Agentic Service-Oriented Computing: A New AI Engineering Frontier

The rise of LLM-powered autonomous agents is transforming software systems, moving beyond static components to goal-directed, adaptive actors. However, as these agents transition from isolated prototypes to complex distributed workflows, they encounter significant engineering challenges that the Service-Oriented Computing (SOC) community has long addressed, such as composition, interoperability, and governance. This paper argues that the current ad hoc development of agentic AI lacks the rigor needed for reliable enterprise and societal deployment. To address this, the authors propose Agentic Service-Oriented Computing (ASOC) as a new field. ASOC focuses on engineering agents as services, orchestrating services through agents, and governing entire ecosystems of agents and services under constraints like trust, cybersecurity, and compliance. Six foundational principles are articulated: harness-ability, composability, lifecycle engineering, trustworthiness by design, goal-driven orchestration, and observability/accountability. The paper outlines a five-dimensional research agenda covering agentic services foundations, composition, governance, security, and evaluation. It posits that the Services Computing community is uniquely positioned to provide the conceptual and engineering backbone for this emerging field, transforming agentic AI from fragmented demonstrations into dependable, service-based systems worthy of trust.