从生成式 AI 智能体,真正的变化是什么

过去两年,生成式 AI 在企业中的普及速度远超预期。但 IDC 指出,生成式 AI并不是终点。当 AI 只能生成内容时,它仍然是工具;而当 AI能够 感知环境、调用工具、执行任务并持续反馈结果,它才真正开始参与企业运行。

智能体正是在这一背景下出现。它不再局限于单点问答或流程辅助,而是以数字劳动力、流程协调者和决策顾问的形式深度嵌入业务流程。企业竞争的分水岭,也随之从是否部署 AI,转向“是否具备规模化、安全化、可治理地运行智能体的能力”。

在《IDC FutureScape:全球 Agentic AI 2026 年预测——中国启示》(Doc#CHC54084526,2026年1月)中,IDC 系统刻画了未来五年中国企业在智能体发展过程中将面临的十个关键转折点。

十大预测:智能体如何重塑企业的运行方式

预测 1|数据就绪度

到2027年,如果企业没有优先构建高质量的AI就绪数据,在扩展AI解决方案时将面临幻觉频发、错误率高的问题,导致生产力下降15%。

数据质量不再只是IT部门的KPI,而是企业的生存红线。如果投喂给智能体的数据是脏的、乱的、没有经过治理的,那么企业得到的将不是效率提升,而是需要耗费更多人力去修正错误的负生产力。

预测 2|定价

到2028年,传统的按席位收费模式将被淘汰。随着智能体作为数字劳动力接管大量重复性工作,70%的软件供应商将不得不重构其商业模式,转向按业务结果、交易量或自动化成果计费的新模式。

在智能体时代,传统的按席位收费模式将越来越难以匹配价值创造的实际形态。当一个智能体在典型场景下一天可以完成过去多个人工岗位累计才能完成的工作量时,按人头收费的定价逻辑将难以为继。

预测 3|智能体项目失败

到2028年,69%的企业自建智能体项目将因未能实现投资回报率目标(ROI)而被放弃,因为企业难以充分认识到项目实施的实际成本和价值。

企业往往会受市场热度裹挟而仓促启动智能体项目。然而,由于未能对潜在应用场景进行深度研判,开发团队被迫仓促推进的项目,往往陷入落地即闲置的窘境。在此背景下,选择能够打通数据、应用、治理全链路,且深度契合业务场景的合作伙伴,无疑是更具可行性的路径。

预测 4|客户体验智能体编排

到 2027 年,45% 的企业将管理跨多个渠道、应用程序和供应商的多智能体(Multi-Agent),从而实现更无缝、上下文更丰富的体验。

这里的编排并非指的是单纯的工作流配置编排,而是指构建支持多智能体动态协作的系统架构。未来的竞争不在于拥有一个超级智能体,而在于编排能力。企业建立智能体系统架构应避免过于刚性的流程,拥抱灵活的协同框架,让智能体与智能体、人类与智能体能够无缝协同工作。

预测 5|智能体服务体验

到2029年,30%的中国500强企业将运用AI客户服务智能体,主动且个性化地联系客户,在客户尚未意识到问题时就解决问题。

服务模式将发生根本性逆转,从被动响应投诉升级为主动解决问题。这种预判式的服务能力,将在存量市场中建立起全新的差异化体验。

预测 6|人工监督作为战略职能

到2027年,50%的AI驱动型企业应用部署将设立新的专业职位,负责监督智能体,作为合规核心,确保自主工作流中的结果可追溯。

智能体的自主性不等于无人值守。随着智能体权力的扩大,人类的角色必须从操作者转变为监督者,以确保在合规与伦理的安全边界内释放AI的能力与价值。

预测 7AI 卓越中心

到2027年,那些建立了成熟AI或智能体卓越中心(CoE)的企业,其创新、速度和服务质量将比竞争对手高出20%。

零散的烟囱式试点难以支撑AI的真正落地和组织的规模化创新,建立AI CoE卓越中心是弥合技术与业务鸿沟、实现跨职能规模化治理的关键组织保障。

预测 8|岗位角色转型

到2026年,中国500强企业中40%的岗位将涉及与智能体的深度协作,重新定义传统的初级、中级和高级岗位。

人才的定义正在被改写。未来的核心竞争力不再单纯是个人执行力,而是智能体的管理协同能力,即构建、指挥、评估和优化数字劳动力工作的能力。

预测 9Agent 战略顾问

到2031年,60%的中国500强CEO将利用智能体进行战略决策,这一趋势由市场波动性、创新速度要求,以及董事会层面对更快决策和智能驱动决策的多重需求推动。

智能体正在从业务一线的手脚进化为董事会的外脑。通过实时处理海量数据并进行情景模拟,它能为高层决策提供人类难以企及的数据广度与速度支撑。

预测 10AI 对业务的颠覆性影响

到2030年,多达20%的中国500强企业将因智能体管控不力引发的高关注度事件,面临诉讼、巨额罚款,甚至导致CIO被问责。

随着智能体掌握更多自主权,缺乏透明框架和审计机制的企业将面临巨大的法律与声誉风险。

这些预测共同揭示的本质变化

IDC FutureScape 2026 反复强调一个核心结论:

