which one you need is most of the skill. What you want to avoid is the fourth option: asking the same question again and hoping for a different answer.
Three things to do this week
01
Say what was wrong, not "try again"
Name the actual problem: too long, wrong audience, missed the objection. Vague feedback gets vague fixes.
02
Add the context that was missing
Most weak first drafts are missing something only you knew. Give it that rather than asking for a rewrite.
03
Start over when three rounds get worse
Wrong format, misunderstood task, or each version worse than the last. Open a fresh prompt instead of digging.
Conclusion
01
The first output is rarely the one you ship. Two or three rounds is normal, so plan for it.
02
Iteration isn't asking again. It's telling it precisely what was wrong.
03
Sometimes restarting beats iterating. Three rounds getting worse is the signal.
Go to next