Abstract
Intelligent Process Automation (IPA) is an emerging technology with a primarygoal to assist the knowledge worker by taking care of repetitive, routine andlow-cognitive tasks. Conversational agents that can interact with users in anatural language are potential application for IPA systems. Such intelligentagents can assist the user by answering specific questions and executingroutine tasks that are ordinarily performed in a natural language (i.e.,customer support). In this work, we tackle a challenge of implementing an IPAconversational assistant in a real-world industrial setting with a lack ofstructured training data. Our proposed system brings two significant benefits:First, it reduces repetitive and time-consuming activities and, therefore,allows workers to focus on more intelligent processes. Second, by interactingwith users, it augments the resources with structured and to some extentlabeled training data. We showcase the usage of the latter by re-implementingseveral components of our system with Transfer Learning (TL) methods.