The best use of AI isn't generating ideas. It's refining yours.

I use it for code. I use it to build one pagers. I use it for graphics. At this point I use it for almost everything, and none of that is the point of this article.

Because the use that's actually changed how I work has nothing to do with output. It's talking through a half-formed idea with something that asks exactly the right question at the right time and never gets impatient waiting for me to figure out what I'm actually trying to say.

You don't need a special skill for this. You just need to stop treating it like a vending machine, ask a question, get an answer, and start treating it like a sounding board. Type the messy version. Say the thing out loud before you've cleaned it up. Let the conversation go where it goes.

Rehearse Before the Real Thing

You've got a presentation Thursday. Instead of just building the slides, run it. Give the AI the backstory, who's in the room, what you're pitching, what's riding on it, then actually deliver the presentation to it. Ask it to push back like a skeptical stakeholder would. Ask where it got confused, where you lost the thread, where a question would stop you cold if someone asked it live.

Same move works for an interview. Hand it the job description and your background, have it interview you cold, then ask what your answers were missing. You walk in Thursday having already survived the version of the conversation that used to only happen once, live, with nothing to learn from if it went badly.

Making a Tough Decision

I was torn between a safe promotion and a risky startup offer, going in circles on my own. Instead of a pro/con list, I talked it through.

That question was the whole thing. I realized I hadn't actually dug into the startup's details, I was just scared of the unknown. I took the offer. It's been great. The AI never told me what to do. It helped me hear what I already wanted.

Getting unstuck on a project

I wanted to start a newsletter but felt paralyzed by decisions: platform, topic, frequency, what if no one reads it. One question cut through all of it: "What if you just focused on sending one email to one person about something you're genuinely interested in? What would that be?" I wrote that email that afternoon, sent it to three friends, and now have 500+ subscribers six months later. The breakthrough wasn't a better plan, it was starting with one real email instead of the perfect newsletter.

Sorting Out Priorities

Three projects all felt equally urgent and I was drowning. Instead of a verdict, I got one question at a time: which would make your boss happiest in the next two weeks, what happens if the reporting system stays broken another month, how important is the dashboard really. A clear order fell out of it: the launch first, the dashboard second, the reporting fix a distant third. I'd already known the priorities. I just needed someone asking the right questions to get there.

Refining an Idea Until Everyone Gets It

This one's the quiet workhorse. You have a good idea, but it's still living in your head in shorthand, the version only you fully understand. Say it out loud, badly, and ask the AI to keep asking what you mean until it could explain it back to a stranger. Not "make this better." More like "tell me where this stops making sense to you."

That's the real value most days. Not new ideas. The same idea, pushed from "I sort of know what I mean" to specific enough that nobody in the room has to guess.