Recommending Next Best Skill in Conversational Robotic Process Automation

Avi Yaeli, Segev Shlomov, Alon Oved, Sergey Zeltyn, Nir Mashkif

Abstract: In recent years, Robotic Process Automation (RPA) has been widely adopted across the industry as an important enabler for business process automation and digital transformation. Recent advancements suggest that next generation RPA will require advanced human-robot collaboration capabilities for providing a more natural conversational interface and supporting more complex automation orchestration needs. Our work focuses on the nascent field of conversational RPA bots that are able to dynamically orchestrate automation tasks through natural language. In this context, recommending possible utterances and next steps to the user is an important capability to enhance human-bot collaboration. We take an exploratory approach to the problem of next-best-skill recommendation in human-robot collaboration. We highlight key characteristics of this problem, examine existing approaches, and call out specific challenges in implementing a solution. We suggest that this problem calls for an integrated strategy for recommendation, and illustrate a possible implementation architecture that can integrate multiple recommendation strategies.