Evaluating the Impact of Assignment Group and Category Classification Prediction of Incoming Service Requests on the Perceived Service Quality

Abstract

Machine learning (ML) is reshaping customer service, tackling the growing complexity and volume of customer requests. This article investigates the effects of ML on perceived service quality (PSQ) across various customer service measures within an organizational context. Utilizing a quasiexperimental design, we analyzed 131 978 service requests submitted to the service desk of a large enterprise software organization. Over a 2-year period, these requests were made by 1252 organizations and were associated with 85 654 predictions and 19 720 returned PSQ postservice questionnaires. Our regularized logistic regression model aimed to categorize service requests into 56 categories and dispatch them into one of nine assignment groups to enhance resolution efficiency. Contrary to expectations, the overall PSQ did not significantly improve, while five specific metrics, such as time to resolve and first-time resolution, improved. This may be attributed to the increase in the first personal response time. The article highlights the complexities of implementing ML-based classification and underscores the importance of organizational structure. We found that expert groups prioritizing accurate problem-solving over quick responses led to an increase in the response time for incoming service requests. Theoretical contributions include an understanding of how classification in customer service affects PSQ, offers practical tactics to counteract negative impressions, and sets the groundwork for future article on ML in customer service management, despite the limitations, such as the potential influence of external factors and the study's generalizability.