内容来源：https://aibusiness.com/agentic-ai/opinion-from-islands-to-ecosystems-why-interoperability-unlocks-scale-for-agentic-ai

内容总结：

打破AI孤岛：企业级生成式AI应用迈向“协作互联”新阶段

当前，生成式AI正成为数十年来最具影响力的企业创新技术之一。从IT、人力资源到客户服务与运营，各类专用AI智能体已逐步承担重复性任务、管理工作流并辅助员工工作，应用步伐持续加快。然而，随着部署规模扩大，一个关键挑战日益凸显：许多AI智能体虽在特定领域表现优异，却因设计孤立而无法相互通信，形成“AI孤岛”。

这种互操作性的缺失导致企业在跨部门、跨流程扩展AI应用时遭遇瓶颈——智能体之间工作重复、信息不畅、效率抵消，难以实现整体转型。业界逐渐认识到，AI的真正价值并非来自单点效率提升，而取决于智能体能否像人类团队一样协同工作。因此，互操作性（即智能体跨系统共享上下文、交换信息与协调行动的能力）已成为规模化部署AI的基石。

实现互操作需三大要素协同支撑：

1. 开放协议：支持跨平台、跨厂商的智能体通信；
2. 统一数据架构：提供安全、实时的信息访问，避免数据冗余；
3. 集中编排层：统筹智能体协作，确保流程透明、高效、可追溯。

以新兴的Agent2Agent（A2A）开放协议为例，该标准允许不同厂商、不同技术底层的AI智能体互相识别能力、分配任务并协调工作流，为构建“人机协同”生态系统提供通用语言。其应用场景已跨行业展开：

• 电信领域：预测型智能体可预警网络中断，服务智能体实时调配容量，客服智能体主动通知用户，多方联动预防业务中断；
• 制造业：维护智能体与供应链智能体协作，实时预防停产并管理突发状况；
• 政务服务：AI服务智能体协助市民办理许可证更新，同时合规智能体确保流程符合法规与隐私要求。

全球动力管理公司伊顿的实践印证了这一趋势。面对9.2万名员工日益增长的IT与人力资源服务需求，伊顿通过采用基于A2A的互操作智能体系统，实现了跨职能协同：一类智能体分派请求，另一类检索政策与知识库，其他则执行常规操作。统一编排层消除了冗余努力，带来处理时间缩短、工单减少、员工体验更主动流畅等成效。公司管理层评价，这一转变标志着从“单点自动化”迈向“智能体+工作流+生成式AI”的融合新模式。

值得注意的是，成功不仅依赖于技术。伊顿将成果归因于高质量数据、健全治理流程以及清晰的早期投资回报验证——这些要素为规模化扩展奠定了信任基础。如今，企业级AI应用已进入“治理与协同并重”的阶段：智能体的决策需可解释、行动需可审计、运行需受监督，而开放协议需内置企业级认证与审计功能，以保障协作既高效又可靠。

展望未来，企业若想将AI从试点项目转化为驱动业务的操作系统，就必须尽早布局互操作性生态。采用A2A等开放标准的企业，将更易打破孤岛、构建跨部门协同的AI运营模式，从而在智能系统与人类团队的高效协作中确立行业标杆。当前的核心议题已不再是“智能体是否应当协作”，而是“企业以多快速度实现协作”，以及“等待的成本是否能够承受”。

中文翻译：

谷歌云赞助
选择您的首批生成式AI应用场景

要启动生成式AI的应用，首先应关注能够提升人类信息交互体验的领域。

AI智能体无法独自改变企业格局，其真正的力量在于协同合作。

AI智能体已成为近几十年来最具影响力的企业创新之一。从IT、人力资源到客户服务与运营，跨行业跨职能的专用智能体正以日益增强的自主性承担重复性任务、管理工作流程并辅助员工。这一趋势势不可挡：越来越多组织正在生产环境中部署智能体，且应用速度持续加快。

但这一趋势也带来了新挑战：AI智能体之间无法互通。尽管智能体擅长特定任务，但其设计模式往往导致它们孤立运作。

许多早期部署在单一领域内取得成功，却在企业范围推广时陷入停滞。若缺乏互操作性，为某个工作流构建的AI智能体无法与负责其他流程的智能体协调，也无法与基于不同模型运行的智能体协作。其结果将是重复劳动、沟通障碍和数字瓶颈，这些风险可能抵消其带来的收益。

