AI-Powered Methods for Smarter Decisions in Automated Machine Learning in Business Analytics

Abstract

The advent of AI-driven Automated Machine Learning (AutoML) has redefined business analytics, enabling organizations to automate data preprocessing, feature engineering, model selection, and hyperparameter optimization, thereby accelerating predictive modeling and expanding access to advanced analytics. This systematic narrative review synthesizes findings from 84 peer-reviewed and industry publications (2007–2024) sourced from Scopus, Web of Science, IEEE Xplore, Google Scholar, and SSRN. Applications across retail, finance, manufacturing, and healthcare demonstrate measurable impacts, including a 25% reduction in equipment downtime, 20% gains in customer engagement, and deployment cycle reductions of up to 70%. Core enabling methods such as neural architecture search, Bayesian optimization, and meta-learning enhance predictive accuracy and operational efficiency, while sector-specific adaptations improve compliance and contextual relevance. Key challenges include model interpretability, computational scalability, data quality, bias mitigation, and integration into existing business processes. Emerging solutions, including federated learning, causal AutoML, and hybrid neuro-symbolic architectures, aim to address these constraints while safeguarding ethical and regulatory alignment. Future research should prioritize domain-specific, resource-efficient, and transparent AutoML frameworks that balance automation with human oversight, fostering robust, explainable, and operationally viable decision-support systems.