The rapid advancement of Large Language Models (LLMs) and agentic AI has enabled autonomous reasoning, planning, and tool-assisted execution for increasingly complex enterprise workflows. However, existing research primarily focuses on improving individual agent capabilities, with limited attention to the architectural foundations required for enterprise-scale deployment across diverse industries. As AI adoption expands across manufacturing, finance, healthcare, logistics, construction, agriculture, and energy, the challenge shifts from optimizing individual agents to designing architectures that support scalability, adaptability, governance, resilience, and continuous learning. This paper proposes the Adaptive Enterprise Agentic Architecture (AEAA), a capability-centric reference architecture that enables Continuous Architectural Adaptation (CAA) for enterprise intelligence. Rather than centering on individual agents, AEAA organizes intelligence around reusable business capabilities, modular domain ecosystems, layered memory, adaptive orchestration, and evidence-driven optimization. Its Architectural Intelligence Layercontinuously monitors enterprise operations and optimizes workflows, knowledge organization, governance, and resource allocation while maintaining explainability, security, and regulatory compliance. The framework is evaluated using FINA as a reference implementation across multiple enterprise domains, comparing static, dynamic, and adaptive architectures using metrics including scalability, adaptability, governance, resilience, architectural efficiency, and operational cost. The proposed architecture provides a generalizable foundation for continuously adaptive enterprise intelligence platforms capable of supporting long-term autonomous operations across heterogeneous organizational environments.
Cassel Scott-Emuakpor (2026) studied this question.