Machine learning (ML) has become a critical enabler of modern logistics systems as global supply chains face increasing complexity, fluctuating demand patterns, and the need for more efficient export operations. This systematic review synthesizes evidence from 72 peer-reviewed studies published between 2020 and 2025 to examine the effectiveness of ML techniques in logistics demand forecasting and export operations optimization. Comprehensive searches were conducted across Scopus, Web of Science, IEEE Xplore, SpringerLink, and Google Scholar, following the PRISMA guidelines. The findings show that ML algorithms including Random Forest, Gradient Boosting, Support Vector Regression, LSTM networks, and hybrid deep-learning architectures consistently outperform traditional statistical models in forecasting accuracy, scalability, and adaptability. ML applications in export logistics, such as container allocation, port congestion prediction, customs risk profiling, documentation automation, and scheduling optimization, are emerging but remain less explored relative to general supply chain forecasting. Key challenges identified across the literature include data quality limitations, integration difficulties, lack of real-time interoperability, algorithmic interpretability issues, privacy risks, and skills gaps within logistics organizations. Despite these constraints, ML demonstrates substantial potential for enabling agile, cost-efficient, and resilient logistics systems. Future research should prioritize explainable ML models, integration of IoT-enabled real-time sensing, reinforcement learning for autonomous logistics planning, and blockchain-supported export documentation workflows. Overall, this review highlights the transformative role of machine learning in shaping next-generation logistics and export-operation ecosystems, offering actionable insights for researchers and industry practitioners.