The decisions made (including data interpretation) in the process can be captured in a pre-defined logic. The exception rate is either low or can be included as well in the business logic.

The input in the process should either be electronic and easily readable or readable using a technology that can be associated with RPA (such as OCR). A good example is an invoice having the fields pre-defined.

We can differentiate 4 types of processes:

• Manual & non-repetitive: the process steps are performed by humans and can be different every time the process is executed
• Manual & repetitive: the steps in the process are performed by the user, and at least some of them are the same every time
• Semi-automated & repetitive: some of the repetitive steps have already been automated (using macros, Outlook rules, and so on)
• Automated: there are processes that have been already automated using other technologies than RPA

Processes that need to stay manual or are non-repetitive, due to the high exception rate or factors that cannot be integrated in a business logic, aren't good candidates for automation.

Processes that have been the same for a certain period of time and no changes are expected within the next months are good candidates for automation, provided they meet the other criteria as well.

RPA works by programming the robot to perform tasks at screen level (when the screen changes, the logic has to be taught). The higher the number of screens, the more elements have to be captured and configured prior to the process automation.

An automation's complexity increases with the number of decision points in the business logic. Basically, each one could multiply by two the number of scenarios.

Some applications are more easily automated (such as the Office suite or browsers), others heavily increase the automation effort (Mainframe, for example). And the more different the applications are, the number of screens will increase, as well (see previous point).

As previously stated, standard input is desirable. Yet there are cases in which one standard input (such as an invoice) has to be configured for each supplier that will be affected by the automation. Moreover, non-standard input can be of different complexity grades, with free text being the most complex.