智能体改变的不是某个流程,而是企业如何运行、如何决策、如何承担责任。当智能体能够自主执行任务、协调流程并影响结果,企业必须重新思考数据、架构、治理、组织和领导力的边界。

IDC 中国研究经理 孙振亚表示,中国企业正在从生成式AI 阶段迈入智能体阶段的关键窗口期,但这个过程并非是简单的技术升级,而是一项系统化的工程。FutureScape 2026 显示,智能体的规模化落地必须要有AI 就绪的数据底座、多智能体的编排平台以及完善的治理机制。对于缺乏这些关键要素的企业而言,智能体带来的可能不是机遇,而是效率与合规层面的重大风险源。

一个面向企业领导层的扩展性建议

IDC 认为,智能体并不是一项可以逐步叠加的技术能力,而是一种会持续放大组织既有优势与短板的系统性力量。当智能体开始承担决策、执行与协调角色,企业原有的数据质量、流程设计、治理成熟度以及组织协同能力,都会被迅速放大并体现在结果层面。

因此,企业不应将智能体视为单一技术投资,而应将其纳入企业运行模式的长期演进路径来规划。这意味着:

  • 在技术层面,必须优先夯实 AI 就绪数据、智能体编排与可观测性能力,而非堆叠模型或工具;
  • 在治理层面,需将人工监督、责任边界和可追溯性制度化,而不是事后补救;
  • 在组织层面,需同步重构岗位角色、能力模型与决策流程,使人机协作成为默认工作方式;
  • 在管理层面,高管团队需要形成对智能体的共同认知,平衡效率和安全,把如何有效治理也纳入战略考量,而非单纯追求速度。

那些能够在规模化之前就完成这些准备的企业,更有可能把智能体转化为持续生产力;反之,智能体的能力越强,潜在风险也会被放大得越快。


行动指南:企业推进智能体的现实起点

结合 FutureScape 2026 的十大预测,IDC 建议企业在未来 12–24 个月内,优先从以下几个方面入手,逐步构建可持续的智能体能力:

第一,先解决基础数据问题
在引入或扩展智能体之前,对关键业务场景开展 AI 就绪数据评估,重点关注数据的完整性、语义一致性、上下文关联能力以及可追溯性。没有高质量数据,智能体带来的将更多是返工与人工干预,而非自动化红利。

第二,从高价值、低歧义的流程切入
优先选择目标清晰、结果可衡量、决策歧义较小的流程作为智能体的落地点,例如客户服务分流、内部运营协调或标准化审批支持,而非一开始就覆盖高度复杂或高风险场景。

第三,把治理与监督嵌入设计之初
在智能体架构设计阶段即明确人工介入点、升级路径与审计机制,确保所有自主决策都具备可解释性与回溯能力,而不是等到智能体进入关键流程后再补治理。

第四,建立跨职能的智能体管理机制
将 IT、数据、业务、合规与人力资源纳入同一治理框架,避免智能体成为某个部门的工具。在多智能体(Multi-Agent)场景下,统一编排、权限与责任归属尤为关键。

第五,为岗位与能力转型预留空间
提前识别哪些岗位将与 智能体深度协作,哪些能力需要被重塑,并通过培训、试点和角色演进,帮助员工适应新的工作方式。智能体的成功,很大程度上取决于组织是否准备好与智能体共事。

通过以上路径,企业可以在控制风险的前提下,逐步释放智能体的规模化价值,避免陷入技术领先但组织滞后的常见陷阱。

如需进一步了解与研究相关内容或咨询 IDC其他相关研究,请点击此处与我们联系。

Zhenya Sun - Research Manager - IDC

Zhenya Sun is a research manager for the IDC team focused on exploring the application of technology and industrial development of AI and AI agents. He is also responsible for providing clients with consulting services on technologies, products, and markets related to large language models (LLMs) and AI agents, as well as delivering speeches at industry conferences and internal seminars. Before joining IDC, Zhenya served as a project management officer (PMO), responsible for internal and external strategic consulting, AI application research and advisory services, AI project framework standardization, management system construction, and technical training on AI applications. Prior to that, he also led initiatives in product development process optimization and user market analysis. Zhenya holds a Master's Degree in Engineering Management with a specialization in Information Systems Engineering from the University of the Chinese Academy of Sciences.

Government organizations across Asia/Pacific are entering a defining phase in their digital evolution. Economic constraints, heightened citizen expectations, talent shortages, and tightening regulatory mandates are converging just as digital systems shift from automation to autonomous orchestration. For government technology leaders, this is no longer about adopting another digital tool. It is about preparing institutions for agentic AI and the operating models required to use it responsibly.

What is Agentic AI and Why it Matters for the Government

Agentic AI represents a step beyond analytical or recommendation-based systems. These systems can interpret intent, plan tasks, and execute actions within policy-defined boundaries. They navigate across systems, channels, and agencies, coordinating activities that previously relied on manual intervention, casework, or administrative adjudication. In a climate where governments are expected to deliver more with fewer resources, agentic AI offers a path to fundamentally reshape how public services are delivered and managed.

