Towards Continual Learning in Interactive Digital Assistants for Process Automation

Abstract

Business process automation solutions often leverage digital assistants to make a set of automation functions accessible to non-tech-savvy end-users who leverage these functions to define process automation workflows. However, these digital assistants, made up of a fixed set of automation skills, are seldom easily customizable to better fit the end-users' process management requirements. In addition, they do not leverage contextual information to adapt and optimize to changing process states. This can impact the widespread adoption of process automation solutions, particularly in highly regulated or fast changing industries. In this work, we argue that the next era of digital assistants is one where these assistants can be taught new process automation capabilities by end-users in an interactive manner using natural language and other intuitive modalities.%, in an effort to expand their capabilities. To achieve this, we identify and present several key research challenges at the intersection of AI, process management, and interactive task learning that must be addressed to realize the vision of digital assistants that continually learn new automation solutions from user interactions. Business process automation solutions often leverage digital assistants to make a set of automation functions accessible to non-tech-savvy end-users who leverage these functions to define process automation workflows. However, these digital assistants, made up of a fixed set of automation skills, are seldom easily customizable to better fit the end-users' process management requirements. In addition, they do not leverage contextual information to adapt and optimize to changing process states. This can impact the widespread adoption of process automation solutions, particularly in highly regulated or fast changing industries. In this work, we argue that the next era of digital assistants is one where these assistants can be taught new process automation capabilities by end-users in an interactive manner using natural language and other intuitive modalities, in an effort to expand their capabilities. To achieve this, we identify and present several key research challenges at the intersection of AI, process management, and interactive task learning that must be addressed to realize the vision of digital assistants that continually learn new automation solutions from user interactions.