The AI Sentinel: Leveraging Big Data Analytics and Predictive Systems to Mitigate Negative e-WOM and Enhance Service Recovery in Hospitality

The paper presents AI Sentinel, a closed-loop socio-technical approach to monitoring, analyzing, and responding to negative hotel reviews through a combination of big data analytics, natural language processing, and machine learning predictive modeling. A total of 85,178 reviews were analyzed for 80 European hotel properties, with 5665 (mean = 6.54) classified as negative and 79,513 (mean = 9.22) classified as positive. Latent Dirichlet Allocation (LDA) was used to discover topics; Gradient Boosting was used to classify high-risk reviews (AUC = 0.919); and a rule-based engine was employed for routing recovery/delivery of service. This analysis identified ten major complaint areas in guest reviews, with Cleanliness, staff behavior, and room quality accounting for 47.0% of negative comments about hotels and forming the Critical tier of intervention. There are three key theoretical contributions made by this study: (1) establishing operationalization of joint socio-technical optimization in AI-augmented service management; (2) introducing algorithmic service sensing as a time-compression mechanism for recovery workflow; and (3) demonstrating that the integration of unsupervised topic modeling with supervised risk classifications can provide a compounded analytical approach. Managerial consequences include risk prioritization at the portfolio level, the design of specific services to target certain traveler segments, nationality-based recovery threshold levels, and an appropriate governance structure that meets the requirements of the General Data Protection Regulation and the new European Union Artificial Intelligence Act.