Before you automate, decide what work should disappear
Automating a poor process usually makes the problem faster, not better.
Before choosing a tool, understand why the work exists, who depends on it and which steps no longer need to happen at all.
Start with the decision
Ask what decision or customer outcome the process supports. This helps separate necessary controls from steps that are there only because they always have been.
Follow one real transaction from beginning to end. The gap between the documented process and the actual work is usually where the useful opportunities appear.
Remove before you connect
Remove duplicate approvals, data entry and handoffs first. Then standardize the information that still needs to move. Automation comes after that.
A small workflow that people can observe and improve is a better first release than an ambitious end-to-end build that is hard to measure or change.
Measure the human outcome
Track time returned to the team, errors avoided, customer wait reduced and decisions improved. Those measures keep the work tied to business value after the technology is delivered.