Keep 20% of Your Work Human

2 min read

The more we let AI do things for us, the more it's going to cost us later. Not in tokens. In something much harder to buy back.

Let me explain what I mean.

The copywriter

Say your job is copywriting. Now there's AI that writes, so you decide to let it write and you just "polish" at the end. It'll probably work at first. You're faster, the client is happy, everyone is happy.

But when everyone does the same thing, and you've already lost the ability to write like a normal human being, what exactly is special about your product? Your product was your writing. If the writing now comes from the same place everyone else's comes from, you're selling the same thing as everyone else, only with your name on the invoice.

The designer

Or take designers. You have an amazing eye. You know how to make wonderful things. Then you let AI do the designing and it really accelerates everything. It's incredible. Output went up five times.

But a few years from now you don't know how to design anymore, because the AI designs. And so does every other designer. You have no edge left. Your output becomes generic, and the whole moat, the thing that was uniquely yours, goes in the bin.

That's the part people miss. The cost doesn't show up this quarter. This quarter looks great. It shows up later, slowly, as a skill you didn't notice you were losing.

What I suggest

At least 20% of the time, do the work by hand. Completely. Yourself. No first draft from the model, no "just tidy this up for me". Keep the uniqueness and keep the capability. Like everything else: if you don't train it, you lose it.

It doesn't have to be the most important 20%. It has to be real work, done end to end, often enough that your hands still remember how.

Why I write my LinkedIn posts without AI

This is also why I write my LinkedIn posts without any AI pass. Not before, not after, not during, not to get ideas. I want to keep my writing ability, and to get better at it.

And also, like everyone, I hate slop. Just as I don't want to read slop, I assume nobody else wants to read it either. Same as getting yet another generic Claude design. You can spot it from across the room, and the moment you spot it you stop paying attention.

It's hard. It hurts efficiency. But I believe that in the long run it will pay off very, very well.

And honestly? It's also fun.

P.S This post was adapted from my human-written Linkedin post and into a blog post by.... AI


All posts

Working on something this applies to?

A free 30-minute intro call. You'll leave with a clear read of your product challenges, ideas how to solve them, and a sense of whether we want to work together.