Why Data Readiness is the Real Barrier to Agentic AI

This shift is already influencing investment priorities. According to IDC FutureScape: Worldwide National Government 2026 Predictions: Asia/Pacific (Excluding Japan) Implications research, in 2026, 40% of national governments in Asia/Pacific excluding Japan (APeJ) will invest 10% of their IT budget in data architecture and governance solutions to address gaps that are preventing them from realizing the benefits of agentic AI. This signals a clear recognition that data readiness —not algorithms—is now the primary barrier to scaling autonomy.

IDC survey data reinforces this outlook. While many government agencies are exploring agent-driven workflows, relatively few have moved beyond pilots. The primary barriers are not technical ambition but gaps in data quality, system integration, and oversight models. As a result, national administrations across Asia/Pacific are increasing allocations toward data management, interoperability, and governance, acknowledging that agentic AI readiness depends more on institutional foundations than on model sophistication.

Agentic AI systems require structured, traceable, and interoperable data to reason and act responsibly across high-stakes domains such as benefits administration, taxation, compliance, emergency response, and infrastructure operations. Without strong data foundations and clear policy constructs, autonomy introduces operational, regulatory, and trust risks rather than value. For government leaders, data architecture and governance are becoming strategic prerequisites for agentic AI, not supporting functions.

Strategic Forces Shaping Government Agentic Adoption

Several macro-level forces are shaping the pace and direction of agentic AI adoption in government:

  • Budgetary pressure: Fiscal constraints persist even as demand for digital service continues to expand.
  • Sovereignty and compliance: Requirements around data residency, algorithmic transparency, and accountability are tightening.
  • Workforce disruption: Structural skill gaps in cybersecurity, data engineering, compliance engineering, and MLOps remain unsolved.
  • Citizen Expectations: Citizens increasingly expect faster, more personalized, and more equitable services, influenced by private-sector experiences.

IDC data shows these forces converging as agentic AI moves from conceptual exploration toward early operational pilots. Government leaders increasingly see agentic capabilities as tools for accelerating workflows, improve decision support, and enhance service quality. However, integration, governance, and compliance remain the primary obstacles to scaling beyond pilots. Without deliberate management, these crosscurrents risk fragmented investments and new digital silos. Addressed strategically, they can accelerate modernization while reinforcing public trust.

How Agentic AI Transforms  Government Functions

Agentic AI opens up new opportunities across three core government domains:

1. Operational orchestration – Agent-driven systems can coordinate multi-step workflows that span multiple agencies or departments, reducing handoffs and administrative lag. This is particularly valuable in benefits processing, regulatory inspections, tax compliance, procurement, licensing, and infrastructure operations, where complexity and interdependence are high. IDC surveys show that a growing share of Asia/Pacific government agencies are prioritizing orchestration capabilities over standalone task automation, marking a shift in architectural strategy.

2. Citizen service delivery – Agentic AI capabilities enable proactive, context-aware, and personalized interactions. Agents can identify citizen needs, trigger workflows, prompt follow-ups, and escalate cases to human supervisors when required. This directly supports government priorities around service equity, responsiveness, and improved case resolution outcomes.

3. Decision support for policy and planning – Agentic systems can synthesize data, model scenarios, and present options for policymakers during crises, planning cycles, or resource allocation exercises. This does not replace human authority; it expands the analytical capacity available to decision-makers when time and complexity are constraints. Across all three domains, trust is the central requirement. Agentic systems deliver sustainable value only when paired with auditability, human oversight, and transparent accountability structures. Without these safeguards, autonomy becomes a liability —especially in regulated or politically sensitive environments.

 What Government Technology Buyers Must Do Now

For CIOs, CTOs, Chief Digital Officers, and procurement leaders, the transition to agentic AI raises several practical considerations: Institutional readiness is the first barrier.

Many agencies continue to rely on siloed legacy systems, inconsistent data definitions, and limited interoperability. Agentic AI amplifies these weaknesses. Without mature integration, clean data, and consistent metadata, autonomy is either unsafe or impractical.

Governance must be built into the workflow.

Because agentic systems act rather than merely recommend, governments must design for traceability, audit trails, human-in-loop controls, and clear escalation paths from the outset. Policy and sovereignty alignment cannot be retrofitted after deployment.

Operating models and workforce must evolve.

Agentic AI reshapes work patterns rather than simply reducing labor. While agencies currently rely heavily on external system integrators and cloud providers, new internal roles in agent orchestration, compliance engineering, and lifecycle management will become essential over time.

The message for technology buyers is clear: agentic AI is not merely a technology decision. It is an institutional capability decision.

Procurement and Vendor Evaluation for Agentic AI

As governments move beyond proofs of concept, procurement teams must distinguish between true agentic platforms and offerings that simulate autonomy through scripted automation or interfaces. IDC recommends evaluating vendors against criteria such as:

  • Orchestration of multi-step, cross-system workflows
  • Integration and interoperability across legacy and multi-cloud environments
  • Auditability, explainability and documentation
  • Alignment with sovereignty and policy mandates
  • Support for open standards and architectural portability
  • Clear responsibility models across the autonomy lifecycle.

