Which Business Processes Should You Not Automate?
Automation is useful.
But that does not mean every manual task should disappear.
Sometimes the best decision is:
Leave it alone.
A process might be too rare.
Too unpredictable.
Too dependent on judgment.
Or simply cheaper to perform manually than to automate properly.
Good automation strategy is not about automating the most things.
It is about automating the right things.
Do Not Automate a Process You Do Not Understand
If nobody can clearly explain how a process works, automation is probably premature.
You might hear:
“Normally we do this, unless it's this type of customer, and then sometimes Sarah checks it first.”
That is a warning.
Before automating it, establish:
- what triggers the process,
- what information it needs,
- what decisions are made,
- who owns each step,
- and what the normal outcome should be.
Otherwise you may simply encode confusion into software.
Avoid Processes That Change Every Week
Automation works best when the underlying rules are reasonably stable.
If management changes the process constantly, every automation change creates more development and testing.
For example:
If your quoting approval rules have changed six times in two months, it may be better to stabilise the process first.
Automation should support the operating model.
It should not constantly chase it.
Be Careful With High-Judgment Decisions
Some work requires context that is difficult to reduce to a rule.
For example:
Should we give this customer a credit?
Is this complaint serious enough for management involvement?
Should we accept this unusual project?
Automation can provide information.
It can prepare the decision.
It can route the task.
But the final decision may still belong with a person.
That is perfectly fine.
Do Not Automate Rare Tasks Just Because You Can
Suppose someone spends:
15 minutes
on a task that happens:
twice per year.
You could automate it.
But should you?
Probably not if designing, building, testing and maintaining the automation takes ten hours.
Frequency matters.
A small inefficiency repeated 5,000 times can be valuable to automate.
A larger inefficiency that happens once a year may not be.
Avoid Automating Broken Exceptions Before the Main Process
Businesses sometimes start with the strangest edge case.
For example:
“Can we automate what happens when a customer changes the product after the supplier has shipped it but before the installer is assigned?”
Maybe.
But first ask whether the normal workflow is automated.
Focus on the 80–90% path before building complicated logic for rare exceptions.
Keep unusual situations visible for humans until they happen often enough to justify automation.
Do Not Automate Poor Data
Automation trusts the information it receives.
If the source data is unreliable, the workflow can make mistakes faster.
For example:
Wrong customer address
↓
Automatically copied to job
↓
Automatically sent to technician
↓
Automatically included in reminder
One bad field has now travelled through the entire business.
Fix the data source first.
Be Careful With Irreversible Actions
Some actions deserve a human checkpoint.
For example:
- sending large payments,
- deleting records,
- issuing credits,
- cancelling important jobs,
- changing contract values,
- or sending sensitive customer communication.
You can still automate the preparation.
For example:
Job qualifies for credit
↓
Draft credit created
↓
Finance approves
↓
Credit issued
Automation removes repetitive work without removing control.
Do Not Automate Something That Already Takes Seconds
Sometimes automation enthusiasts create a five-step workflow to save somebody:
30 seconds per month.
Technically impressive.
Commercially pointless.
Always compare:
Time saved
against:
build cost + maintenance + failure risk.
The simplest solution is often the best one.
Automate the Repetition, Keep the Judgment
This is usually a useful dividing line.
Automation is strong at:
- moving data,
- checking conditions,
- creating records,
- sending routine notifications,
- calculating values,
- monitoring deadlines,
- and surfacing exceptions.
People are strong at:
- negotiation,
- judgment,
- unusual customer situations,
- conflict,
- strategy,
- and decisions where context matters.
A good workflow combines both.
Start by Asking One Question
Before automating a process, ask:
What problem are we actually solving?
If the answer is:
“Because automation would be cool.”
stop.
If the answer is:
“This costs us 20 hours every week and creates regular mistakes.”
you probably have a much stronger candidate.
At 5M Consulting, we help businesses decide not only what should be automated, but what should stay manual.
The goal is not maximum automation.
It is a better operating system for the business.