Where to Start When Everything Looks Automatable
Brendon Smith
Most teams arrive at automation with a list. Twenty processes, all of them annoying, all of them apparently automatable. The list is the problem — not because it is wrong, but because it flattens twenty very different jobs into one undifferentiated pile.
Start where the work is boring and identical
The best first candidate is rarely the most painful process. It is the most repetitive one. Pain usually means complexity, judgement and exceptions — exactly the things that make an automation brittle. Repetition means the rules are already stable, even if nobody has written them down.
A useful test: could a new hire do this task correctly on their second day, given a checklist? If yes, it is a good candidate. If it takes six months of context to get right, leave it alone for now.
Map it before you build it
Before any tooling decision, watch the process run once end to end and write down every input, every decision point and every place a human has to check something. Most teams discover the process they described is not the process that actually happens. That gap is where automations quietly break.
Once one workflow has run unattended for a month, the second is dramatically easier — you will have the integrations, the error handling and, more importantly, the organisational trust.