Business process automation has a low failure rate technically and a surprisingly high failure rate in practice — not because the tools don't work, but because businesses often automate the wrong thing first, lose momentum, and conclude automation "didn't work for us."
A better starting approach is to score candidate processes on two dimensions: how frequent and repetitive the task is, and how well-defined the rules behind it are. Processes that are both frequent and rule-based — invoice generation, appointment reminders, data entry between systems that don't talk to each other — are the highest-value, lowest-risk starting points. Processes that are infrequent or require judgment calls are poor candidates for a first automation project, regardless of how much time they seem to cost.
It also helps to automate the process as it should be, not exactly as it currently is. Teams often ask for automation of a workflow that's only convoluted because it grew organically around a limitation that no longer exists. A short review of "why do we do it this way" before automating frequently reveals steps that shouldn't exist at all.
Finally, treat the first automation as a proof of value, not a finished program. A single well-chosen process, automated cleanly and visibly saving staff time, builds the internal case for further investment far more effectively than an ambitious automation roadmap presented without a working example yet.