Start by mapping the process as it really runs
Documented procedures rarely match daily practice. Before automating, we shadow the work, collect samples of the messy cases and measure volume, cycle time, error rates and exception paths. Automating a flawed process only makes the flaw faster, so some steps are removed or simplified before any tool is chosen. Interviews with the people doing the work usually reveal workarounds that never appeared in the official procedure.
We then rank candidates by volume, rule clarity, system access and risk. High-volume tasks with stable rules, such as invoice matching or onboarding checklists, usually come first. Judgment-heavy work with rare cases may stay human with better tooling. Process mining on system logs can add evidence, showing where cases stall and which paths are rare. Combined with interviews, it replaces opinion with data and helps leaders agree on the sequence of automation, so the first release delivers a visible, measurable improvement.