Governments that structure RFx around interoperability, auditability, and policy alignment will be better positioned to deploy agentic capabilities responsibly without increasing regulatory or operational risk.

The Leadership Mandate for Agentic AI in Government

Agentic AI is no longer distant. It is a leadership mandate. As economic pressure, regulatory expectations, workforce disruption, and citizen demands intersect, government leaders must move beyond isolated pilots toward responsible orchestration at scale.

That mandate requires alignment across strategy, data foundations, governance, and operating models. Agencies that establish these foundations will translate agentic AI into resilience, accountability, and measurable public value. Those that do not will remain locked in pilot mode—unable to scale autonomy without unacceptable risk.

Register now for the live webinar on 24 February 2025 at 1:30 pm SGT to join IDC in charting the agentic future with confidence.

Ravikant Sharma - Research Director, Government Insights for Asia/Pacific - IDC

Ravi Kant Sharma serves as the Research Director for Government Insights for Asia/Pacific (excluding Japan and China) (APEJC) at IDC’s office in Bangalore, India. He is tasked with guiding public sector agencies and the tech industry by collaborating with them in strategic planning. He has led numerous complex consulting engagements, represented IDC at industry events, and hosted workshops focusing on understanding how digital investments impact key economies in Asia/Pacific within the context of global trends — all while collaborating with IDC’s global Government Insights teams.

Asia/Pacific enterprises are entering a new era of cybersecurity defined by the convergence of human expertise, autonomous AI agents, and trust frameworks. IDC calls this the Cyber Trinity, a security model that integrates human judgment, autonomous AI agents, and embedded trust frameworks. Drawing on IDC FutureScape: Worldwide Security and Trust 2026 Predictions – Asia/Pacific Excluding Japan (Implications), this analysis examines how AI-driven SOCs, embedded AI governance, synthetic identity threats, sovereign AI requirements, and quantum-era risks are reshaping security strategies across the region.

As organizations accelerate toward AI-first operating models, security and trust are no longer reactive controls. They are now engineered, governed, and continuously validated capabilities that determine enterprise resilience, regulatory compliance, and long-term competitiveness.

Why security and trust are being redefined in Asia/Pacific

The security landscape in Asia Pacific excluding Japan (APeJ) is undergoing rapid change. IDC forecasts that total security spending in the region will reach US$39.5 Billion in 2026, growing at a 10% CAGR to US$52.4 billion by 2029. This growth reflects more than rising threat volumes. It signals a structural shift in how organizations must build and govern trust in an AI-driven world.

As enterprises adopt agentic AI, face fragmented regulatory requirements, and contend with sophisticated adversaries using AI-powered techniques, traditional security models are proving insufficient. Trust, once implicit, must now be engineered, governed, and continuously validated.

Five security and trust shifts shaping 2026

IDC’s analysis points to five major shifts that will define security and trust strategies across Asia/Pacific over the next 18–24 months.

1. Autonomous, AI-driven security operations

Security Operations Centers (SOCs) are evolving from human-centric environments to AI-augmented and increasingly autonomous operations. AI agents are deployed to triage alerts, reduce false positives, normalize incident response, and orchestrate remediation at machine speed. IDC’s Asia Pacific Security Study 2025 states that 39% of enterprises plan to apply AI/GenAI solutions in the next 12 months to optimize threat detection and analysis capabilities. This shift is essential as skills shortages and exploding telemetry volumes overwhelm traditional SOC models.

2. Embedded AI governance and sovereign AI requirements

Governments across Asia/Pacific are tightening controls on data usage and AI systems. Only 7% of enterprises are highly prepared in terms of GRC skills to support these new requirements, driving demand for privacy-by-design, compliance-by-design, and sovereign AI architectures Enterprises are reassessing cloud strategies, adopting retrieval-augmented generation (RAG), and exploring private compute environments to meet data residency and regulatory requirements while scaling AI responsibly.

3. Synthetic identity as a core trust threat

According to IDC’s 2025 Future Enterprise Resiliency & Spending (FERS) study, 49% of APeJ enterprises have paid at least US$10,000 in ransom due to ransomware breaches. Adversaries are using AI to create synthetic identities that blend real and fabricated data, undermining authentication systems across financial services, e-commerce, and government platforms. These attacks erode digital trust at scale, forcing organizations to modernize identity protection and adopt AI-powered anomaly detection to distinguish legitimate users from synthetic fraud.

4. Quantum readiness and cyber risk quantification

As quantum computing advances, enterprises are beginning to assess the long-term viability of existing cryptographic systems. Crypto-agility and quantum readiness are emerging as strategic imperatives. By 2028, IDC predicts that 20% of Asia’s top 2000 enterprises will engage cybersecurity professional services firms to conduct quantum risk assessments. The ability to quantify cyber risk in financial terms is also becoming a board-level requirement, shaping budgets, insurance strategies, and M&A decisions.

5. Dynamic playbooks and endpoint-level trust

Static security playbooks are giving way to dynamic, AI-generated response models that adapt in real time to evolving threats. 30% of enterprises will be prioritizing the expansion of its MDR capabilities across assets, endpoints and applications. The rise of deepfakes and AI-enabled deception is also accelerating demand for endpoint detection capabilities that balance privacy, performance, and resilience.

