Why Not All Automation Is Equal
Not all automation is equal. The best automation supports judgment and selection instead of producing more low-value output.
June 23, 2026
One of the easiest mistakes to make in the age of AI is assuming that more automation is automatically better. If a tool can help you produce faster, then it seems obvious that the best strategy is to automate as much as possible. But that misses something important. Once output becomes cheap, the real question is not whether you can automate. The real question is whether you are automating something worth scaling in the first place.
Why cheap output creates a new bottleneck
When it becomes easy to generate writing, code, images, plans, and ideas on command, production stops being the hard part. The bottleneck shifts. The hard part becomes selection. Which direction is actually promising? Which format is worth repeating? Which idea has real leverage? Most output will not matter much, even if it is polished.
Why most automation just scales slop
This is why not all automation is equal. If the thing you are automating is generic, weak, or misguided, then automation just gives you more of the same. It can make bad judgment look productive. It can make low-signal work multiply faster. That is useful only if your goal is volume. It is not useful if your goal is disproportionate outcomes.
Why the power law makes selection more important
The reason this matters is the power law. In most domains, a small number of ideas, experiments, or directions create most of the value. A small number of posts drive most of the attention. A small number of product bets drive most of the growth. A small number of insights unlock most of the progress. When output is cheap, this becomes even clearer. The challenge is no longer making more things. The challenge is identifying the few things worth making more of.
What to automate only after you find signal
A better model is to run many small experiments, look for signal, and then automate the part that proves itself. Social media creators already do this. They test formats, notice when one breaks out, and then create variations of the thing that is working. The same logic applies everywhere else. First find the direction with real traction. Then automate around that. Automation is powerful, but only when it is attached to something that deserves to scale.