Automation is attractive because it is visible. A manual step disappears. A workflow moves faster. A dashboard updates in real time. A system sends the reminder automatically instead of relying on memory. All of that can look like progress.
But automation does not automatically improve a process. In some cases, it simply makes a bad process run faster. McKinsey puts this bluntly: automating a low-productivity line can simply speed up poor quality. That is a useful warning for any organisation thinking about digital transformation, operational excellence, or process improvement.
So how do you improve a process without automating waste? You start by improving the process itself.
Start with the process, not the tool
One of the most common mistakes in improvement programmes is starting with the technology question.
What system should we buy?
What can we automate?
What dashboard do we need?
What workflow can be digitised?
Those questions are not wrong, but they are often premature. BCG’s research on large-scale tech programmes shows that more than two-thirds are not expected to be delivered on time, on budget, or within scope, with poor management of interdependencies a major reason. That matters because when organisations focus on the technology before understanding the process, they tend to underestimate complexity and overestimate what the system will fix.
A better starting point is simpler:
What is the process trying to achieve?
Where is the actual waste?
What is slowing the flow?
What decisions are unclear?
What do people do outside the formal process just to get the work done?
That is where real improvement begins.
Separate activity from value
A process can be full of activity and still add very little value.
Approvals may be happening, reports may be generated, forms may be completed, data may be entered, and meetings may be held. But none of that proves the process is good. Often, process waste hides inside respectable-looking work: duplicated checks, repeated data entry, unnecessary handoffs, unclear ownership, and controls that create delay without improving quality.
This is why improvement should begin by separating value-adding work from non-value-adding work. Before automating any step, ask:
Does this step help produce the outcome?
Does it reduce real risk?
Would anyone miss it if it disappeared?
If the answer is no, that step should probably be removed or redesigned before it is digitised.
Simplify before you standardise
Many organisations try to standardise a process that is still too complicated. That creates another form of waste: formalised complexity.
A complicated process can look more controlled once it is documented and standardised, but if it still contains unnecessary decisions, duplicate approvals, confusing routes, or too many exceptions, the waste remains. It just becomes harder to challenge because it now looks official.
Improvement works better when simplification comes first. Remove unnecessary steps. Clarify decision points. Reduce avoidable variation. Only then should you think about locking the process into standard work or software.
This matters even more in digital environments, where systems can hard-code poor logic very effectively.
Understand where the real failure sits
A process often looks broken at the point where the problem becomes visible, but that is not always where the cause sits.
A late approval may actually be a role-clarity problem. A recurring documentation error may really be a usability problem. A quality issue may stem from an upstream handoff. A repeated escalation may reflect weak decision rights rather than weak effort.
McKinsey’s more recent operational-excellence thinking argues that high performers go beyond classic tool use and rethink how performance, people, and technology fit together. That broader view matters because waste is often systemic rather than local.
So before automating anything, trace the problem properly. Ask:
Where does the failure first become likely?
What condition makes this delay or error repeatable?
Which upstream issue is creating downstream rework?
If you automate at the symptom level, you usually preserve the cause.
Design around real work, not ideal work
Another common mistake is improving the documented process instead of the real process.
The documented process may look clean, logical, and compliant. But the real process often includes workarounds, side conversations, local spreadsheets, informal approvals, and hidden judgement calls. These usually exist for a reason. They often compensate for gaps in the formal process.
If improvement ignores that reality, automation tends to fail in a predictable way: the official process goes live, but people continue using unofficial workarounds because the new process does not fit how work actually happens.
That is why good process improvement requires observation, not just documentation review. Go to where the work happens. Speak to the people doing it. Understand where they hesitate, where they wait, where they re-enter data, and where they have to compensate for the system.
You do not want to automate the workaround. But you do need to understand why it exists.
Improve decisions, not just steps
A lot of process waste is really decision waste.
The process may not be slow because there are too many actions. It may be slow because nobody is sure who decides, what information is enough, when escalation is needed, or what rule should apply. In those cases, automating steps without improving decision logic simply moves confusion into a digital environment.
So part of improving a process is asking:
Who owns this decision?
What information is actually needed?
What can be decided at first point of contact?
What should trigger escalation?
What can be resolved without another layer of review?
When decision logic becomes clearer, processes usually become shorter, faster, and easier to automate well.
Build for people, not just efficiency
The European Commission’s Industry 5.0 framework is useful here because it shifts attention toward human-centricity, sustainability, and resilience, not just efficiency. That means a process is not truly improved if it becomes faster but harder for people to use, less understandable, or more brittle under pressure.
This is a crucial point. A process that looks efficient on paper can still create waste if users find it confusing, if the interface is awkward, or if the workflow ignores the realities of work on the ground.
So before automating, ask:
Will this make work clearer or more confusing?
Will users trust it?
Will it reduce cognitive load or add to it?
What happens when something goes wrong?
A process that people can use well is usually a better candidate for automation than one that is merely technically elegant.
Measure improvement before digitalisation
Another useful discipline is to improve the process manually first, where possible.
If a simplified version of the process cannot perform better before automation, it is risky to assume software will fix it later. Piloting a cleaner manual process often reveals what really matters: which steps are essential, which rules are unclear, which data points are unnecessary, and where the real bottlenecks remain.
This also gives you a baseline. You can see whether lead time, error rate, rework, or user effort improves before you invest in technology. That makes later automation much more intelligent because you are automating a better design, not just an older habit.
Use automation last, not first
Automation still has a major role. It is valuable when it removes repetition, reduces administrative burden, strengthens control, improves visibility, or speeds up stable and well-designed workflows.
But automation should come after:
- waste has been identified
- unnecessary steps have been removed
- roles and decisions are clearer
- the real process is understood
- usability has been considered
- the process performs better in principle
BCG’s 2026 work on AI impact argues that transformation fails not because of technology itself, but because leaders apply weak transformation discipline. That idea applies directly here: poor process design plus fast technology is not transformation. It is accelerated waste.
What good looks like
A well-improved process usually has a few clear characteristics before automation is applied.
It is shorter.
It has fewer handoffs.
It contains fewer duplicate checks.
Decision rights are clearer.
People understand the flow.
Exceptions are visible.
The work makes sense without requiring constant workaround behaviour.
Only then does automation truly add value. At that point, digital tools are reinforcing a better process, not rescuing a weak one.
Conclusion
To improve a process without automating waste, start with the work, not the software. Simplify first. Understand the real process. Trace the cause of the problem properly. Improve decision logic. Design around users. Test the cleaner process before digitising it.
Automation is powerful, but it is not a substitute for process thinking. As McKinsey warns, automation can simply speed up poor quality, and as BCG’s transformation research shows, weak discipline and unmanaged complexity still derail large programmes. The organisations that improve most effectively are the ones that remove waste before they digitise it.