Why We Don't Let AI Write for Us

We build AI tools for a living. We spend our days making software that helps people understand their data faster, forecast better, and make decisions with more confidence. And yet this blog is written entirely by humans.

That’s not hypocrisy. It’s the whole point.

Writing is not a byproduct of thinking. Writing is thinking. When you sit down to explain why a product decision was made, or what you believe about some corner of your industry, you are forced to discover what you actually think. The resistance you feel when a sentence won’t come together isn’t writer’s block — it’s your brain noticing that the idea isn’t fully formed yet.

When you outsource that process to a language model, you skip the formation. You get the output without doing the work the output is supposed to represent. The text may look fine. But the thought behind it is hollow, because there was no struggle. And people can feel that hollowness when they read it — even if they can’t articulate why.

We’ve watched a lot of company blogs turn into AI press releases over the past two years. Smooth sentences. Impeccable structure. Nothing to say.

There’s a useful distinction between AI as a tool for leverage and AI as a replacement for judgment. We happily use AI to help with code, to summarize documents, to generate first drafts of boilerplate. These are tasks where the quality of the thinking isn’t the point — the output is. But when the output is the thinking, when we’re trying to communicate something real, we write it ourselves.

This blog is an attempt to do exactly that. Not polished. Not strategic. Just what we actually think, in our own words, at this particular moment.

If you read something here that doesn’t quite land, that’s also fine. Bad writing that was actually written is worth more than good writing that wasn’t.