Why Your Best Prompts Still Aren't Working

I was on a coaching call recently watching a member ask Claude to fix a draft.

First try, the draft was off.

Second try, a little better but still kinda generic.

Third try, she added more direction. Better, but the voice was still wrong.

By the fifth try, she was getting frustrated. By the seventh, she was ready to give up on AI entirely.

I stopped her there.

Because here’s something I’ve noticed on these calls:

If the first three or four iterations don’t get you close, the problem isn’t the iteration.

It’s the context you gave Claude at the start.

More reps won’t fix a missing voice doc. Or a fuzzy goal. Or no examples of what “good” looks like to you.

It’s like asking a new employee to redo a project five times when you never explained what you wanted in the first place. The work doesn’t get better. Both of you just get tired.

So here’s the shortcut.

When you hit iteration four and you’re still nowhere close, stop.

Start a new chat. Re-brief from the top. Give it the about-you doc, the voice samples, the actual outcome you’re after.

Usually that next chat gets it on the first try.

The fix is almost never downstream. It’s upstream.

Matt “reset rules” Ragland

P.S. I’m starting the next Claude for Creators cohort late this summer. Click here to learn more and grab your spot. I limit the group to 10 people so everyone gets attention and support.