3 Questions I Ask Before Automating Anything.
By Hazen Gardner · 21 October 2026 · 5 min watch · 2 min read
Almost anything can be automated. That was never the right question.
Before I build anything, I run it through three filters. Does it fail silently if something goes wrong, does it happen often enough to justify the upkeep, and is there an actual clean moment to build around.
Most tasks fail at least one of those, and that's fine, they're better off staying manual. The ones that pass all three are usually the only ones worth the build time.
1. Does it fail silently?
Some failures are loud. A supplier doesn't deliver, and you're on the phone with them within the hour. Other failures are quiet. A subscription renewal slips past its date, or an automated email lands in a spam folder and never gets opened. Nobody checks for those by default, because nothing forces the check. Those are the processes worth the most attention, because the risk isn't the task itself, it's the silence around it.
2. Does it happen often enough to matter?
Renewing a business license once a year isn't worth a full system, no matter how annoying the paperwork is. Something that comes up dozens of times a week is a different story. The maintenance cost of an automation doesn't disappear just because the task is rare, it still needs monitoring, and it still needs updating when something upstream changes. Frequency is what turns a one off annoyance into something worth the ongoing overhead of a real system.
3. Is there a clean trigger?
Automation runs on triggers. If the trigger is genuinely fuzzy, if it needs a person's judgment to decide whether something's actually happened, that's usually a sign the task isn't ready to be automated yet, or it needs breaking into a smaller piece that does have a clean trigger. A payment either goes through or it doesn't, that's a clear yes or no with nothing to interpret. Deciding whether a client relationship feels off is not clean, that one still needs a person.
Run your own processes through it.
These three won't catch everything, but they rule out most of the wrong candidates before any build time gets wasted on something that shouldn't exist. Worth running a few of your own processes through this before automating anything. Which ones would actually pass all three.