Domain-Specialized Agentic AI Analytics for Secure Multi-Context Enterprise Decision System
Main article
Abstract
Agentic artificial intelligence is increasingly used to connect enterprise data, automate workflows, and support decisions across business domains. However, enterprise decisions are rarely contained within a single data source or a single policy context. They span finance, supply chain, customer operations, compliance, cybersecurity, and legacy platforms with different rules and risk thresholds. This article develops a domain-specialized framework for secure multi-context enterprise decision systems. Building on the core ideas of mesh-based multi-context coordination, the paper argues that safe agentic analytics requires domain nodes that combine local knowledge, validated tools, data-governance policies, and auditable decision rights. The framework is presented as a layered architecture and a six-stage decision cycle: request parsing, policy evaluation, local action, remote delegation, result aggregation, and assurance. A scenario-based evaluation compares three architectures--centralized agent, hub-style multi-context access, and domain-specialized mesh analytics--across decision accuracy, privacy control, fault isolation, legacy compatibility, and auditability. The results suggest that domain specialization improves privacy, resilience, and accountability even when it increases initial design effort. The paper contributes a practical roadmap for incremental adoption and a research agenda for evaluating agentic AI decision systems in regulated enterprise environments.
