Process automation is one of the highest-return investments available to most organizations, and one of the most commonly started badly. The organizations that start badly typically begin with the process that is most visible, most requested, or most enthusiastically championed by the team that will do the automation — not the process that offers the highest return on the investment of automation effort. The result is a successful automation that does not meaningfully improve business performance, a disappointed leadership team, and a skepticism about automation that is hard to overcome.
A Framework for Prioritizing Automation
The framework I use evaluates process automation candidates on three dimensions. Volume: how many times does this process execute per day, week, or month? High-volume processes amplify the return on any automation investment. Manual effort: how much human time does a single execution require? Processes with high manual effort per execution have the most to gain from automation. Error rate and consequence: how frequently does manual execution produce errors, and what is the cost of those errors? Processes with high error rates in high-consequence contexts — financial reporting, compliance submissions, customer-facing transactions — have the most to gain from the consistency that automation provides.
Plot every candidate process on these three dimensions and the priorities become clear. The processes that combine high volume, high manual effort, and high error consequence are the ones that will produce results that the organization will notice. Start there.
RPA vs. API Integration vs. Workflow Automation
Not every process that can be automated should be automated using the same approach. Robotic Process Automation — software that interacts with applications through their user interfaces, the same way a human would — is appropriate for processes that involve legacy applications with no API, where the alternative is either manual execution or expensive custom development. It is fragile when interfaces change, and it requires monitoring and maintenance. API integration — direct system-to-system data exchange through programming interfaces — is more robust and more efficient where APIs exist, and should be preferred over RPA when it is available. Workflow automation — routing approvals, notifications, and task assignments through a workflow engine — is appropriate for processes that require human judgment at defined decision points and systematic tracking of where work stands in the process.