Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{201173,
        author = {K Shreeshanth and Madhushree Stalin and Chethan Y and A. Priyadharshini},
        title = {A Review of Frameworks for Autonomous Multi-Agent Personal Assistants in Agentic AI Systems},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {12},
        number = {12},
        pages = {2848-2860},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=201173},
        abstract = {The shift from conventional command-based systems to intelligent, self-sufficient digital assistants has been made possible by recent developments in artificial intelligence. In order to create next-generation AI assistants, this review paper looks at the development and integration of Large Language Models (LLMs), multi-agent systems, memory structures, and governance mechanisms. This study aims to examine the ways in which these technologies support the creation of an Autonomous Multi-Agent Personal Assistant Framework, or ATLAS. Key contributions from current research are reviewed in this pa-per, including safety mechanisms like constitutional AI, multi-agent coordination frameworks, retrieval-augmented generation (RAG), and reasoning approaches like Chain-of-Thought prompting. It assesses how various methods deal with issues such system stability, task automation, contextual understanding, and personalisation. The paper also examines multimodal interaction strategies that improve accessibility and user experience, such as speech recognition and avatar-based interfaces. There is also discussion of the shortcomings of existing systems, including their computational complexity, weak memory integration, and security flaws. The results indicate that a scalable and safe system for intelligent automation can be achieved by integrating LLM-driven planning with modular execution agents and governance layers. The potential of these systems to revolutionise digital workflow management, productivity, and human-computer interaction is highlighted in the paper’s conclusion, which also outlines future research directions in adaptive intelligence and embodied AI systems.},
        keywords = {},
        month = {May},
        }