AI Summer School #2: Naked AI

"No self-respecting professional would use AI to write."

I hear this constantly (often on Radio 4), usually said with complete confidence/disdain, and I get it. LinkedIn is awash with AI slop, and em-dashes, and AI-isms spottable at 50 paces.

But most people's entire experience of AI writing is what we call naked AI: default settings, unconvincing rhythm, popcorn phrasing. And just because it knows your name doesn't mean it knows what you want.

Judging AI on its default settings is like judging a new assistant you've never briefed a word to. Nobody expects a person to be any good without being told what you want and how you like things done, yet we open an AI tool, type "write me an email", and conclude the technology can't write. If your team has written off AI writing, it might have missed this crucial step.

I've spent thirty years in comms. I have my own style and I wouldn't dream of letting anything vaguely slop-like leave my desk. So I have spent time briefing my AI properly: who I am, what I do, how I write, and a list of rules - and I keep it updated. Things like use plain English, no flattery, keep the connective tissue that makes my writing read like my writing. I've even banned it from using the word "lands". Take 10 minutes to give it every petty rule and AI annoyance you've spotted, and give it examples.

With that set up, what comes back sounds just like me, usually a really excellent first draft. It's helping me, not replacing me: it's still my thinking, my judgement, and nothing goes out without me all over it.

And remember, good AI usage can be indistinguishable from writing from scratch. Don't rely on being able to spot it.