Abstract
Modern enterprises integrate heterogeneous on-premises, multi-cloud, SaaS, IoT, and AI-enabled systems, creating coordination, scalability, resilience, governance, and real-time context challenges for conventional integration approaches. Although Event-Driven Architecture and event streaming are established, the literature provides limited vendor-neutral guidance that integrates Event Mesh federation, event governance, observability, hybrid-cloud deployment, and selected agentic-AI integration concerns into one enterprise reference framework. This study proposes the Enterprise Event Mesh Reference Architecture (EEMRA) to address that gap. Using Design Science Research Methodology, the study identifies the integration problem, synthesizes prior literature, develops the EEMRA artifact, performs a qualitative capability comparison, and examines illustrative enterprise cases. The framework organizes enterprise systems, event enablement and production, the Event Mesh fabric, integration services, event consumers, and cross-cutting governance and observability. Principal contributions include EEMRA, a six-responsibility architecture model, the Enterprise Event Taxonomy Framework, an Agentic Integration Architecture, an Event Mesh + Model Context Protocol integration pattern, and the Enterprise Nervous System Principle. Findings indicate that Event Mesh can provide a useful asynchronous, federated integration substrate for event-reactive workloads, while performance and business-impact claims remain implementation-dependent and are not experimentally validated in this study. The broader implication is a structured, vendor-neutral basis for designing governed event fabrics and integrating them with broader AI-agent architectures.