I've asked Claude why users drop off, why a flow should work, why a line of copy should land, more times this week than I'd like to admit. Most of those times, I couldn't have answered it myself.
My manager found that out before I did. He asked why this, and not that, about an activation flow I'd built entirely on Claude. I had nothing. Not a bad answer. No answer. I didn't know what my user was actually feeling at that screen, I'd just shipped a decision I never made myself.
It wasn't really an AI problem. I'd stopped asking why a button sits where it sits, why a sentence is worded the way it's worded, why someone drops off exactly there and not a step earlier.
So every week, I pick one real thing apart, an app's feature, a line of copy that shouldn't have shipped, a signup flow losing people for reasons nobody bothered to name, and sit with the why until it's actually mine. AI still helps me move faster, I just try not to let it think for me anymore.
This isn't about using AI less. It's about understanding your product, your user, and your own copy a little more before you reach for it.
I might be wrong some weeks. Your thinking, or someone else's, might land better than mine. That's fine, that's how I get there. I'm still figuring this out myself, slowly. If any of this sounds familiar, I'd like to have you here.
