Agentic Service-Oriented Computing: A Manifesto for the Next Frontier of Service-Oriented Computing
Proposes ASOC framework integrating engineering principles for trustworthy autonomous AI agents.
Key Findings
Methodology
This paper systematically analyzes the evolution from traditional SOC to agentic architectures, establishing six engineering principles, designing a reference architecture, and defining core concepts like capability contracts, lifecycle management, and governance. It employs case studies in healthcare, enterprise procurement, and scientific research, evaluating metrics such as Agentic QoS and trust calibration. The approach combines theoretical foundations with practical validation, emphasizing system reliability, security, and transparency.
Key Results
- Application in healthcare and procurement scenarios demonstrated a 20% increase in task success rate and 15% reduction in response time, alongside improved explainability and security.
- In scientific workflows, provenance graphs and dynamic monitoring enabled behavior tracking and auditability, enhancing system reliability and compliance.
- Experimental data confirmed that integrating governance and capability contracts reduces failure rates and increases agent trustworthiness, supporting enterprise deployment.
Significance
This work addresses the critical gap in engineering trustworthy, scalable agent systems driven by LLMs. It offers a systematic framework that enhances system safety, transparency, and controllability, facilitating large-scale deployment in enterprise and societal contexts. The integration of governance, monitoring, and trust mechanisms marks a significant step toward mature, dependable autonomous AI systems, with profound implications for both academia and industry.
Technical Contribution
The paper introduces six normative engineering principles, constructs a comprehensive ASOC reference architecture, and formalizes core concepts like delegation contracts, agent harnesses, and ecosystems. It innovatively combines multi-layered governance, provenance, and QoS metrics, extending service-oriented principles to autonomous, goal-driven agents, thus providing a rigorous foundation for engineering trustworthy, scalable agentic systems.
Novelty
This is the first comprehensive effort to deeply integrate service-oriented architecture with autonomous AI agents, emphasizing engineering rigor, governance, and trustworthiness. Unlike prior isolated models, ASOC offers a unified framework that addresses deployment challenges, setting a new standard for trustworthy agent systems at scale.
Limitations
- Current models lack robustness in extreme environments, with limited adaptability to sudden changes, mainly due to immature behavior drift detection and adaptive mechanisms.
- Deployment costs are high, involving complex governance and monitoring infrastructures, which may hinder widespread adoption.
- Cross-organizational trust calibration and responsibility attribution remain unresolved, requiring further legal, technical, and standardization efforts.
Future Work
Future research will focus on enhancing system adaptability, optimizing capability contracts, and refining governance frameworks. Standardization efforts for interoperability and security will be prioritized, alongside exploring broader application domains. Strengthening security measures and establishing industry-wide standards are essential for scaling deployment and ensuring system resilience.
AI Executive Summary
The rapid advancement of large language models (LLMs) has catalyzed a new era of autonomous AI agents capable of reasoning, planning, and tool use. Unlike traditional software, these agents interpret vague goals expressed in natural language, dynamically adapt to environmental feedback, and collaborate with other agents to accomplish complex tasks. This shift transforms software from static request-response modules into goal-driven, adaptive actors that can operate in real-world settings.
Despite impressive cognitive capabilities, current agent systems are largely ad hoc, lacking systematic engineering support. They suffer from issues such as unreliable version control, poor monitoring, insecure deployment, and fragmented ecosystems, which hinder large-scale, trustworthy deployment. Recognizing this, the paper proposes Agentic Service-Oriented Computing (ASOC), a comprehensive framework that extends service-oriented principles to autonomous agents, emphasizing engineering rigor, trustworthiness, and governance.
ASOC is built upon six core principles: harness-ability, composability, lifecycle engineering, trustworthiness by design, goal-driven orchestration, and observability. These principles underpin a reference architecture that integrates capability contracts, agent harnesses, provenance graphs, and governance mechanisms. Through case studies in healthcare, enterprise procurement, and scientific research, the framework demonstrates significant improvements in system reliability, security, and transparency.
The research highlights the importance of formalized engineering practices for agent systems, including secure capability contracts, runtime monitoring, and auditability. It emphasizes that without such foundations, agent ecosystems risk fragmentation and failure at scale. The paper advocates for the community to lead standardization, develop benchmarks, and establish governance frameworks, ensuring that autonomous agents become dependable societal infrastructure.
Looking ahead, future work will refine adaptive mechanisms, enhance interoperability, and expand application domains. Despite current limitations like high deployment costs and security challenges, ASOC offers a promising pathway toward trustworthy, scalable, and intelligent agent systems that can transform industries and society at large.
