Upskilling software robots : how artificial intelligence can be applied to increase robotic process automation capabilities

Abstract

Robotic process automation (RPA) is an emergent business process automation technology. RPA automates predefined sequences of steps to mimic human actions interacting with the user interfaces of other information systems. While RPA allows organizations to automate a variety of activities with software robots, the technology strongly depends of scripts based on predefined rules and structured data. This limits the use of technology in complex human activities such as reacting to unexpected events, analyzing unstructured information, and performing other activities that require cognitive skills. In computing, the area that studies and develops solutions to reproduce human cognitive skills by machines is Artificial Intelligence (AI). In this study, we investigated what tasks are feasible and unfeasible to automate with RPA and what kind of complexity can be overcome with the increment of AI techniques and technologies. Our study produced results based on the combination of knowledge found in the academic and industry communities using scientific and reproducible methods. With a Systematic Literature Review (SLR) we assessed academic articles and papers indicating the combined use of RPA and AI. The SLR selected and analyzed 91 studies with contributions to the understanding of the current abilities and limitations of RPA and how to provide new abilities to software robots by combining AI technologies with RPA. The SLR was complemented with a survey performed with 86 professionals from industry who implement and use the technology, capturing and analyzing their impressions on practical use, limitations, and what applications have they done with AI to improve robotic automations. As a result, we were able to identify the abilities and limitations of RPA regarding data manipulation, process logic, and software applications operated during work execution. We have also found 39 new abilities that can be implemented by combining AI techniques and technologies with RPA. These abilities were organized in five categories: process improvements, data processing improvements, robot-human interaction improvements, increment of robot autonomy and development of improved cognitive skills. With this study, we created a complete panorama of the current state of RPA and the possible evolution of the technology, producing a useful report to assist organizations and solution providers to create more autonomous and intelligent robotic automation solutions.