Automate what's stable, instead of speeding up chaos
Automation multiplies the existing process logic. Unclear decisions, exceptions, and responsibilities must therefore first be made visible.
"Only automate stable processes" is considered here from the perspective of "process, tool selection, and cost-effectiveness." For operations teams and agencies, "stable input" and "faster error detection" are particularly important.
Published: 3 min read · Author: Sebastian Geier
How can you tell if a process is ready for reliable automation?
Before automation, process variants, owners, quality criteria, and common errors are made visible. Only steps with unambiguous rules and sufficient repetition are automated; unclear decisions initially remain as deliberate manual handoffs.
Stable input
Test criterion
Stable input
The format, origin, completeness, and meaning of the required data do not change uncontrollably between instances.
Test criterion
Determinable rule
The step can be decided with verifiable conditions without imitating hidden experiential knowledge.
Observable result Success, failure, and limiting cases have verifiable states instead of just the message that the process was executed.
Observable result
Control signal
Signal 1
Percentage of automated cases that reach the defined target state without subsequent expert correction.
Control signal
Signal 2
Frequency of unknown exceptions and manual repairs per automated process step.
Faster error
Faster error An unclear standard can be incorrectly applied through automation on a large scale before anyone notices the deviation.
Exception concealment Rare but business-critical cases are treated as technical errors, even though a business decision is lacking.
Responsibility Gap The team relies on the workflow, even though no one permanently possesses the rules, input data, and results.
Diagnostic Case: "Faster Error"
A release process currently varies from team to team and contains unresolved exceptions. Instead of automating everything immediately, mandatory checks and status updates are standardized; only these stable parts are automated, and special cases remain visible to the responsible party.
Determinable rule
The actual process is monitored, including variants, waiting times, decisions, errors, and responsible roles.
Repeatable rules are stabilized, and edge cases are defined as explicit manual or alternative paths.
A small automated section runs with monitoring, quality comparison, and clear recall.
What to check before and after "Automate only stable processes"
A relevant follow-up question answered Controlling Automations with Unique IDs and Status Values"How do IDs and status values prevent duplicate or lost processing?"
A second connection for "Automate only stable processes" leads to Model Digital Processes First, Then Choose SoftwareThis article remains focused on the question "Why should the target process be defined before selecting software?"
If you want to practically implement "Automate only stable processes," you can refer to Robust Website Systems This article focuses on "Process, tool selection, and cost-effectiveness" and "Stable input."
Conclusion: Only automate stable processes
Automation reinforces existing process logic. Only a stable, observable standard makes speed an advantage instead of creating scaled chaos.
Sources and Further Information
The following sources substantiate the technical and methodological guidelines used for "Only automate stable processes."
Eliminating Toil – Google SREPrimary source for identifying repetitive manual work and the limits of meaningful automation.
The Evolution of Automation at Google – Google SREPrimary report on the benefits, limitations, costs, and careful application of automation in production systems.
Key Thesis
A process is mature when inputs, rules, exceptions, responsibilities, and expected results are stably defined. Open-ended business decisions initially remain human.
What This Is Not About
Automation does not resolve unclear responsibilities, changing exceptions, or contradictory input definitions.
What it's about
It handles repeatable steps with clear inputs, decisions, exceptions, and outputs after the process has been stabilized.
Automation & Workflow Design
Explicitly plan for error states in automations.
"Automate only stable processes" includes, as a separate check, the question: What error states should an automation system be aware of before going live?
Automation & Workflow Design
Combining Automated Content with Human Quality Control
"Automate only stable processes" is supplemented by a separate decision: How can human quality control be truly effective in content automation?
Insights Overview
All VELUNO Insights at a Glance
Further analyses on Website Systems, digital visibility, and robust working models.
Determinable rule: Path to implementation
A recurring process is first documented using several real-world examples. The most stable rule-based part forms the first limited automation step.