Process automation
Which business process should you automate first? A practical SME guide
Start with a repetitive, stable and bounded process with accessible data and rules your team can explain. Measure its current cost and define how exceptions will be handled before automating it.
Copying enquiries from an email into a spreadsheet may look like a small task. Repeated over several weeks, it competes with other work and requires checks to avoid losing information. Understanding the actual workload helps you decide whether automation is worthwhile.
What the research tells us
Syed and colleagues (2020) reviewed 125 publications on robotic process automation, or RPA. They identified clear rules, volume and process maturity among task-selection criteria. The review includes academic and professional literature; it does not establish a universal saving. RPA automates interactions with applications and does not represent every approach to systems integration.
Reference: Syed et al., 2020
1. Describe the task from start to finish
Choose a recognisable unit of work: registering an enquiry, preparing a notification or transferring a validated record. Describe the trigger, where the information appears and the required output. Include checks, which can easily disappear from an optimistic time estimate.
Observe enough cases to capture ordinary work and less frequent exceptions. Ask the people doing the task how problems are actually resolved.
- Input: source, format and required fields.
- Output: the record, document or notification required.
- Owner: who confirms that the result is correct.
- Exceptions: missing data, duplicates or unavailable systems.
2. Compare candidates using the same worksheet
This worksheet is a practical tool developed for this guide, not a scientifically validated scoring scale. Use it to expose uncertainties and compare options.
| Criterion | Record | Check before proceeding |
|---|---|---|
| Volume | Runs per week or month | Whether frequency justifies implementation |
| Time | Execution and review minutes | Variation between easy and difficult cases |
| Rules | Decisions and conditions | Ambiguity or disagreement |
| Data | Source, access and quality | Duplicates, missing fields and permissions |
| Changes | How often the process changes | Dependencies that could break the workflow |
| Error impact | Consequences of an incorrect output | Validation, alerts and recovery |
3. Calculate net time, including supervision
Hypothetical example: registering 120 enquiries per month at six minutes each takes 12 hours. If a pilot left two minutes of review per enquiry and required one monthly hour of supervision, five hours of work would remain, releasing seven hours of capacity. These are illustrative assumptions, not ALKATECH results.
Include implementation, licences, maintenance and incident handling in the assessment. Released time does not automatically mean lower payroll or higher revenue; decide how the capacity will be used.
4. Choose the solution after understanding the process
Check whether an existing application already offers a suitable feature. When connecting applications, establish which data they can exchange and under which permissions. If a direct connection is unavailable, assess other approaches and their maintenance requirements.
For a constantly changing process or disputed rules, begin by clarifying the workflow. Repeating an incorrect rule automatically does not make it correct.
5. Run a pilot that supports a decision
Limit the pilot to one enquiry type and retain a manual fallback. Test ordinary, duplicate and incomplete cases. Record execution time, corrections, exceptions and supervision.
Compare the results with the baseline and explain the differences. The outcome should support a decision to expand, adjust or stop. The worksheet also gives a potential supplier a concrete process to discuss.
Frequently asked questions
Does every repetitive process deserve automation?
No. Volume, maintenance costs and the consequences of errors may favour a manual improvement or an existing application feature.
Is artificial intelligence required?
Not always. Clearly specified decisions may only require conventional automation. AI should address a specific need.
What should I prepare for an initial meeting?
A non-sensitive example, the current steps and tools, approximate volume and a list of exceptions.
Sources and scope of evidence
Research figures describe the samples and conditions stated, not ALKATECH results. The examples and templates in this guide are illustrative.
Syed, R., Suriadi, S., Adams, M., Bandara, W., Leemans, S. J. J., Ouyang, C., ter Hofstede, A. H. M., van de Weerd, I., Wynn, M. T., & Reijers, H. A. (2020). Robotic Process Automation: Contemporary themes and challenges. Computers in Industry, 115, 103162. DOI: 10.1016/j.compind.2019.103162.
Read the research — Syed et al., 2020