There’s a real fear underneath the automation conversation: that if you automate the making, you hollow out the craft, and eventually there’s nothing left that’s actually you. It’s a reasonable worry, and the answer to it isn’t reassurance — it’s a line. Some things should be automated. Some never should. Knowing which is which is the entire skill.
What automation actually removes
Automation is good at exactly one thing: doing the same decision the same way, repeatedly, without getting tired or bored. That’s it. And it turns out a huge amount of creative work is that — resizing exports, renaming files, applying the same grade, laying out the same structure for the hundredth time. None of that repetition was ever the craft. It was the tax you paid to get to the craft.
What automation can’t remove — because it can’t do it in the first place — is the curiosity. The “what if we tried it the other way.” The noticing that something’s slightly off. The decision about what the thing should even be. Those aren’t repetitive, which is exactly why no machine reaches them.
| Safe to automate | Never automate |
|---|---|
| Repetitive, identical every time | Different every time by nature |
| The answer is already decided | The answer is the whole question |
| Boring when done by hand | Interesting when done by hand |
| Exporting, resizing, renaming, formatting | Deciding what it should be, and whether it’s any good |
| Frees your attention | Is your attention |
The trap: automating the curiosity by accident
The danger isn’t automation itself — it’s automating one square too far. The moment you let the machine decide what to make, not just how to churn it out, you’ve handed over the part that was yours. This is subtle, because it often feels like efficiency. “Just have it generate the concepts too” sounds like the same kind of time-saving as “have it resize the exports.” It isn’t. One removes tax; the other removes the point.
The test is the left-hand column above: automate the thing whose answer is already decided. Protect the thing where the answer is the work.
Curiosity is the thing you’re freeing up for
Done right, automation isn’t a subtraction from craft — it’s a reallocation toward it. Every hour you don’t spend renaming files is an hour of attention you get back for the part only you can do. That’s the whole promise: not less work, but the interesting fraction of the work taking up a bigger share of your day. (This is the practical cousin of a point I make in The Thing That Makes the Thing — building the system that handles the repetition so you can think about better questions.)
How to Proceed
- List the tasks in your week that are identical every time. Those are your automation candidates — the tax, not the craft.
- For each, confirm the answer is genuinely always the same. If it varies with judgment, it's not repetition — leave it alone.
- Draw your line explicitly: name one thing you will never automate because deciding it *is* the work.
- After automating something, notice where the freed hour goes. If it's not going to the curious part, you automated the wrong thing.
- Watch for scope creep toward "let it decide too." The moment automation starts choosing what to make, pull it back.