I was trying to write about a simple, few-click automation that had made one part of my day easier. AI turned it into a “powerful reminder” about “the biggest transformations.”
That was the point where I thought, no. Absolutely not.
The automation itself wasn’t some big, grand system. It was a straightforward setup that helped capture new enquiries into my CRM and made the process easier to manage. Useful, yes. A transformation? No.
But AI had taken something small and practical and made it sound inspirational:
“This is such a powerful reminder that sometimes the smallest shifts can create the biggest transformations.”
That sentence still makes me cringe because it’s just not something that would come out of my mouth. It’s too over the top. Too dramatic. “Such a powerful reminder”... like, no way.
The problem wasn’t only the wording. AI had changed the size of what happened. It had turned a small improvement into a lesson about transformation, which wasn’t what I meant at all.
At the time, I was using AI chat by chat. Every new task meant opening a new conversation, giving it a rough idea and hoping the result would be close enough.
Sometimes it was. Other times, I’d get back something polished that was technically fine but didn’t sound like me. It might exaggerate the point, fill in something I hadn’t said or take the idea further than I intended.
That’s when the editing started. I wasn’t just changing one word or deleting one dramatic sentence. I’d be rewriting whole paragraphs, pulling the tone back and trying to put myself into something AI had already shaped.
The draft was fast. Fixing it really wasn’t.
Eventually, I realised I didn’t want to keep getting better at correcting AI after it had already written the piece. I needed to catch the missing detail before it got that far.
That meant doing a little more at the beginning. Making sure I had explained what actually happened, what I thought about it, and what AI wasn’t allowed to exaggerate or fill in for itself.
So I started using a simple process to organise the idea and add what was missing before asking AI to write the draft.
And I know the obvious response to that: this sounds like more work. AI is supposed to make content quicker.
I get that. We’ve been sold the promise that AI will help us move faster without asking much from us. But there is some effort, and pretending otherwise is part of the problem.
You need to establish how you want to use AI and give it the raw words and detail that make the point yours. You also need boundaries around what it can and can’t do.
You can’t put in a rushed prompt or half-formed idea, expect AI to write for you, and then wonder why the result doesn’t sound like you.
The work is already happening. The real question is where it happens.
Without a process, most of it appears at the end. You rewrite, reprompt and sometimes start again. With a process, more of that effort happens before AI writes, while the idea can still be clarified without pulling apart a finished draft.
That is what I mean when I say move the effort upstream.
It doesn’t mean making the process needlessly complicated. It means noticing what’s missing before AI builds a full piece around the gap. The idea gets organised, you’re asked for more detail where it matters, and the approved version moves forward.
A better prompt may improve one interaction, but it doesn’t replace a process that carries your decisions and boundaries through the work.
This matters because when AI smooths out your judgement or adds meaning you never intended, the content can make you sound less considered than you really are.
That’s why I needed more than another saved prompt. I needed a way of working that kept me involved before the draft took shape, rather than handing over the first version of an idea and spending an afternoon trying to edit myself back into it.
This isn’t about removing the human work. It’s about using that work at the point where it actually helps.
And it doesn’t mean forcing every idea into the same box. The boundaries are there to stop AI from exaggerating, making things up or including something I didn’t say. My opinion, story and judgement still need to be there because I am the source of truth.
AI can support the process. It shouldn’t decide how important something was, what I believe or what lesson should be drawn from it.
That simple automation is a good example. What happened was useful but ordinary. AI made it sound bigger and shinier than it was, and the editing was me trying to close that gap after the fact.
Now, I’d rather catch the gap before the draft exists.
Using AI well still requires human input. The choice is whether that input happens before the writing, or through repeated repairs afterwards.
What’s the phrase or pattern you keep deleting from AI drafts because it doesn’t sound like you?
Keep thinking clearly,
– Shannon
