There's a thing that happens the first time you show someone what AI can do with their own writing. They go quiet. Not impressed-quiet. Uneasy-quiet. Like watching a stranger pick up your coffee cup and drink from it.
That reaction is data. And most people never examine it.
Here's what I want you to do. Open any AI chat tool — ChatGPT, Claude, Gemini, whatever you have already. Take a paragraph of your own work: an email you're drafting, a slide note, a section of a report, anything you're allowed to share. Paste it in. Then give it this prompt, word for word:
"Rewrite this to be 30% shorter, in the same voice. Don't change the meaning."
Wait for the result. Read it. That's the whole task.
Two minutes. Maybe less.
Now here's the one thing I want you to notice: what does your gut do when you read the AI's version?
Most people land in one of two places. The first is a small wave of relief — oh, that's better, that's cleaner — and then an immediate temptation to paste it straight in without another look. The second is a low-level indignation. A sense that the AI got it wrong, even when you can't quite say how. That it sounds a bit flat, or too smooth, or somehow less like you even though it's saying the same thing.
Both reactions are interesting. Neither is the full story.
The relief people are right that the AI often produces cleaner prose. Editing for compression is one of the things AI genuinely does well — it's patient, it has no attachment to your sentences, and it doesn't get tired at line twelve. The danger is skipping the step where you check it. The AI doesn't know what you decided not to say, or why a particular phrase mattered, or that the rhythm of that third sentence was deliberate. It's shortening to the centre of the signal. Sometimes that's fine. Sometimes you've lost the actual point.
The indignation people are onto something real, but they're often defending the wrong thing. There's a concept in behavioural science called the endowment effect — we overvalue things simply because they're ours. We do this with possessions, with opinions, and, it turns out, with sentences. Your draft feels more correct partly because your hand wrote it. That's not the same as it being better. The useful question isn't does this AI version sound like me? It's does it do the job?
This is where the two-minute exercise gets genuinely instructive. You've just learned something about where your resistance lives. That's harder to get from reading about AI than from this one small encounter with your own material.
The honest complication: sometimes the AI version is worse. Genuinely. Not because of the endowment effect, but because the original sentence was doing something precise that compression broke. I've seen it happen with legal language, with technical qualifications, with anything where the "redundant" words were carrying legal or structural weight. When that happens, ignore the output. Use the exercise anyway — you've now tested a hypothesis and found the boundary. A prompt that fails is information. It tells you what this tool is good at and what it's not.
That's the actual point of trying things this small and this fast. You're not looking for a result. You're developing a sense of where AI earns your trust and where it doesn't. That sense is not built by reading case studies or watching demos. It's built by using the tool on real problems, noticing what happens, and adjusting. Tiny loop. Repeat.
The people who get genuinely good at working with AI aren't the ones who used it most. They're the ones who paid attention most. Who noticed the unease, or the relief, and asked what it meant.
If that two minutes was interesting — if the result surprised you, or unsettled you, or made you realise you'd been avoiding this for no clear reason — that's exactly the territory The AI Habit is built for.
It's a daily training programme: 15 minutes a day, first 14 days free. The exercises are small, concrete, and built around your actual work. No theory lectures. No software to install. Just a daily habit of trying something real, noticing what happens, and getting incrementally sharper at something that's only going to matter more.
https://harrysharman.com/projects/ai-habit/