What Cyber Trinity means for enterprise leaders

Together, these shifts signal a fundamental change: security and trust are no longer reactive controls. They are strategic foundations for innovation. Organizations that succeed will be those that can:

  • Balance human oversight with autonomous AI decision-making
  • Embed governance directly into AI and security architectures
  • Treat trust as a measurable, managed asset
  • Anticipate regulatory and technological disruption rather than respond after the fact

From Insight to Action

These themes form the foundation of IDC’s FutureScape 2026 Security & Trust Predictions for Asia/Pacific, which will be explored in depth by IDC analysts Sakshi Grover and Yih Khai Wong in an upcoming webinar. The discussion will focus on how organizations can architect, govern, and operationalize the Cyber Trinity to strengthen resilience and lead with confidence in an autonomous security landscape. Register now.

About the Authors

Sakshi Grover - Senior Research Manager - IDC

Sakshi Grover is a senior research manager for IDC Asia/Pacific Cybersecurity Services, supporting its research and client engagement activities across Asia/Pacific markets. Additionally, she serves as the lead security analyst for IDC India. Sakshi is responsible for delivering syndicated custom research and consulting engagements on next-generation emerging and disruptive technologies. Her tasks include developing and socializing IDC's point of view within security services, covering both legacy and modern cybersecurity technologies. Her role involves close collaboration with technology vendors and buyers, developing market insights, and providing research, consulting, and advisory services in the fields of security software and services. This includes partnering on research efforts with relevant country analysts in the local IDC offices. Sakshi's views on security have been quoted in numerous publications, such as the Economic Times, Business Standard, Data Quest, CRN, and others.

Yih Khai Wong - Senior Research Manager - IDC

Yih Khai Wong is a senior research manager for IDC Asia/Pacific's Cybersecurity practice, supporting cybersecurity research and client engagements through the Asia/Pacific Security Opportunities: Trust and Resilience program. Yih Khai's area of focus is on security technologies, including cloud-native application protection, identity, endpoint and network security. He works closely with technology vendors and buyers, delivering actionable market insights and advice within the cybersecurity ecosystem. Before rejoining IDC, Yih Khai was a principal analyst covering the cloud, datacenter, and edge computing market with ABI Research. Prior to that, Yih Khai was in EY, in his capacity as an assistant director at EY's research and insights group. Yih Khai started his analyst career with IDC Malaysia as an analyst covering the enterprise applications market.

一个正在发生的变化软件不再只是人写的

在过去二十年里,软件工程的核心始终围绕“人如何写代码、交付系统”展开。即便进入 DevOps 时代,自动化更多也只是加快了既有流程。但 IDC 指出,随着 Agentic AI 的成熟,软件开发正在发生一次结构性转变:开发不再完全由人主导执行,而是由人类开发者与自主 AI 智能体协作完成。

在《IDC FutureScape:全球开发者和 DevOps 2026 年预测——中国启示》(Doc# CHC54059126

,2026年1月)中,IDC 明确提出:未来五年,Agentic AI 将深度嵌入从开发、测试到运维和安全的整个生命周期,迫使 DevOps 从“工具链升级”走向“运行模式重构”。

IDC 的核心洞察:DevOps 的问题,已经不只是效率

在中国市场,许多企业仍将 DevOps 视为提升交付速度、降低沟通成本的方法。但 IDC 认为,这种理解正在失效。
当 AI 智能体开始自动生成代码、执行测试、修复缺陷并参与决策,真正的挑战不再是“怎么用工具”,而是:

  • 谁来管理和监督智能体?
  • 如何保证 Agent 的行为可解释、可审计?
  • 人类开发者的角色将如何转型?
  • 企业是否具备规模化运行智能体的治理与平台能力?

