I've sent exactly one email while angry that I wish I could take back. It went to a supplier who'd shorted us on materials twice in a month, and I wrote it hot, at 6am, before coffee. I hit send before I let it sit. It didn't end our business relationship, but it made the next six months more awkward than they needed to be.
These days, if I write anything with an edge to it — an email to a contractor, a comment on a neighborhood Facebook group, a text to my brother about money — I paste it into an AI chatbot first and ask, plainly: "Does this sound too harsh? How might this land with someone who's already a little defensive?" It won't stop me from sending something firm. It just shows me the version of my words that a stranger, without my tone of voice or my face, would actually read.
That gap is bigger than people realize. You know exactly what you mean when you write "per my last message," because you can hear your own voice saying it. The person reading it just sees three cold words with no context. An AI reading it fresh, the same way your recipient will, is a decent stand-in for that blind spot.
This is especially useful before posting anything public — a review, a comment, a post defending yourself in some online argument you probably shouldn't be in anyway. Ask it to read it back to you as a stranger would. If it flags something, that's not censorship, that's a free preview of how your words are about to be received by someone who doesn't already like you.
You're still the one who decides what to send. This just buys you thirty seconds of distance from your own heat, which is usually all it takes to catch the sentence you'd regret.
If you want the slow, no-jargon version of all this — the stuff I wish someone had handed me before I fumbled through it myself — that's exactly what I built On Ramp AI for. Check it out here.