RPA: How to Choose Which Processes to Automate First
Published on 21/06/2026 · Data Architecture Team, Innovations BI
Robotic Process Automation (RPA) has a common trap: starting with the wrong process. A poorly chosen bot becomes eternal maintenance; a well-chosen one pays for itself in weeks. These are the criteria we apply before writing a single line of code.
1. The process follows clear rules
RPA works best when every decision in the process can be written as a rule: "if the invoice exceeds X amount, it requires approval." If the process depends on human judgment at every step, the right automation isn't a bot — it's an approval workflow.
2. It repeats with frequency and volume
A process that takes 15 minutes but runs 40 times a day is worth more than a 2-hour one that happens once a month. Multiply frequency by time by the cost of the person doing it: that's your potential annual savings.
3. The systems involved are stable
If the system the bot must operate changes its interface every two weeks, maintenance will eat the savings. Prioritize processes on mature systems, or automate via API when one exists.
4. Human error has a real cost
Data-entry processes between systems (copying data from email to the ERP, from the ERP to a spreadsheet) are the favorites to start with: high volume, clear rules, and every typo costs money or rework.
One final tip
Document the process before automating it. In half the projects we've done, simply mapping the process step by step revealed unnecessary steps that were eliminated without needing a bot. Automating a bad process just makes the bad thing happen faster.