这些问题,正是 FutureScape 2026 十大预测试图回答的核心。

十大预测:Agentic AI 将如何重塑开发者与 DevOps 生态

预测 1|智能体开发采用

到 2028 年,面对智能体部署量增长 10 倍的局面,50% 的中国 1000 强企业将采用智能体开发生命周期,以实现企业级智能体 AI 的有效规模化落地。

这意味着,传统 SDLC 已不足以支撑智能体开发,企业必须引入专门面向 Agent 的开发与治理方法论。

预测 2|多智能体编排

到 2029 年,多智能体编排的风险与复杂性将促使企业强化战略布局、扩充卓越中心(COE)资源,并将 AI 治理与监控工具的支出增加 30%。

当单一 Agent 变成 Agent 集群,治理与可见性将成为规模化落地的前提。

预测 3|自主式智能体 AI 工作单元

到 2030 年,80% 的开发者将与自主 AI 智能体展开协作,推动人类开发者向规划、设计与编排角色转型,并重塑开发者工具生态系统。

开发者将不再只是“写代码的人”,而是“引导和监督智能体的人”。

预测 4|氛围编程采用

到 2027 年,随着企业级能力的成熟,35% 的专业开发者将采用氛围编程开发平台构建生产级应用。

自然语言正在成为新的开发接口,但前提是企业级治理与质量控制能力同步成熟。

预测 5|嵌入 DevOps 的智能体应用

到 2030 年,65% 的企业将把 AI 智能体嵌入 DevOps 和 DevSecOps 流水线,用于执行开发与安全工作流。

Agent 将成为流水线中的“常驻成员”,而非外部插件。

预测 6|前沿模型采用

到 2027 年,在开发者偏好的驱动下,70% 的 AI 用例将仅由少数几个前沿模型提供支持。

模型选择正在从“多而杂”走向“少而精”。

预测 7|智能体 AI 项目失败

到 2028 年,70% 的“自建型”智能体 AI 项目将因未能达成投资回报率目标而被放弃。

低估治理、运维和组织成本,是失败的主要原因。

预测 8AI 质量保障扩展

到 2028 年,AI 质量保障将推动智能体测试和跨应用生命周期管理的采用率至少提升 30%。

没有质量保障的 Agentic DevOps,无法进入生产核心。

预测 9AI 加速应用开发

到 2029 年,通过使用智能体 AI 软件开发工具,企业的应用开发与现代化迭代速度将提升 400%。

速度跃迁的前提,是平台化与治理并行。

预测 10|开发者模型微调

到 2027 年,微调将取代检索增强生成(RAG)成为大语言模型改造的主流模式,这将推动开发者对开源权重模型的使用率提升 80%。

模型工程正在走向更深度的定制化。

分析师观点

IDC 中国研究经理王彦翔认为,开发者和 DevOps 正站在从“自动化时代”迈向“智能体时代”的关键门槛。FutureScape 2026 显示,真正拉开差距的,不是是否引入 AI 编码工具,而是企业是否具备平台工程、治理能力和开发者角色转型的整体规划。那些仅在局部场景试点智能体的组织,将很难释放规模化价值;而将 Agentic AI 作为企业级能力来建设的组织,更有可能在速度、质量和创新能力上形成长期优势。

一个面向技术与业务领导者的综合建议

IDC 并不建议企业急于“全面智能体化”。更重要的是,以 DevOps 为核心,系统性重构开发流程、平台能力与治理机制:建立智能体开发生命周期(ADLC)、强化多智能体编排与监控、同步推进开发者技能转型,并将 AI 治理嵌入每一个交付环节。


只有这样,Agentic AI 才能成为持续创新的引擎,而不是新的技术债务来源。

行动指南:企业可以从哪里开始?

  • 从 高价值、低风险的开发或运维场景 切入,验证 Agent 的实际收益
  • 建立 跨职能的 AI / Agent 卓越中心(COE),统一治理与平台策略
  • 投资 平台工程与 AI 质量保障,而不仅是开发工具
  • 提前规划 开发者角色与能力转型,为人机协作做好准备

如需进一步了解与研究相关内容或咨询 IDC其他相关研究,请点击此处与我们联系。

Bryan Wang - Senior Market Analyst - IDC

Bryan Wang is a senior market analyst for Cloud Computing in the Emerging Technology sector for IDC China. He focuses on research and analysis of China's cloud computing market, including infrastructure as a service (IaaS), platform as a service (PaaS), and software as a service (SAAS). Bryan is also responsible for providing market analysis and research in relevant fields together with IDC's regional and global research teams. Before joining IDC, Bryan worked as a cloud computing solution architect for well-known manufacturers and systems integrators. He was mainly responsible for presales consulting, project design, industry insight, project management, and other work. He has rich experience and a profound understanding of the cloud computing field. Bryan graduated with a B.A. in Inorganic Nonmetallic Materials Engineering from Central South University.

企业连接,正在从基础设施演进为战略能力

在 AI 工作负载快速增长、业务连续性要求不断提高的背景下,企业连接已不再只是网络团队的技术议题,而正在成为影响 业务韧性、运营效率与创新速度 的核心能力。
IDC 认为,企业正在进入一个全新的连接阶段:连接不再只是“管道”,而是由 AI 驱动编排、可感知并持续演进的数字底座。

为什么这份 FutureScape,对企业连接战略具有参考价值?

《IDC FutureScape:全球企业连接2026年预测——中国启示》(Doc#CHC52329725,2025年12月)中,IDC 指出,随着 AI、Agentic AI 与边缘计算深度融入企业运营,连接能力正在被重新定义。
企业对连接的期望,已从“稳定可用”升级为 敏捷、自治、安全、面向 AI 的能力体系。这一转变,使网络与连接成为 AI 规模化落地不可或缺的前提条件。

IDC FutureScape 给出的十大关键预测:

预测1AI 重塑云通信

到2027年,50%的企业将部署由 Agentic 人工智能(AI)驱动的云通信 API,从而以更高水平的个性化和自动化重塑通信与协作的使用方式。
要点:个性化和智能化化重塑通信与协作。

预测2AI 赋能的数字全连接底座(LEO 卫星)

到2029年,50%的企业将采用低轨道(LEO)卫星连接来补充地面网络,将关键的卫星直连消费者(D2C)、直连终端(D2D)以及高速宽带纳入统一的数字全连接底座。
要点:连接韧性开始向“天地一体”扩展。

