Agentic Artificial Intelligence and Autonomous Decision-Making: Implications for Intelligent Systems

Abstract

In this paper, a systematic literature review (SLR) is provided to the autonomous decision-making and intelligent system to explore the role of agentic artificial intelligence (AI). The identified studies were those published from 2020 to 2026 in the Scopus, Web of Science, IEEE Xplore and Google Scholar databases with the following pre-defined keywords: Agentic AI, Autonomous agents, Intelligent systems, Decision-making. After screening of duplicates and title/abstract and detailed full text screening, 612 records were found and 57 peer-reviewed studies were identified for detailed thematic synthesis, based on predetermined inclusion and exclusion criteria. The studies included in the collection fell into five abstract clusters: agentic AI architectures and frameworks, decision-making mechanisms, domain applications, ethical and governance issues, and future directions for research. The synthesis revealed that the dominant technical foundation for agentic AI is the large language model (LLM) based single- and multi-agent architectures that integrate planning, tool use, memory, and reasoning with human-in-the-loop oversight, with decision-making increasingly moving towards integrating reinforcement learning with LLM-based reasoning and human-in-the-loop. The reviewed applications included business/enterprise, health care, finance, cyber security, education, manufacturing, and scientific discovery, and the tourism and the destination-management contexts were relatively under-researched. Explainability, accountability, safety, bias, and management of increasingly autonomous decision authority were persistent challenges. The review also concludes that, although the use of agentic AI provides significant benefits in efficiency, adaptability, and decision quality within intelligent systems, the responsible integration of agentic AI will depend on coordinated advancements in technical architecture, human oversight mechanisms, and regulatory frameworks. Suggestions for future primary research and longitudinal evaluation are provided.