Automation does not start with the tool. It starts with understanding the work. | Image credit: Tanía Diaz / Ordinal
Repetitive work often looks like the obvious place to start with automation. Frequency alone, however, tells businesses very little about whether a process is stable, predictable or suitable for automation.
Main takeaways
- Repetition does not equal predictability.
- Exceptions can undermine automation.
- Stability matters as much as frequency.
- Variability can reduce performance.
- High volume does not guarantee suitability.
A common starting point for automation seems to be asking a team which repetitive tasks take up too much of their time, in order to identify potential automation candidates. It seems logical.
Repetitive work is visible, relatively easy to quantify and often unpopular with the people doing it. If a task happens hundreds of times a month, removing even a few minutes from each occurrence can appear to offer an obvious saving.
But repetition tells us how often something happens. It tells us far less about how predictable each occurrence is.
A frequent task can still depend on inconsistent inputs, exceptions, judgement calls or information arriving in different formats. Two cases may belong to the same process while requiring very different decisions from the person handling them.
Frequency can therefore identify work worth examining. It does not establish, by itself, that the work should be automated.
Repetition and Predictability Are Different Questions
Research into robotic process automation, or RPA, reflects this distinction.
Herm et al. (2023), drawing on 35 RPA implementation cases alongside expert interviews and workshops, found that execution frequency was an important consideration when organisations selected processes for automation. It appeared alongside other criteria:
- Process standardisation
- Process stability
- Complexity
- Frequency of exceptions
- Transaction volume
Stability Matters as Much as Frequency
The researchers also found examples of organisations improving processes before automating them.
That distinction becomes clearer with a simple example.
Repetition Can Hide Variation
Imagine a team reviewing 200 customer requests each week. All of them may sit under the same process category. But some arrive complete, while others require clarification. Certain customers may operate under different contractual conditions. Some cases may require an employee to decide whether an irregularity can be accepted or needs escalation.
The work repeats, but the decisions inside it may not. As a result, human judgement is essential.
Research into intelligent process automation reaches a similar conclusion. Mayr et al. (2024) distinguish between simple rule-based automation and more advanced AI-driven automation.
Rule-based automation works best when the same inputs lead to the same decisions and the process follows clear, stable rules. AI-driven automation can handle more complex tasks, but that does not remove the need to understand where judgement, exceptions and changing conditions enter the process.
That matters because repetitive tasks can still contain decisions that are easy to overlook when you only look at how often the task occurs.
So we can establish that judgement is one source of complexity. Variation is another.
Variability Can Undermine Reliability
Even when a task follows the same general process, changes in the information, conditions or exceptions can make automation less reliable.
Průcha (2024) examined RPA bot performance across 12 processes and found a strong relationship between process variability and bot success rates. Greater variability in processing time was associated with poorer performance, while isolated outliers had far less influence.
An occasional unusual case may be manageable. A process in which inputs, formats or conditions change regularly can require more monitoring, exception handling and maintenance.
A ten-minute task performed 500 times a month may look like an obvious saving, but less so if many of those cases still require manual intervention.
Automation can reduce manual work without eliminating the complexity surrounding it.
Repetition Is Only One Part of the Decision
This is why process selection cannot rely on volume alone.
Costa et al. (2023) treated automation selection as a multi-criteria decision, arguing that choosing unsuitable processes can contribute to failed automation attempts and weaken confidence in future projects.
High-volume, stable and rule-based activities can still be strong candidates for automation.
Frequency tells a leader mainly about scale, not the variation between cases, the judgement required along the way or how often the expected path changes.
Repetitive work is therefore a useful place to look. It is not automatically the right place to automate.
References
- Costa, D. S., Mamede, H. S., & Mira da Silva, M. M. (2023). A method for selecting processes for automation with AHP and TOPSIS. Heliyon, 9(3), e13683. https://doi.org/10.1016/j.heliyon.2023.e13683
- Herm, L.-V., Janiesch, C., Helm, A., Imgrund, F., Hofmann, A., & Winkelmann, A. (2023). A framework for implementing robotic process automation projects. Information Systems and e-Business Management, 21, 1–35. https://doi.org/10.1007/s10257-022-00553-8
- Mayr, A., Stahmann, P., Nebel, M., & Janiesch, C. (2024). Still doing it yourself? Investigating determinants for the adoption of intelligent process automation. Electronic Markets, 34, 56. https://doi.org/10.1007/s12525-024-00737-9
- Průcha, P. (2024). Towards discovering erratic behavior in robotic process automation with statistical process control. Information Systems and e-Business Management, 22, 741–758. https://doi.org/10.1007/s10257-024-00686-y