When Is a Workflow Stable Enough to Automate?
A practical test for deciding whether repeated work has earned automation—or needs observation first.
Automate a stable rule, not a recurring feeling. A workflow is ready for a small automation test when the trigger, steps, exceptions, owner, and expected result can each be written in plain language.
Five stability checks
- The same trigger appears often enough to matter.
- The path is mostly the same each time.
- Exceptions can be named rather than discovered after launch.
- A person can inspect the outcome without reconstructing the process.
- Turning it off will not damage a critical relationship or system.
If any check fails, use WAIT. Run the work manually for a bounded period and keep an exception list. That is not delay for its own sake; it is how you avoid automating a moving target.
Google’s SRE guidance describes toil as repetitive, predictable work and stresses measurement, risk assessment, and iterative development before treating automation as a cure-all. The scale is different, but the operating lesson holds: start narrow, observe the maintenance created, and retain a clear way back. Read the source.
Run this workflow through the Matrix
A 14-day test
Automate one step, not the whole system. Record the time saved, exceptions created, failed handoffs, and maintenance required. Review it in 14 days. If exceptions grow faster than the work removed, return to WAIT or DELEGATE.
For the opposite failure mode, read seven signs you automated too early.
Reader signal
Was this useful?
Decision dispatch
Keep the decision close to the constraint.
Decision notes, worked trade-offs, and corrections will open after the first useful case is published.