Hey,
I think first thing is to describe use cases:
1. Going to doctor:
a) You describe your problems;
b) depend on your description and your medical card information doctor choose:
diagnosis services ( various blood analysis, tests and etc.)
c) after all needed information gathered doctor can asign you treatment:
home;
hospital;
sanatory.
From every doctor steps process can shift dramatically, but we can describe main abstract steps - analysis, treatment.
This is process as described above, you can chose sub tasks from main process. Sub task list must depend on main task, available data, already executed sub tasks (you can not assign two x-rays in half a year) and so on.
So rule based available task list generation would be best option.
2. Account errors correction - where is various ways to solve accounting errors, depend on error, time when you solving it, accounting policies and so on. So better to give user task list and let him pick best options. Also task can be small processes, rule dependent.
3. Simple selling process - where could be a lot of turns of processes, so user can pick some task in middle of execution an change process flow. Again you need rules engine to describe which task at witch moment to execute.
So we need task repository, rule based selection of task.
My two cents.