预测3|一体化蜂窝物联网

到2027年,60%的企业将利用蜂窝物联网应用、多 SIM 卡与嵌入式SIM卡(eSIM)方案,以及窄带物联网(NB-IoT)和 5G,构建面向关键业务场景的泛在连接网络。
要点:蜂窝物联网连接正在成为业务规模化的基础。

预测4|无线广域网(WLAN)加速扩张

到2027年,80%的企业将全网集成 AI 驱动的无线广域网(wireless WAN),以提供可扩展、安全且具备高弹性的园区和分支机构连接。
要点:AI 正在重塑无线广域网的运维与自治能力。

预测5|超大规模与网络平台(Cloud WAN

到2028年,50%的“云优先”企业将为其 AI 工作负载采用云广域网(cloud WAN),强化云服务提供商在网络中的角色。
要点:网络能力正加速平台化、云化。

预测6|边缘侧推理部署

到2028年,50%的企业将把推理类用例部署在边缘侧,以驱动新增收入、改善客户体验和 / 或优化内部流程。
要点:AI 推理开始向连接边缘迁移。

预测7|零信任网络架构

到2027年,仍有30%的企业在安全访问服务边缘(SASE)的实施上保持碎片化策略,在其 SD-WAN 部署向零信任网络演进的过程中逐步调整。
要点:零信任是方向,但路径并不一致。

预测8|虚拟 AI 网络工程师

到2027年,Agentic AI 将在不显著扩张人力规模的前提下,使网络团队的有效人效实现近乎翻番。
要点:网络团队正在被“数字员工”增强。

预测9AI 重塑专业服务

到2027年,AI / 生成式 AI 以及 Agentic AI 将全面融入咨询与集成服务,使服务交付能力提升25%,并将设计与配置时间缩短60%。
要点:网络专业服务进入 AI 驱动时代。

预测10AI 保障数据合规与可信(ESG 连接)

到2029年,围绕 ESG 强制性要求的提升,将导致仅有40%的企业会主动投资用于遥测数据采集的网络连接。
要点:连接与 ESG 的关系更加务实与现实。

这些预测对企业意味着什么?

IDC FutureScape 2026 表明,企业连接正在经历一次角色升级:从支撑 IT 运行的基础设施,转向 支撑 AI、业务连续性与组织敏捷性的战略平台。企业若仍以传统网络视角规划连接,将难以支撑 AI 推理、边缘自治与实时决策的需求。

IDC中国助理研究总监崔凯表示,未来五年,领先企业将把连接视为“AI 原生能力”的组成部分,通过 AI 驱动的云通信、无线 WAN、边缘推理与多网络融合,构建可持续演进的数字连接底座。FutureScape 2026 显示,只有将连接、AI 与安全统一规划的企业,才能在复杂环境中保持韧性与创新速度。

给企业管理层的近期行动建议

  • 将 AI 工作负载需求系统性映射到连接与网络规划中
  • 评估无线 WAN、Cloud WAN 与卫星连接的组合策略
  • 在网络运维中引入 Agentic AI 与 AIOps
  • 将边缘推理与数据主权纳入连接架构设计
  • 与服务提供商建立长期、平台化合作关系

未来 12–24 个月值得关注的信号

  • AI 驱动连接在园区与分支机构的规模化落地
  • 边缘推理对带宽与时延需求的重塑
  • 网络团队角色从“运维者”向“平台工程师”转变

进一步推荐

如需进一步了解与研究相关内容或咨询 IDC其他相关研究,请点击此处与我们联系。

Kai Cui - Associate Research Director - IDC

Kai Cui is an associate research director for IDC China's Telecommunications and Internet of Things (IoT) Group. His research covers telecommunications, enterprise communication, and IoT industries. He is responsible for tracking and analyzing relevant areas as well as providing research and consulting services based on customized requests. Kai has more than 15 years of experience in the communications and telecom industry. Prior to joining IDC, Kai worked at Polycom, Huawei, and other communication enterprises, where he engaged in technical support, project management, and solution planning. He has an in-depth knowledge of deployment and application of communications solutions in vertical industries. Kai graduated from the Beijing Union University in 2001, with a bachelor's degree in Computer Science.

制造业,正站在从自动化迈向自主化的门槛

制造业不确定性持续上升、产品复杂度与柔性需求并行增长的背景下,中国制造企业正面临一次关键抉择:是继续以点状自动化与局部优化应对变化,还是系统性迈向以数据、模型与智能体驱动的自主化运营。IDC 认为,这一选择将直接决定企业未来五年的运营韧性、创新效率与全球竞争力。

IDC 认为,中国制造业正在进入一个新的关键阶段:AI 不再只是提升局部效率的工具,而是推动生产系统向“自主化运营”演进的核心引擎。

为什么这份 FutureScape,对中国制造业尤为关键

《IDC FutureScape:全球制造业2026年预测——中国启示》(Doc# CHC52915025,2025年12月)中,IDC 指出,中国制造业未来五年的智能化升级,将不再以单点应用为主,而是围绕自动化平台开放化、AI 驱动的计划与控制、OT 数据智能整合、人机协同、工业安全以及混合云与智能体治理等关键能力展开。
这些能力的成熟度,将直接决定制造企业能否从“数字化工厂”跨越到“自主化工厂”,并在全球竞争中建立可持续优势。

