Evaluating Agentic AI Across Domains: A Comparative Study of Applications and Models
Publisher: IEEE
Authors:
Modi Nandini, India； Gandhinagar； School of Technology； Pandit Deendayal Energy University；Department of Computer Science Engineering Kumar Yogesh, India； Gandhinagar；Department of CSE； School of Technology； Pandit Deendayal Energy University Amer Ayman, Faculty of Engineering； Jordan； Zarqa Univeristy Muda Zakaria Che, Malaysia；Faculty of Engineering and Quantity Surveying INTI-IU University Nilai Jose Abey, School of Allied Health University of Limerick； Ireland Srinivasu Parvathaneni Naga, India；Amrita School of Computing； Amrita Vishwa Vidyapeetham； Amaravati Manzoor Muhammad Umair, Australia；School of Engineering RMIT University； MelbourneIjaz Muhammad Fazal, Australia；Torrens University
Abstract:
Agentic Artificial Intelligence (AI), defined by its autonomy, reactivity, proactivity and adaptive learning capabilities, is rapidly transforming diverse sectors ranging from healthcare and finance to manufacturing, education and security. Despite its growing impact, there is limited systematic evaluation of Agentic AI performance across application domains. This study addresses the gap by using multiple machine learning models including Logistic Regression, LinearSVC, MultinomialNB, SGDClassifier, Extra Trees, Random Forest, XGBoost, LightGBM, and ensemble Voting Classifiers on a curated dataset of Agentic AI applications. The methodology integrates preprocessing, feature engineering, and model optimization, followed by rigorous evaluation using accuracy, precision, recall and F1-score. Results demonstrate that ensemble approaches, particularly the Hard Voting Classifier, achieve the highest overall accuracy (94.7%), while domain-specific performance varies. Further analysis reveals patterns of application adoption across industries underscoring the domain-dependent nature of Agentic AI deployment.
