Not everything needs a model

I don't think you should write your emails with a model.

Not because the model can't — it can, and the email will be fine. Fine is the problem. When everyone routes their thinking through the same few models, everyone arrives at the same competent, forgettable middle. The email that gets read is the one that sounds like a person had a thought. You can't outsource that and keep it.

That's the whole argument, really: AI is extraordinary where it makes a real difference, and a quiet liability where it just replaces a human doing something well.

Use it where it earns its place

Cleaning a million records. Reading a thousand support tickets to find the three that matter. Watching your data for the anomaly a person would never catch in time. That's where a model changes what's possible — work no human could do at that scale, done well. Use it there, hard. That's the whole point of building AI into the revenue engine rather than bolting it on.

Don't use it because everyone else is

Somewhere in the last two years, "we use AI" became a thing companies say instead of a thing that helps them. Teams are bolting models onto work that was already working, because not using AI started to feel like falling behind. It's cargo-cult adoption — the ritual without the result. And it has a cost: every task you hand to a model that didn't need one is a bit of judgment you've quietly stopped exercising.

Human-first is not anti-AI. It's the opposite.

The teams that win with AI aren't the ones that use the most of it. They're the ones that stayed human where it counts and used the machine where it counts — and knew the difference. A human-first culture in an AI-driven world produces better results, not worse ones. Get it backwards — machine-first, human-as-fallback — and you get speed with nothing behind it.

The question is never "can AI do this." It almost always can. The question is "should it, here." That's a human question. Keep it that way — it's the entire basis of how we use AI at all.