Beyond automation: Unveiling the potential of agentic intelligence

摘要

Agentic intelligence encompasses artificial intelligence systems imbued with autonomous capacity, facilitating independent decision-making beyond conventional automation frameworks. This study applies the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) protocol across three authoritative databases and others, examining peer-reviewed literature from 2020 to 2025 (March). From 1800 identified publications, 191 studies met inclusion criteria, revealing research concentrations: 35.1 % (64 studies) address computational scalability, privacy/security vulnerabilities, ethical/legal ramifications, and interpretability constraints; 24.0 % (44) each examine Industry 4.0 transformation and healthcare intervention systems; 9.8 % (18) investigate autonomous cybersecurity architectures; 16.4 % (30) each focus on personalized digital assistance and theoretical foundations; while 14.7 % (27) each explore intelligent energy distribution networks, quantum computational integration with robotics, and sustainability imperatives. The analysis juxtaposes agentic intelligence against traditional automation, elucidating distinctive characteristics within variable environments. Foundational technological enablers, large language models, reinforcement learning algorithms, multi-agent system frameworks, cognitive architectures, and edge-cloud computing integration propel advancements in capabilities. Implementations spanning energy distribution networks, cybersecurity mechanisms, manufacturing ecosystems, healthcare platforms, and digital assistance technologies demonstrate transformative potential. The investigation identifies computational limitations, ethical challenges, and trust-related impediments, providing a comparative assessment of advantages and constraints. Future research trajectories, including quantum computation, are proposed to advance theoretical and practical dimensions. This synthesis provides researchers, practitioners, and policymakers with an authoritative examination of agentic intelligence foundations and implications for advanced autonomous system development. • Reviews evolution from traditional AI to agentic intelligence for autonomous systems. • Discusses enabling technologies: LLMs, RL, multi-agent systems, and cognitive architectures. • Explores applications in smart grids, cybersecurity, healthcare, and Industry 4.0. • Analyzes challenges such as scalability, ethics, and trust; proposes future directions. • Comparative analysis highlights agentic AI’s advantages in flexibility and decision-making.