IDC FutureScape 给出的十大关键预测:

预测1|软件定义工厂

受自主化运营的潜力驱动,到2029年,将有30%的中国工厂通过开放、虚拟化、软件定义的自动化平台,在中央统一配置和管理自动化控制系统。
要点:自动化控制正在从封闭专有系统走向开放、可编排的平台化形态。

预测2AI APS

到2026年,超过40%已部署 APS 的中国制造商将升级为 AI 赋能的 APS,从而开始实现自主化流程。
要点:生产计划与排程正从“人主导”迈向“持续自优化”。

预测3IT/OT 融合 Agent

到2027年,随着标准化水平提升以及面向特定数据类型的 AI 智能体(AI Agents)广泛应用,40%的 OT 数据将能够在应用与平台之间实现自主集成。
要点:智能体正在改变工业数据工程的效率边界。

预测4AI 后服务

为打通设计与服务之间的闭环,到2027年底,25%的中国头部 OEM 及售后服务企业将利用 AI 连接现场与工程数据,从而提升产品与服务质量。
要点:产品创新开始真正进入全生命周期闭环。

预测5|可预测工业数据安全

为应对数据模型安全风险,到2029年,60%的大型制造企业将采用 AI 驱动的 OT 网络防御系统,将威胁检测时间缩短60%。
要点:工业 AI 安全正从“被动防御”转向“预测与自主响应”。

预测6|人机技能互学

到2028年,未能建立人机技能闭环的中国企业,将面临比同行高出20%的停机和再培训成本,其生产效率也将明显低于已实施双向培训机制的企业。
要点:人机协同能力成为生产韧性的核心变量。

预测7|设计仿真 Agent

到2028年,65%的中国头部制造企业将在设计与仿真工具中结合 AI 智能体(AI Agents),以持续验证设计更改、配置与变体是否符合产品要求。
要点:仿真与工程决策正在加速智能化。

预测8AI 员工培训

到2028年,超过30%的中国头部制造企业将采用 AI 驱动的知识管理工具,对员工进行再培训和技能升级,并促进产业生态内的协作共享。
要点:知识数字化成为应对劳动力波动的关键手段。

预测9|复合式工业 AI

到2030年,70%的中国头部制造企业将借助 AI 智能体(AI Agents)构建数据模型并管理混合云工作负载,从而将质量成本降低2%。
要点:混合云与多智能体协同成为工业 AI 的主流架构。

预测10|工业模型管理

到2029年,40%的中国头部制造企业将依托超大规模云生态,构建、部署并扩展新一代 AI 解决方案,加速数字化转型进程。
要点:行业云与模型生态将重塑制造软件格局。

这些预测对制造企业意味着什么

IDC FutureScape 2026 清晰表明,中国制造业的竞争焦点正在发生结构性转移:
从单点自动化,转向系统级自主化;从经验驱动,转向数据与模型驱动;从局部优化,转向跨设计、生产与服务的全局协同。无法建立这些基础能力的企业,即便部署了 AI,也难以真正释放规模化价值。

IDC中国高级研究经理杜雁泽表示,中国制造业正在跨越“数字化到智能化”的关键拐点。FutureScape 2026 显示,领先企业正通过开放式自动化平台、AI 智能体和混合云架构,将 AI 深度嵌入生产控制、工程决策与知识传承之中,从而构建可持续、自主演进的运营体系;而仍停留在封闭系统与点状应用阶段的企业,将在效率、韧性和创新速度上持续承压。

给制造企业管理层的近期行动建议:

  • 评估现有自动化系统向开放、软件定义架构演进的可行性
  • 以 APS、质量与能耗等高价值场景为切入口推进 AI 自主化
  • 建立 OT 数据标准化与智能体协同的数据治理基础
  • 将 AI 安全与模型治理纳入工业网络安全战略
  • 投资人机技能闭环与 AI 知识平台,提升组织韧性

未来 12–24 个月需要重点关注的信号

  • 软件定义自动化在中国头部工厂的规模化进展
  • AI 智能体在工业数据工程与仿真中的成熟度
  • 工业 AI 安全从“合规导向”向“风险导向”的转变

进一步推荐

如需进一步了解与研究相关内容或咨询 IDC其他相关研究,请点击此处与我们联系。

Yanze Du - Senior Research Manager - IDC

Yanze Du is a senior research manager for the Manufacturing Insights group in IDC China, and he is responsible for conducting research on and analysis of the China manufacturing industry and supply chain. He is also involved in regional and global consulting, and business development in related markets. Prior to joining IDC, Yanze had an in-depth working experience in the digital transformation field and had wide exposure to various businesses in the manufacturing industry including smart manufacturing and industry software. These experiences gave him a deep understanding of both the status quo and future trends of the market. Yanze graduated from Tianjin University with a bachelor's degree in Computer Science.

Press Releases

Stay informed with our latest research findings, market insights, and breakthrough technology analysis.