Garima Agarwal
The increasing incorporation of artificial intelligence into software systems in various enterprises has further complicated architectural concerns related to scaling, security, governance, and adaptability. Although cloud-based and event-based architectural approaches are promising for modern enterprise systems, existing architectural approaches rarely consider artificial intelligence incorporation, cloud-based systems, and event-based systems as integral parts. These further limits the potential for artificial intelligence-based decision-making systems in various enterprises. This paper proposes a systematic architectural framework for AI-augmented enterprise software systems, which combines cloud-native design principles and event-driven coordination. By using a design science research approach, a systematic approach is followed in developing the architecture using synthesized architectural requirements from enterprise systems, AI service lifecycles, and distributed cloud environments. In addition, a layered abstraction is introduced in the proposed architecture for enterprise software systems, where AI services are treated as first-class components. The framework is validated using analytic, quality attribute-driven architectural reasoning, thus verifying its ability to tackle key concerns of an enterprise, such as scalability with varying AI-driven workload, secure interactions between services, resistance to partial failures, and long-term maintainability. By abstracting architectural patterns rather than tying them to implementation technologies, we provide a generalized blueprint for designing enterprise-grade AI systems. The proposed framework offers practical advice to enterprises looking to move from initial AI system deployment to scalable, secure, adaptable, and cloud-native systems.