Automation is one of the technology initiatives with the strongest potential return — and one of those that disappoint most often. The main reason is simple: processes get automated before they have been properly defined. Automating a bad process does not improve it. It industrialises it. The mistakes happen faster, at greater scale, and with no opening for manual intervention.
The method that delivers: identify, simplify, then automate
The step organisations skip most often is the second. A process that can be automated is one whose steps are clear, stable and documented. If it is not, it has to be rationalised first. Two workshops with the business teams are usually enough to find the simplifications available.
- Identify the right candidates: a process run daily, with stable steps and little human judgement required. A monthly process riddled with exceptions does not justify the investment.
- Simplify before automating: rationalise the process, remove the steps that add nothing, write down the decision rules. This groundwork determines how robust the result will be — and it pays for itself every time.
- Choose the tool to fit the process: robotic automation for multi-system processes with no API, no-code for approval workflows, API integration for structured data exchange, the ERP's native workflow for processes inside its perimeter. The tool comes last — not first.
Stability and volume: the two criteria organisations underestimate
The most common trap is automating a process because it is visible and painful, not because it can be automated. A process with many exceptions, or with decisions that have never been written down, produces a brittle robot that breaks at the first change.
A process that changes every six months costs more to re-automate than to leave manual. Stability is the most underestimated criterion: checking it before starting is elementary caution. The same logic applies to volume — a low-volume monthly process rarely justifies the investment.
Without governance from the start, the automation estate becomes unmanageable
Organisations have deployed dozens of robots over several years with no overall view, only to end up with an estate that is undocumented, hard to maintain and impossible to audit.
Putting governance in place at launch is not administrative overhead; it is what makes scaling possible. Four questions to settle before the first deployment:
- Who decides that a process should be automated, and against what criteria?
- Who maintains the robot, and with what availability?
- How is the real gain measured, once it is live?
- What happens when the underlying process changes?
The challenge of automation is not technological: the tools are accessible and mature. It is methodological — choosing the right processes, preparing them properly, and not skipping steps.
