Large-scale enterprise Java applications often rely on hundreds of third-party libraries. Over time, many of these libraries become outdated, vulnerable, or incompatible with newer environments. Manually managing these vulnerabilities is time-consuming, error-prone, and increasingly difficult as systems scale. This paper presents an AI-assisted approach to automate and prioritize the remediation of dependency vulnerabilities in enterprise systems. By integrating static dependency analysis, security advisories—including Common Vulnerabilities and Exposures (CVEs), which catalog publicly known software flaws—and machine learning models trained on historical resolution patterns, the system can recommend upgrade paths, detect potential breaking changes, and propose targeted refactoring strategies. We evaluate this framework on a real-world enterprise application with over 200 dependencies. Our approach achieves a 60% reduction in manual triage time and improves detection of latent security issues. Furthermore, integration with continuous integration/continuous deployment (CI/CD) pipelines, such as Jenkins, enables proactive and continuous monitoring of dependency health. These findings contribute to both the theory and practice of secure software maintenance in enterprise-scale Java systems.