根据 Dataiku 的一项调查，大多数首席信息官（CIO）无法全面了解员工在未经批准的系统中开发的人工智能（AI）应用程序。在接受调查的 685 名 CIO 中，有 81% 表示他们无法完全监控那些通过非正式渠道创建的 AI 应用程序。
Most chief information officers cannot fully see the AI agents their staff build outside approved systems, a survey by Dataiku found. 81% of the 685 CIOs polled said they lack complete oversight of AI agents created outside formal channels.
这项调查由 Harris Poll 公司于 7 月 9 日至 29 日期间为 Dataiku 在线进行，调查对象涵盖了来自美国、英国、法国、德国、阿联酋、日本、韩国和新加坡的 CIO。Dataiku 是一家总部位于巴黎的公司，目前总部设在纽约。该公司在年度会议 “Succeed” 上发布了《全球 AI 现状报告：CIO 版》。
The Harris Poll ran the survey online for Dataiku from 9 to 29 July. It covered CIOs in the US, the UK, France, Germany, the UAE, Japan, South Korea and Singapore. Dataiku, a Paris-founded company now based in New York, released the Global AI Confessions Report: CIO Edition on Thursday at Succeed, its annual conference.
虽然 90% 的 CIO 表示能够全面追踪自己开发的所有 AI 应用程序，但仍有 72% 的 CIO 表示无法持续确认这些应用程序是否实现了预期的业务目标。另有 67% 的 CIO 估计公司目前有 51 个或更多 AI 应用程序正在实际运行中；83% 的 CIO 表示整个组织内缺乏统一的 AI 应用程序生命周期管理机制；47% 的 CIO 表示今年已经停用了 20 个以上的 AI 应用程序；仅有 21% 的 CIO 能够实时了解各团队或具体使用场景下的 AI 成本情况。
Tracking agents, but not their results 90% of CIOs said they have complete tracking of all their agents. Yet 72% said they cannot consistently confirm whether those agents deliver the business outcomes they were built for. 67% estimate that 51 or more agents are running in production. 83% lack standardised lifecycle management for agents across the organisation. 47% have already decommissioned more than 20 agents this year. Only 21% have full, near-real-time visibility into AI costs by team or use case.
84% 的 CIO 认为，员工开发 AI 应用程序的速度远远超过了 IT 部门的管理能力；60% 的 CIO 表示他们的应用程序、工具和 IT 环境中缺乏统一的 AI 管理机制。
67% estimate that 51 or more agents are running in production. 83% lack standardised lifecycle management for agents across the organisation. 47% have already decommissioned more than 20 agents this year. Only 21% have full, near-real-time visibility into AI costs by team or use case.
关于出现问题的 AI 应用程序应由谁负责，CIO们的意见存在分歧：23% 的 CIO 认为应由相关团队负责，21% 认为应由中央 IT 部门负责，20% 认为应由数据或 AI 团队负责，18% 认为应由安全、风险控制及合规部门负责。
84% agree that employees are creating agents and applications faster than IT can govern them. 60% lack a central AI governance layer across their apps, tools and IT environments.
Dataiku 的联合创始人兼首席执行官 Florian Douetteau 表示：“监控只能说明某个 AI 应用程序正在运行中，而管理则决定了它是否应该继续运行。目前，几乎没有人能够‘终止’一个正在运行的 AI 应用程序。”
CIOs also split on who owns an agent that goes wrong. Their answers were shared teams (23%), central IT (21%), the data or AI team (20%) and security, risk and compliance (18%).
AI 的成败直接关系到 CIO 的声誉和职业发展：88% 的 CIO 认为 AI 的成功与否会直接影响他们的职业前景；87% 的 CIO 表示他们的职位稳定性取决于 AI 的应用效果；76% 的 CIO 认为，如果公司在 2027 年底前无法从 AI 中获得任何可衡量的收益，他们的职位将面临风险。
“Monitoring tells you an agent is running. Managing tells you whether it’s earned the right to keep running, and right now, almost nobody can fire an agent,” said Florian Douetteau, co-founder and CEO of Dataiku. Jobs and budgets on the line 88% of CIOs said success with AI will shape their reputation or career. 87% said their CEO has told them their job security depends on AI outcomes. 76% expect their role to be at risk if their company shows no measurable gains from AI by the end of 2027.
72%的受访者表示，如果他们在2026年底未能实现人工智能（AI）项目的目标，他们的AI预算很可能会被削减或冻结。97%的受访者表示，公司管理层对AI项目的回报表现出了越来越大的压力（即要求AI项目必须产生实际的经济效益）。
Jobs and budgets on the line 88% of CIOs said success with AI will shape their reputation or career. 87% said their CEO has told them their job security depends on AI outcomes. 76% expect their role to be at risk if their company shows no measurable gains from AI by the end of 2027.
Dataiku的研究显示，美国企业的首席信息官（CIO）们的态度尤为坚决：94%的受访者表示，员工开发AI模型的速度远远超过了IT部门对这些模型的管理能力；87%的受访者预计如果项目目标未能实现，他们的预算将会被削减或冻结；82%的受访者对过去18个月内做出的某个重要的AI供应商或平台选择感到后悔。
72% expect a cut or freeze to their AI budget if they miss performance targets by the end of 2026. 97% report at least some rise in board pressure to show a return on AI. Dataiku said US CIOs posted some of the starkest numbers. 94% of them say employees build agents faster than IT can govern them. 87% expect budget cuts or a freeze if targets slip. 82% regret at least one major AI vendor or platform choice from the past 18 months.
在模型选择方面，75%的受访者计划使用更多或不同的AI模型；74%的受访者正在考虑使用开源模型或混合模型（即结合开源与商业化模型的解决方案），以应对某些模型可能无法继续使用的风险。
On models, 75% plan to use more or different ones. 74% are considering open-source or open-weight models as a hedge in case a model becomes unavailable.
Dataiku自身也提供了相应的软件解决方案来帮助企业管理AI项目。该公司在发布“Agent Management”工具的同时，还推出了新的产品功能，该工具能够将所有AI模型统一纳入公司的管理系统中，并与AWS Bedrock、Databricks、Google Vertex、Microsoft Copilot Studio、Azure Foundry、Salesforce Agentforce以及Snowflake Cortex等平台实现互联互通。该软件将于10月正式上市，定价方式为每年按每个实例计费，同时还会对每个AI模型提供监控服务。
Dataiku’s own product Dataiku sells software for this problem. On the same day it launched Agent Management, which pulls agents into one inventory. It connects to AWS Bedrock, Databricks, Google Vertex, Microsoft Copilot Studio and Azure Foundry, Salesforce Agentforce and Snowflake Cortex. It goes on general sale in October, priced per instance each year, with monitoring charged per agent.
另外，Gartner在2025年6月预测，到2027年底，超过40%的AI项目将会被取消。Gartner指出，导致这些项目被取消的原因包括成本上升、业务价值不明确以及风险控制措施不足。在亚马逊（Amazon）公司，大约90%的早期开发的AI原型最终未能投入实际生产使用。
Separately, Gartner predicted in June 2025 that over 40% of agentic AI projects will be cancelled by the end of 2027. It cited rising costs, unclear business value and inadequate risk controls. At Amazon, AWS says almost 90% of early agent prototypes never reached production.