Herm, L. V., Janiesch, C., Reijers, H. A., & Seubert, F. (2021). From Symbolic RPA to Intelligent RPA: Challenges for Developing and Operating Intelligent Software Robots. In A. Polyvyanyy, M. T. Wynn, A. Van Looy, & M. Reichert (Eds.), Business Process Management – 19th International Conference, BPM 2021, Proceedings (pp. 289-305). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 12875 LNCS). Springer. https://doi.org/10.1007/978-3-030-85469-0_19
Abstract
Robotic process automation (RPA) is a novel technology that automates tasks by interacting with other software through their respective user interfaces. The technology has received substantial business attention because of its potential for rapid automation of process-driven tasks that would otherwise require tedious manual labor. This article explores the dichotomy between the practical reality of symbolic RPA, which requires handcrafting robots using process models and rulesets, and the promise of intelligent RPA, which relies on artificial intelligence technology to implement intelligent robots. Our research is based on a scholarly literature review as well as an interview study to derive and discuss challenges for this transition. We found that issues such as the lack of training data, human bias in data, compliance issues with transfer learning, poor explainability of robot decisions, and job-security-induced fear of AI robots all need to be addressed to enable the transition from symbolic to intelligent RPA.