Why Working With AI Is Like Working With Kids

I've got two kids, and if there's one thing parenting teaches you fast, it's that "just do the thing" almost never works the way you meant it to. You have to actually think about how a request lands on the other end.

Turns out, the same thing is true with AI. Not because it's dumb — it's not — but because it behaves a lot more like a very capable kid than like a computer. And once I started treating it that way, I got noticeably better results. Here's what I mean.

It doesn't know what you know — unless you say so

If I tell my kid "grab your shoes, we're going to Nan's," they know which shoes, they know who Nan is, they know roughly what that trip involves, because they've got years of context I didn't have to repeat. AI has none of that. It doesn't know your business, your project, your history with the thing you're asking about — unless you put it in the message.

So what does that mean for you? Don't assume the AI is picking up on the backstory. If you're three months into a project and ask "does this look right?", it has no idea what "right" means for you unless you've told it, or it's still sitting in the same conversation where you explained it earlier. Give it the context you'd give a new hire on day one, not a long-term colleague.

It takes you extremely literally

Tell a kid to "clean your room" and there's a real chance everything just gets shoved under the bed. Technically the floor is clean. That's not defiance — it's that "clean" wasn't actually specific, and they filled the gap with the easiest interpretation available.

AI does the exact same thing. Ask for "a short summary" and you might get three sentences or you might get a page, because "short" was never actually defined.

So what does that mean for you? Replace vague words with numbers and examples wherever you can. Not "make it shorter" — "cut it to under 100 words." Not "make it more professional" — "remove the exclamation marks and contractions." The more specific the ask, the less room there is for it to guess wrong.

It'll tell you what sounds right, not always what's true

Kids do this constantly — not out of malice, but because "I don't know" feels worse than an answer, even a made-up one. Ask a five-year-old why the sky is blue and you'll get a confident, completely wrong theory before you'll get "I'm not sure."

AI has the same instinct. It's built to produce a fluent, confident-sounding answer, and fluency isn't the same thing as accuracy. If you don't push back, you won't always know which one you got.

So what does that mean for you? Ask "are you sure?" more than feels necessary. Ask it to explain why it landed on that answer, not just what the answer is. If something feels off, it usually is — trust that instinct and make it show its work.

It learns faster from an example than an explanation

Try explaining how to tie a shoelace to a five-year-old using only words and you'll watch their eyes glaze over somewhere around "make a loop." Show them once, slowly, with your actual hands, and it clicks in a way no amount of verbal instruction ever quite manages. Kids copy what they see far more reliably than they follow what they're told.

AI's the same. Describe the tone or format you want in the abstract — "make it punchy," "keep it professional" — and you'll get its best guess at what those words mean, which may or may not match what's in your head. Show it one example of the actual thing you want, though, and it'll match that far more closely than any amount of describing ever gets you.

So what does that mean for you? When you can, show instead of tell. Paste in a paragraph you've written before and say "more like this." Give it one good example of the output you're after rather than three sentences trying to describe it. It's the difference between telling a kid "just tie it" and actually showing them the loop — one of those actually works.

None of this makes AI less useful — if anything, it's the opposite. Once you stop expecting it to read your mind and start treating it the way you'd coach someone new, the results get noticeably better, fast. Turns out the same patience that works on a five-year-old works pretty well here too. Who knew.