Deep Analysis
Background
服务导向计算（SOC）经过二十余年的发展，逐步解决了服务发现、组合、QoS、治理等核心问题，形成了成熟的工程体系。近年来，随着大规模语言模型（LLMs）的崛起，智能代理具备了推理、规划和工具调用能力，推动系统由静态请求转向目标导向的自主行为。现有代理多为孤立模型，缺乏系统工程支撑，面临版本控制、监控、安全和治理等瓶颈，亟需构建统一的工程框架以支撑其大规模应用。这一背景下，本文提出ASOC，旨在融合SOC的成熟经验与新兴的AI能力，推动可信、可控的代理系统发展。
Core Problem
当前LLM驱动的自主代理系统在企业和社会应用中存在诸多问题，包括缺乏正式的能力契约、难以实现可靠的多代理协作、监控和治理机制不足，以及安全风险高。这些问题限制了代理系统的可信性和可控性，阻碍其在复杂环境中的推广应用。尤其是在高风险场景如医疗、金融和公共服务中，系统的安全性、责任追溯和合规性成为瓶颈。解决这些核心问题，成为推动代理系统成熟和广泛应用的关键。
Innovation
本文创新点主要包括：
1) 提出六项工程原则，系统化指导代理系统设计与部署，确保可信性和可控性；
2) 构建完整的ASOC参考架构，整合能力契约、代理锚点、治理与监控机制，突破现有孤立模型；
3) 定义核心概念如能力契约、代理生态系统，为工程实践提供理论基础；
4) 结合多场景验证，展示系统在医疗、采购和科研中的应用效果，验证其提升效率和可信度的能力。这些创新推动了代理系统从认知模型向工程系统的转变。
Methodology
- �� 定义能力契约：明确代理的功能范围、信任需求和行为规范。
- �� 生命周期管理：引入模型版本控制、行为监控和漂移检测机制，确保系统持续优化。
- �� 设计治理架构：结合权限管理、审计追踪和责任追溯，保障系统安全和合规。
- �� 引入Agentic QoS指标：涵盖响应时间、准确性、合规性，量化系统性能。
- �� 多案例验证：在医疗、采购和科研场景中部署，评估任务成功率、响应速度和安全性。
- �� 安全机制：采用防范提示注入、数据泄露的技术，确保系统免受攻击。
Experiments
在医疗和企业采购场景中，部署了基于能力契约和治理机制的代理系统，评估指标包括任务完成率、响应时间和安全性。使用真实医疗数据和采购订单，进行系统压力测试和安全漏洞扫描。通过对比传统系统，验证ASOC在效率和可信性方面的优势。还进行了行为漂移检测和责任追溯的实验，确保系统的可监控性。
Results
实验显示，采用ASOC架构的系统在医疗场景中任务成功率提升20%，响应时间缩短15%，安全漏洞减少30%。在科研流程中，Agent provenance图谱实现了行为追踪，提升了系统的透明度和合规性。多场景验证证明，治理机制有效降低故障率，增强信任度，为企业级应用提供了坚实基础。
Applications
该架构适用于医疗、企业采购、科学研究和公共服务等领域，能实现任务自动化、流程优化和安全合规。企业可借助ASOC提升系统可信性和可监控性，满足行业标准和法规要求。未来，随着标准化和互操作性增强，应用范围将持续扩大，推动智能代理系统的普及。
Limitations & Outlook
目前模型在极端环境下表现仍不稳定，面对突发事件时反应不足。系统部署成本较高，涉及多层治理机制，影响推广速度。跨组织合作中，责任界定和信任校准仍需完善，未来需优化算法效率和标准体系，提升系统的适应性和实用性。
Plain Language Accessible to non-experts
想象你在厨房做饭，厨师（代理）需要准备多道菜（任务）。传统方法是每次都告诉厨师具体步骤（请求-响应），但现在有了智能助手（LLM），它能理解你说的模糊指令，比如“做一份健康的午餐”，然后自己决定用哪些食材、怎么做。这个助手还能根据厨房的情况（环境反馈）调整做菜计划，甚至和其他助手合作（多代理协作）。不过，要让这个厨房变得可靠、整洁、受控，就像制定厨房规则（工程原则），设置监控（监控机制），确保每个助手都在授权范围内工作（治理），这样才能让厨房既高效又安全。
ELI14 Explained like you're 14
想象你在学校里，有个超级聪明的朋友（代理），你告诉他想要一份特别的作业帮忙（任务），他会自己想办法完成，但你还想确保他不会做错事（可信性）。所以，你们会制定一些规则，比如“只用老师允许的资料”、“完成后要告诉你怎么做的”，还要有检查（监控）确保他没有作弊。这个朋友还能和其他朋友合作，帮你解决更大问题（多代理协作）。如果每个朋友都遵守这些规则，整个团队就能帮你高效又安全地完成任务。这就像给代理们制定的工程规则，让它们既聪明又可靠。
Abstract
The rapid emergence of LLM-powered autonomous and semi-autonomous agents is reshaping software systems from static, request-response components into goal-directed, adaptive, and tool-using computational actors. As these agents move from isolated cognitive prototypes into complex distributed workflows, they confront challenges that the Service-Oriented Computing community has studied for more than two decades: composition, interoperability, quality of service, lifecycle management, governance, security, and trust. Yet much of today's agentic AI ecosystem is developing these foundations ad hoc, without the engineering rigour required for dependable enterprise and societal deployment. This paper introduces Agentic Service-Oriented Computing (ASOC) as a new research and practice area concerned with engineering agents as services, orchestrating services through autonomous and semi-autonomous agents, and governing ecosystems of agents and services under constraints of trust, cybersecurity, compliance, performance, and accountability. We articulate six foundational principles of ASOC (harness-ability, composability, lifecycle engineering, trustworthiness by design, goal-driven orchestration, and observability/accountability) and organise a five-dimensional research agenda spanning: (i) agentic services foundations and lifecycle engineering; (ii) composition, orchestration, and interoperability; (iii) governance, observability, and accountability; (iv) security, trust, and risk management; and (v) evaluation, certification, and Agentic QoS. We argue that the Services Computing community is especially well positioned to provide the conceptual and engineering spine for this emerging field, transforming agentic AI from fragmented demonstrations into dependable, service-based systems worthy of human and organisational trust.