Automating a bad process makes a bad process faster. The first question is always whether the work needs doing, then whether a person needs to do it.
Usually not. In most engagements the same headcount stops doing data entry and starts doing work that actually needs a human. If reducing headcount is the goal, say so up front so the plan is built for it honestly.
That is the normal starting condition. Mapping the mess is the first phase. Automating on top of an unexamined process is how businesses lock in their own bad habits.
Rarely. Most of the gain comes from connecting what you already have and removing the manual hops in between.