What a year of using AI badly taught me about communication
For the better part of a year, I used AI like a toy.
I asked it lazy questions. I watched it hand back mush. And I did what a lot of smart, skeptical people did around that time — I decided the whole thing was overhyped and went back to work.
Then one afternoon it occurred to me that the mush was mine.
The machine had been reflecting back, with uncomfortable precision, exactly how clearly I’d asked. Sloppy thinking in, sloppy answer out. When I finally got specific — when I said what I actually meant instead of what I half-meant — the thing got sharp. Genuinely sharp.
That was the moment the novelty wore off and something more useful took its place. Because what I’d run into wasn’t a technology problem. It was a communication problem. And it belonged to me.
The shortcut that isn’t
There’s an industry being built right now on the opposite premise.
You’ve seen the ads. A $27 PDF that promises AI will finally let you make the money you could never make without it. A quiz that “reveals” your hidden path. A download you can supposedly turn around and sell yourself — as if the file were the hard part.
Sit with that first promise for a second. If you didn’t know how to build something valuable before, a chatbot is not going to hand it to you. The tool amplifies what you already bring. It does not manufacture competence you don’t have. Selling people the fantasy that it does isn’t innovation. It’s the oldest grift in a new outfit.
The skill nobody’s selling
Here’s what AI actually rewards — the thing that doesn’t fit on a sales page:
The ability to say clearly what you mean.
That’s it. That’s the whole skill. Not prompt tricks. Not secret templates. The plain discipline of getting clear — first with yourself, then on the page — about what you’re actually trying to do. The machine just happens to be a brutally honest mirror for how well you’ve done it. It can’t read your tone. It can’t fill the gaps with goodwill the way a patient friend does. So when your request is vague, the vagueness comes right back at you, undisguised.
Most of us have never had to face that before. It’s bracing.
What twenty years in data taught me
I spent two decades working in data before any of this. And data work teaches you one law that turns out to explain almost everything:
If two people don’t agree on what a word means, every system built on it eventually breaks.
There’s a whole discipline devoted to fixing this. It’s called governance, and it is deeply unglamorous — mostly the patient work of defining terms, agreeing on what things mean, and refusing to let ambiguity slide. Skip it and your data is garbage no matter how powerful your tools are. Garbage in, garbage out — not because the system is broken, but because the meaning was never shared to begin with.
The same law governs your conversations with AI. And if you’re honest, it governs most of your conversations with people, too.
Same with a marriage. Same with your team on a Monday morning. When something “doesn’t get it,” the problem usually isn’t the other party. It’s that the meaning was never actually shared. We just don’t notice it with people, because feelings paper over the gaps. The machine offers no such mercy — which is exactly what makes it such a useful teacher.
Where I’ve landed
So here’s the conclusion I keep arriving at, and where I’m going to spend my writing for a while.
The most valuable skill in the age of AI is not technical. It’s the oldest one there is: the ability to mean something clearly, and to share that meaning so completely that another mind — human or machine — can act on it.
That can’t be sold for $27, because it isn’t a shortcut. It’s a practice. It’s slow at first and then it compounds. And it quietly makes you better at everything downstream — your work, your writing, your relationships, and yes, the tools.
I used AI badly for a year before I understood that. I’d rather you skip the year.
David Willis 