Transforming Library Operations Through Artificial Intelligence: The Role of Intelligent Library Management Systems in the Digital Era

AI is now a practical, everyday part of the daily life of 21st-century academic libraries. The integrated library management systems that have traditionally been used for circulation, cataloguing and acquisitions are now supported by chatbot-assisted reference services, automated metadata creation, recommender-based discovery and large language model-enabled scholarly search. This paper reviews the perception of librarians about Intelligent Library Management Systems (ILMS) in the context of its impact on the transformation of library operations in Pakistani universities. The cross sectional survey was given to N = 220 librarians in ten Pakistani universities from January to July 2025. The six latent constructs measured with a five-point Likert scale are: AI Literacy (AIL), Intelligent Library System capabilities (ILS), Operational Efficiency (OPE), Integration Readiness (INT), User Service Quality (USR, mediator) and Transformation Outcomes Perception (TOP, outcome). All constructs had acceptable reliability values (Cronbach α = 0.84 – 0.88; composite reliability values were 0.78 – 0.84; and average variance extracted values were ≥ 0.47). A confirmatory factor model fit the data at conventional thresholds (χ²/df = 1.27; CFI = .951; TLI = .945; RMSEA = .036, 90 % CI [.026, .046]; SRMR = .044). The results of the OLS regression predicting transformation outcomes revealed that AI Literacy (β = +.18, p = .004) and Operational Efficiency (β = +.34, p < .001) each predict outcomes of transformation positively and independently, with Operational Efficiency being the strongest marketing-direct predictor. However, coefficient of the User Service Quality was the second largest (β = +.26, p < .001) and partially mediated the relationship between ILS capabilities and TOP (Sobel z = +3.58, p < .001; indirect effect ≈ 37 % of the total). The direct coefficient was no longer significant when controlling for USR, which was consistent with a "fully-mediated" interpretation of the direct ILS → TOP pathway, rather than "partially-direct". There was no interaction between ILS and Integration-Readiness, as the common assumption that the ILS effect would be multiplicative with Integration-Readiness was rejected (β = +.05, p = .47). Overall model explained 36.1 % of the variance (R² = .361). The pattern indicates that the way library operations are transformed by AI is through two main levers: the operational efficiency provided by the system itself, and the service quality made possible by AI-powered discovery and reference tools, and that upstream investment in AI literacy for librarians is a precondition for the delivery of either of these two levers.