AIRPA: An Architecture to Support the Execution and Maintenance of AI-Powered RPA Robots

Abstract

Robotic Process Automation (RPA) has quickly evolved from automating simple rule-based tasks. Nowadays, RPA is required to mimic more sophisticated human tasks, thus implying its combina tion with Artificial Intelligence (AI) technology, i.e., the so-called intelli gent RPA. Putting together RPA with AI leads to a challenging scenario since (1) it involves professionals from both fields who typically have different skills and backgrounds, and (2) AI models tend to degrade over time which affects the performance of the overall solution. This paper describes the AIRPA project, which addresses these challenges by proposing a software architecture that enables (1) the abstraction of the robot development from the AI development and (2) the monitor, con trol, and maintain intelligent RPA developments to ensure its quality and performance over time. The project has been conducted in the Serv inform context, a Spanish consultancy firm, and the proposed prototype has been validated with reality settings. The initial experiences yield promising results in reducing AHT (Average Handle Time) in processes where AIRPA deployed cognitive robots, which encourages exploring the support of intelligent RPA development.