企业AI的未来不在于部署更多孤立的智能体，而在于实现它们的协同工作。正因如此，互操作性——即智能体跨系统共享上下文、交换信息和协调行动的能力——不仅是一项技术特性，更是实现规模化的基石。

若缺乏跨系统交换上下文或协调的机制，AI智能体只能创造零散的效益，而非推动企业整体转型。互操作性将彻底改变这一局面。

互操作性的核心需要三大要素协同作用：

• 开放协议：使智能体能够跨平台和供应商通信。
• 统一数据架构：在避免昂贵重复的前提下，提供安全实时的信息访问通道。
• 集中编排层：监督AI智能体的交互方式，确保协作透明、高效且可追溯。

新兴协议如Agent2Agent（A2A）正是这一架构的核心。A2A作为开放标准，旨在让AI智能体能够宣传自身能力、委派任务并协调工作流，无需考虑供应商或底层技术差异。通过A2A，企业可以构建智能体如人类团队般无缝协作的生态系统，打破限制智能体效能的孤岛。

这正是开放标准的意义所在：它们不仅连接系统，更建立了一种通用语言，使可扩展的跨供应商协作成为可能。这类协议推动互操作智能体规模化应用的力量已延伸至各行各业：

• 电信行业：预测型智能体可预判网络中断，服务智能体同步调整容量路由，客服智能体主动通知用户——多方协同防止业务中断。
• 制造业：维护智能体与供应链智能体协作，实时预防停机并管理突发状况。
• 政府服务：AI服务智能体协助市民更新证件，合规智能体则确保流程符合法规与隐私要求。

所有这些案例的价值都不源于孤立的效率提升，而来自协同编排的智能。

跨国电力管理公司伊顿展示了互操作性如何将AI从试点项目转化为操作系统。面对92,000名员工及日益增长的IT和人力资源服务需求，伊顿意识到孤立的机器人方案已不足够。

通过采用基于A2A的互操作AI智能体，伊顿构建了跨职能协调的系统：一个智能体分派请求，另一个检索政策与知识库，其余则执行常规操作。共享编排层保障了连续性，消除了重复劳动。

成果显著：问题解决时间缩短、工单数量减少，并为员工提供了更贴近对话的主动式交互界面。伊顿管理层将这一转变描述为“从零散自动化升级为智能体、工作流与生成式AI融合的增效配方”。

关键在于，他们将成功归因于技术之外的因素：高质量数据、严谨的治理流程，以及通过早期验证价值、争取扩展支持的清晰投资回报视角。如今，伊顿正将互操作AI智能体拓展至新领域，推动智能体AI从前景广阔的试点项目转型为企业的操作系统。

尽管技术不断进步，但仅靠互操作性仍不足够。企业还需要建立信任。AI智能体的每个决策都必须可解释，员工需要确信AI在安全边界内运行，监管机构与客户则需要系统具备可追责性的保障。

因此，治理是互操作性可持续发展的守护机制。理解AI智能体的决策逻辑、审计其行动记录，以及管理者介入的能力，对于负责任地实现规模化互操作至关重要。

在这方面，A2A再次发挥作用。该协议设计之初就考虑了企业级身份验证与可审计性，能够支持稳健的治理框架。与合作伙伴共同嵌入这些保障措施，可确保智能体协作既高效又可信。

智能体AI的愿景已然清晰：员工得以聚焦更高价值的工作，而AI智能体则处理常规性、重复性甚至预测性任务。为实现这一愿景，企业必须立即行动——当下优先布局互操作性至关重要。

采用A2A等开放标准的企业，将最具优势从零散的试点走向覆盖全企业的AI驱动运营模式。它们将为智能系统与人类团队的协作树立标杆，引领行业展示如何实现智能体AI的互联、编排与负责任治理。

