AI-ENABLED CUSTOMER-INTERACTION MODELS FOR IMPROVING SERVICE EFFICIENCY IN U.S. HOSPITALITY AND RETAIL OPERATIONS

Abstract

This study examined the relationship between AI-enabled customer-interaction models and service efficiency outcomes in U.S. hospitality and retail operations using a quantitative, observational research design. The analysis was informed by an extensive review of 82 peer-reviewed academic studies spanning service operations, information systems, retailing, hospitality management, and technology-mediated service research, which provided the conceptual and methodological foundation for variable selection, model specification, and interpretation of results. Empirical analysis was conducted using a multi-source dataset comprising 68,420 customer-service interaction episodes collected from 132 hospitality and retail sites, integrating operational interaction logs, site-level performance records, and linked customer feedback data where available. AI exposure was operationalized through the share of interactions handled within AI-mediated channels and the extent of automated containment, while capability quality was captured using response latency, knowledge coverage proxies, escalation behavior, and containment outcomes. Service efficiency was measured using time-based indicators, including average handling time, time-to-resolution, wait time, and service cycle time, alongside productivity and cost metrics. Descriptive results indicated that AI-mediated interactions accounted for approximately 51% of total recorded service volume, with containment achieved in 64% of AI-handled interactions compared to 19% in human-mediated channels. Regression findings demonstrated that higher AI exposure was associated with statistically significant reductions in average handling time (β = −1.12 minutes, p < .001), time-to-resolution (β = −4.85 minutes, p < .001), and overall service cycle time (β = −5.60 minutes, p < .001), after controlling for channel type, inquiry category, severity, peak demand, and site scale. Response latency and escalation were positively associated with longer resolution cycles, while containment was associated with lower recontact risk and higher first-contact resolution. Multilevel modeling showed that AI-related effects were primarily driven by within-site variation, although meaningful site-level differences persisted. Overall, the findings provided quantitative evidence that AI-enabled customer-interaction models functioned as operational mechanisms that improved service efficiency through faster resolution, higher containment, and reduced repeat contact when deployed within coherent omnichannel service architectures.