The most dangerous manual steps are usually not the dramatic ones. They are the normal ones.
They are the steps everyone knows, nobody likes, and nobody quite gets around to removing because they still seem manageable.
That is exactly why they become part of the culture.
Manual steps become invisible once they become routine
Small teams often tolerate manual steps for too long because each one feels small on its own.
But once a step becomes part of a normal deploy, a normal incident, or a normal environment setup, it starts costing the team every week.
The real cost shows up in:
- slower releases
- more fragile handovers
- higher incident stress
- hidden knowledge
- inconsistent outcomes between engineers
That is why I think small teams should remove recurring manual operational work early, before habit makes it feel acceptable.
The first manual steps I want gone
The usual candidates are familiar:
- deployment commands only one person remembers
- console changes required to finish a release
- secrets copied by hand between environments
- incident checks that depend on tribal memory
- rollback steps that have never been written down or tested
None of these are impressive problems. That is exactly why they are dangerous.
They quietly train the team to rely on memory instead of system clarity.
Kill the manual steps that touch production most often
I usually prioritize manual work in this order:
- anything on the path to production
- anything on the path to rollback
- anything required during incidents
- anything repeated for environment setup or configuration
That order matters because it focuses on the steps that multiply risk fastest.
Automation is not the only answer
Sometimes the right fix is automation.
Sometimes the right fix is removing complexity that created the manual step in the first place.
That is an important distinction. A bad workflow should not automatically become a script. Some manual steps exist only because the design around them is already too awkward.
My default advice on manual operational steps
If a manual step shows up in a repeated workflow, assume it is a candidate for removal.
If it touches production, assume it matters more than it looks.
And if only one person can explain it clearly, treat it as operational debt immediately.
Small teams do not get stronger by becoming good at repeated manual work. They get stronger by making that work unnecessary.