Major part of work nowadays is done, at least partly, with computers. In many cases a single company might contain dozens of different information systems, and as systems have evolved a lot during their history, they are often from different eras. As such, especially the older systems often contain gaps in their use when they are connected to newer systems, creating a need for human employees to fill those gaps with often simple series of actions, which are mundane but necessary. Robotic process automation is an emerging field that aims to automate these kind of straightforward processes in knowledge work, such as copy-pasting from a system to another or making certain decision-processes automated for certain, so- called "happy path" cases. RPA works in the presentation layer of the software, so no underlying programming code needs to be touched. When compared to more traditional IT-development, RPA is faster and cheaper, yet in some ways more error- prone, as simple user interface updates might cause need for robot reconfiguration. A single company may consist of hundreds or thousands of processes and sub- processes, and rooting out which ones should be automated is a huge task. Aim of this research is to find out how business processes should be chosen for robotic process automation implementation, and in which order they should be automated. The focus is in the preliminary assessment, ie. before any actual robot development has been done. Example questions are "what makes a process viable for robotisa- tion" and "what makes this process beneficial to automate." Currently almost all companies use different kinds of tools to assess these kind of questions, yet they are mostly subjectively and/or intuitively created without proper benchmarking. This research covers 3 companies, of which 2 or RPA clients and 1 is RPA vendor. Their models for preliminary assessment were studied, and the acquired data was complemented with interviews with key personnel of the preliminary assessment process. Then the results are analysed with analytical hierarchical processing (AHP) essential factors for the preliminary assessment mapped.
