Building automation is the easy part. Knowing when to turn one off is the part people skip.
Zero is an answer
I once built a short-form video pipeline with a lead form behind it. The videos went out and the form got zero submissions.
The temptation is to park it: tweak the hook, try another week, keep it running just in case. I closed it instead. A system with no demand signal is finished, and keeping it alive costs attention every time it breaks or needs a login renewed.
Narrow beats broad
I also ran a broad morning brief that tried to summarize everything. I cancelled it. What survived is a narrow daily digest: it runs on a schedule, it writes to one place, and if it fails, there's a fallback. I read it every weekday at 7:30am.
The pattern holds for client work too. The automation that lasts does one job, on a schedule, writes to one place, and has a plan for the day it breaks. The clever one that tries to do everything is the one nobody trusts by month two.
When is it done?
My test for a routine: it's done when it has run on its schedule, unattended, three times, and I saw the result each time without going to look for it. Before that, it's a demo.
If you have automations your team quietly works around, that's worth a conversation too. Sometimes the best fix is turning one off.
