Iterate, Don't Start Over

Most people use AI wrong: they get an imperfect response, delete everything, and write a new prompt from scratch. That's the slow way.

The fast way is iteration — staying in the same conversation and refining the output with targeted follow-up prompts.

Why Iteration Works

When you stay in the same conversation, the AI retains everything it generated. You can point to specific parts and ask for changes without re-explaining the whole context. This is faster, more precise, and produces better results.

The Iteration Toolkit

These follow-up prompts work after almost any response:

Adjust length:

• "Cut this to 3 sentences"
• "Expand the second paragraph with a concrete example"
• "Make this 50% shorter without losing the key points"

Adjust tone:

• "Make this more conversational, like you're talking to a colleague"
• "Make this more formal — it's going to a board of directors"
• "Remove the jargon — explain it like I'm not technical"

Adjust structure:

• "Reformat this as a numbered list"
• "Add headers to each section"
• "Turn this into a table with three columns: Feature, Benefit, Example"

Target specific parts:

• "Keep everything except the introduction — rewrite just that paragraph"
• "The third point is wrong — replace it with [X]"
• "The conclusion is too weak — make it more decisive"

Push for more:

• "Give me 3 alternative versions of the headline"
• "What are the weaknesses of this argument?"
• "What am I missing?"

A Real Iteration Example

Initial prompt: "Write a LinkedIn post about our new product launch"

First response: Too long, too formal, reads like a press release.

Follow-up 1: "Make this 50% shorter and more conversational. Less corporate, more human."

Second response: Better length and tone, but the call to action is weak.

Follow-up 2: "The ending is too passive. Replace 'Feel free to reach out' with something more direct that drives clicks to our demo page."

Final result: A post you'd actually want to publish — reached in 3 prompts instead of starting over twice.

Try It

Write a prompt, get a response, then use at least two follow-up iterations before accepting the output. Notice how much better the final result is compared to the first draft.