问题已不再是智能体是否应该协同工作，而在于企业能以多快的速度实现这一目标，以及它们是否还能承受等待的代价。

英文来源：

Sponsored by Google Cloud
Choosing Your First Generative AI Use Cases
To get started with generative AI, first focus on areas that can improve human experiences with information.
AI agents can’t transform the enterprise alone. Their power comes when they collaborate.
AI agents have become one of the most impactful enterprise innovations in decades. Across industries and functions — from IT and HR to customer service and operations — specialized agents are taking on repetitive tasks, managing workflows and assisting employees with increasing autonomy. The momentum is undeniable: More organizations are deploying agents in production, and the pace of adoption is accelerating.
But this momentum also brings a new challenge: AI agents that can’t talk to each other. While AI agents excel at specific tasks, their design often means they operate in silos.
Many early deployments succeed within a single domain, but stall when scaled across the enterprise. Without interoperability, an AI agent built for one workflow can’t coordinate with an AI agent managing another, or one running on a different model. The result is duplicated work, miscommunication and digital bottlenecks that risk outweighing the benefits.
The future of AI in the enterprise depends not on deploying more isolated agents, but on making it possible for them to work together. That is why interoperability — or the ability for agents to share context, exchange information and coordinate actions across systems — is not simply a technical feature. It’s the foundation for scale.
Without a way to exchange context or coordinate across systems, AI agents create fragmented gains rather than enterprise-wide transformation. Interoperability changes that equation.
At its core, interoperability requires three elements working in concert:
Open protocols: allow agents to communicate across platforms and vendors.
Unified data fabrics: provide secure, real-time access to information without costly duplication.
Centralized orchestration layers: oversee how AI agents interact, ensuring collaboration remains transparent, efficient and accountable.
New protocols like Agent2Agent are at the heart of this architecture. A2A is the open standard designed to let AI agents advertise their capabilities, delegate tasks and coordinate workflows regardless of vendor or underlying technology. With A2A, enterprises can create ecosystems where agents collaborate as seamlessly as human teams, eliminating silos that limit agentic impact.
That is why open standards matter. They don’t just connect systems, they establish a common language that makes scalable, cross-vendor collaboration possible. The power of these types of protocols scaling interoperable agents extends across industries:
Telecommunications: Predictive agents can anticipate network outages while service agents reroute capacity and customer care agents proactively notify subscribers — all working together to prevent disruptions.
Manufacturing: Maintenance agents can collaborate with supply chain agents to prevent downtime and manage disruptions in real time.
Government services: AI service agents could help citizens renew licenses while compliance agents ensure the process follows regulations and privacy requirements.
In every case, the value comes not from isolated efficiency but from orchestrated intelligence.
Eaton, a multinational power management company, illustrates how interoperability transforms AI from pilots into operating systems. With a workforce of 92,000 employees and rising demand for IT and HR services, Eaton knew siloed bots weren’t enough.
By adopting interoperable AI agents powered by A2A, Eaton created a system where agents could coordinate across functions: one to triage requests, another to retrieve policies and knowledge and others to execute routine actions. A shared orchestration layer ensured continuity and eliminated redundant effort.
The results were clear: faster resolution times, fewer tickets and a more conversational, proactive interface for employees. Eaton’s leaders described the shift as moving from one-off automation to a recipe where agents, workflows and generative AI combine to deliver better outcomes.
Critically, they credited success not just to technology, but to strong data quality, governance processes and a clear ROI lens that proved value early and secured support for expansion. Today, Eaton is extending its interoperable AI agents into new domains, scaling a model that turns agentic AI from a promising pilot into an operating system for the enterprise.
For all the technical advances underway, interoperability alone is not enough. Enterprises also need trust. Every decision made by an AI agent must be explainable. Employees need confidence that AI is operating within guardrails; regulators and customers need assurance that systems are accountable.
That’s why governance is the safeguard that makes interoperability sustainable. Transparency into how AI agents arrive at decisions, auditability of their actions, and the ability for leaders to intervene are critical to unlocking interoperability responsibly and at scale.
Here again, A2A plays a role. The protocol is designed with enterprise-grade authentication and auditability in mind, supporting robust governance frameworks. Working with partners to embed these safeguards ensures that agentic collaboration is both powerful and trustworthy.
The promise of agentic AI is clear: Employees have more time for higher-value work while AI agents handle the routine, the repetitive, and even the predictive. To realize that promise, enterprises must not wait — it is crucial to prioritize interoperability today.
Those who adopt open standards like A2A will be best positioned to move beyond fragmented pilots to AI-powered operating models that scale across the enterprise. They will set the standard for how intelligent systems and human teams collaborate and lead the way in showing how agentic AI can be connected, orchestrated, and governed responsibly.
The question is no longer whether agents should work together. The real question is how quickly organizations will make it possible, and whether they can afford to wait.