Organizations invest heavily in artificial intelligence (AI) enabled Customer Relationship Management (CRM) systems to improve operational efficiency, customer responsiveness, and revenue growth, yet the empirical evidence on whether such investment produces measurable performance gains remains mixed. Grounded in the Resource Based View, Dynamic Capabilities Theory, and the IT productivity paradox literature, this paper tests a capability-mediated model of AI enabled CRM using data from 307 manager-level respondents across product-oriented (n = 109) and service-oriented (n = 198) organizations. The instrument comprised 30 items across nine latent constructs, refined through a three-phase pilot (expert review, informed pilot, blind pilot with exploratory factor analysis). Covariance-based structural equation modeling was conducted in R with the lavaan package using the Maximum Likelihood Robust (MLR) estimator. Confirmatory factor analysis supported a nine-factor model (CFI = .913, RMSEA = .068, SRMR = .045) with all standardized loadings exceeding .80 and Cronbach alpha values between .916 and .962. The structural model explained 84.6 percent of variance in business performance. The results reveal an investment paradox: strategic AI investment exerts no significant direct effect on performance (β = −.02, p = .832) but operates through three significant capability pathways validated by bias-corrected accelerated bootstrap (5,000 resamples): innovation capability (indirect = .288, 95 percent CI [.119, .464]), employee expertise (.216 [.093, .352]), and CRM process automation (.171 [.093, .260]). Industry type significantly moderates four predictor-performance relationships, with stronger effects in product-oriented firms. The study reframes AI as enabling infrastructure that creates value only through complementary organizational capabilities.