_{Not dating advice.}

A while back, I drafted a newsletter that felt… off. The topic was inspired by something I'd read, and when I sat back and looked at what came out, it read less like my thinking and more like a slightly reshuffled version of someone else's. It wasn't plagiarized — but it wasn't mine, either. That uneasy feeling in my gut was telling me something, and I almost ignored it because the draft was "done."

Instead, I asked my AI a simple question: How might this article cause me to lose credibility rather than gain it?

What came back was uncomfortable… and exactly right. Even though the piece read brilliantly, I had buried my own insights and leaned too heavily on the source material. So I did the thing we’re all afraid to do when someone hands us a nice, polished draft.

I threw it all out.

_{If my AI had feelings, it would have filed for emotional damages.}

I re-considered what my core unique insight was, and I restructured the article completely. I’d say it turned into one of the strongest pieces I've published.

Here's the thing most people miss about AI when they start using it to just write emails: the real gift isn't that it writes for you. It's that it writes fast for you. And that speed changes the economics of your attention. When a first draft takes five minutes instead of five hours, you stop clinging to it. You can afford to throw it away. You can finally spend your time on the part that actually matters — thinking critically about whether the work is achieving its purpose.

AI has a reputation of being a “cheerleader” that tells you you’re the smartest thing since sliced bread. I pick my primary AI tool based on how willing it is to honestly tell me I’m wrong. Over 12 months I’ve gone ChatGPT → Gemini → Claude, like a serial dater who keeps leaving partners that stop being honest after the honeymoon phase (not dating advice). Even now, I will always, always ask my AI tool to challenge me even more than it does naturally.

When it feels too easy, it’s probably not right.  (Still not dating advice)

This is the habit I want you to steal: once you have a draft of anything, ask your AI "why does this suck?" Or more precisely — why would this fail to meet its goals? Where is the risk? Where is the ambiguity? What would make someone stop reading? You're not asking it to rewrite. You're asking it to poke holes at it, so you can think better and ultimately make it stronger.

I'll sometimes run a piece through two or three different AIs, taking each one's critique and folding it back in, until the only objections coming back are ones I genuinely disagree with. That's when I know I've landed somewhere real — not when the AI says it's good, but when I can honestly defend every choice the AI says might not be good.

Next time you're writing a policy memo, a funding application, or even a difficult email, try it. Draft fast. Then ask: why would this fail? Why would someone not receive this well? Take what comes back and restructure until you feel your goals are truly being met. The draft is cheap. Your judgment is the expensive part — and now you finally have time to use it.

TransitGPT is where I share practical AI insights for transit leaders. Subscribe if this resonates or forward to a colleague who needs it. And hit reply — I read everything.

Go shatter your glass ceiling.