A Survey on Multi-Agent Systems for AI-Driven Data Automation

Abstract

The increasing reliance on data-driven decision- makingdemandsintelligentsystemsthatcanautonomouslyplan, govern, and enforce data processes while ensuring compliance and scalability. This paper surveys advancements in multi-agent platforms and introduces a conceptual framework comprising fourspecializedagents:aTaskPlannerAgent,aDataGovernance Agent,aDataEnforcementAgent,andaSyntheticDataCreation Agent. Leveraging Large Language Models (LLMs) where ap- propriate, the Task Planner Agent interprets natural language input to generate actionable roadmaps, enabling automated orchestrationofcomplexworkflows.TheDataGovernanceAgent oversees adherence to policies and regulatory standards, while the Data Enforcement Agent ensures their execution in realtime, safeguarding integrity and compliance. The Synthetic Data Creation Agent generates privacy-preserving data to support ex- perimentation and model development. Collectively, these agents establish an LLM-enhanced platform designed to address plan- ning,governance,enforcement, anddataaugmentation challenges in modern data ecosystems

Introduction

The work proposes a multi-agent, LLM-powered platform designed to manage modern data workflows in distributed knowledge environments. It addresses challenges in digital collaboration such as fragmented tools, high cognitive load, and lack of scalable automation. The system integrates four specialized agents—task planning, data governance, enforcement, and synthetic data generation—working together to automate planning, ensure compliance, enforce security policies, and generate privacy-preserving datasets.

The literature review highlights rapid progress in LLM-based multi-agent systems, planning frameworks, and synthetic data generation, while also identifying gaps such as limited real-time enforcement, weak inter-agent coordination, and insufficient privacy-preserving solutions. The research gap further emphasizes the lack of unified systems that combine governance, structured planning, and multi-modal data handling in a scalable way.

Methodologically, the system is built using a full-stack architecture with React frontend, Node.js backend, PostgreSQL database, and LLM tools like GPT-4, LangChain, and Hugging Face. Apache Kafka enables inter-agent communication, while tools like OPA and Vault ensure security and policy enforcement. The system also includes monitoring, explainability, and synthetic data modules.

Experiments show that the platform successfully executes complex multi-step workflows with high efficiency, achieving about 97% task success rate and reducing workflow time by roughly 90% compared to manual processes. It demonstrates strong performance in automation, compliance enforcement, and synthetic data quality, while reducing user effort and improving decision-making speed.

Limitations include dependence on external LLM APIs, limited integration with external systems, scalability concerns under heavy loads, and imperfect synthetic data coverage for rare cases. Ethical considerations such as privacy, explainability, and user trust are also emphasized, along with the need for better transparency and gradual adoption in real-world settings.

Conclusion

This paper presents the design, architecture, and implemen- tation of a multi-agent, LLM-enhanced platform for planning, governance, enforcement, and synthetic data generation. By coordinating specialized agents and leveraging LLM reason- ing, the system automates complex multi-phase workflows while maintaining compliance and privacy, demonstrating the potential of intelligent, integrated platforms in modern data ecosystems. Future work will focus on expanding functionality, improv- ingscalability,andenhancingusertrust,asoutlinedinTableII.

References

REFERENCES [1] Y. Liu et al., ”LaMMA-P: Generalizable Multi-Agent Long-HorizonTaskAllocationandPlanningwithLM-DrivenPDDLPlanner,”inProc.Int. Conf. Autonomous Agents and Multiagent Systems (AAMAS), 2024. [2] H.Zhangetal.,”LLM-BasedMulti-AgentDecision-Making:Challengesand Future Directions,” arXiv preprint arXiv:2403.12345, 2024. [3] A.Torresetal.,”Multi-AgentSystems:ASurveyAboutItsComponents,Framework and Workflow,” IEEE Access, vol. 12, pp. 45123–45145,2024. [4] M. Chen et al., ”Agentic AI: A Comprehensive Survey of Tech-nologies, Applications, and Societal Implications,” arXiv preprintarXiv:2405.09876, 2024. [5] R.Pateletal.,”TheConfluenceofEvolutionaryComputationandMulti-Agent Systems: A Survey,” Information Sciences, vol. 672, pp. 118019,2024. [6] S. Gupta et al., ”LLM-Driven Multi-Agent Architectures for IntelligentSelf-Organizing Networks,” in Proc. IEEE Int. Conf. Communications(ICC), 2024. [7] J. Wang et al., ”A Human-Like Reasoning Framework for Multi-PhasesPlanning Task with Large Language Models,” in Proc. Conf. EmpiricalMethods in Natural Language Processing (EMNLP), 2024. [8] P. Kumar et al., ”An Overview of Data Quality Frameworks,” ACMComput. Surveys, vol. 56, no. 2, pp. 1–35, 2024. [9] L. Fernandez et al., ”MAG-V: A Multi-Agent Framework for SyntheticData Generation and Verification,” in Proc. NeurIPS, 2024. [10] K. Singh et al., ”Agentic Workflows for Improving Large LanguageModel Reasoning in Robotic Object-Centered Planning,” in Proc.IEEE/RSJ Int. Conf. Intelligent Robots and Systems (IROS), 2024. [11] Y. Zhao, J. Wu, A. Van Moorsel, and Y. Chen, ”VT-GAN: CooperativeTabular Data Synthesis using Vertical Federated Learning,” in Proc.IEEE Int. Conf. Data Engineering (ICDE), 2023.

Copyright

Copyright © 2026 Suyog Thigale, Indraneel Jadhav, Arya Gaikwad, Pranshu Devhade, Prof. Krittika Goswami . This 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.