Chatbots for customer service: A quantitative research

Abstract

In this thesis we have investigated the advantages and disadvantages of chatbots for customer service. Chatbots gained popularity because of their potential to partly automate customer service centres. Their success depends on the company’s capability to implement the solution, and the customers motivation of using the chatbot. Previous research suggests that cost cutting, setting a reliable scope, transparency of chatbot capabilities and well written responses, are key contributors to chatbot success. The customers motivations for utilizing chatbots are efficiency, productivity, previous experience, human-likeness, and trust in chatbots. We performed five types of experiments in order to challenge previous research on established truths regarding chatbots. Our experiments included comparative analysis, sentiment analysis, regression, and random forest. Our results indicate that customers are less patient with chatbots than human agents. We found that the users’ sentiment towards chatbots is more negative compared to conversations with human agents. Furthermore, we found that when customers use less time and has to write fewer messages to get their inquiry resolved it has a positive effect on customer satisfaction. Asking a chatbot uncomplicated questions instead of complex questions also had a positive effect on customer satisfaction. Based on findings from previous research, our research, and interviews, we recommend several measures to managers that are considering acquisition of chatbots. These are: 1) gradual increase in functionality, 2) include human-likeness, and 3) transparency.