甲骨文将代理式人工智能工具的重点聚焦于Fusion应用开发者。

内容来源:https://aibusiness.com/agentic-ai/oracle-fusion-app-developers-agentic-tools
内容总结:
甲骨文深化AI代理平台建设,推出面向开发者的融合应用新工具
【环球网科技综合报道】甲骨文公司于本周二发布了新的企业级功能,旨在帮助开发者为其Oracle Fusion应用程序构建、协调并治理AI代理。Fusion应用套件涵盖财务、供应链管理、制造、人力资源、销售、客户服务及营销等多个业务领域。
此次更新侧重于提供面向专业开发者的无代码与专业代码(传统人工编程)工具,这与此前于2025年3月推出的、主要面向业务用户的Oracle AI代理工作室(AI Agent Studio)中的编码能力形成互补。通过甲骨文所称的“AI原生构建器体验更新”,开发者可接入来自第三方厂商的流行AI编码模型与代理,包括OpenAI Codex、Claude Code、微软VS Code、Google Antigravity以及开源分布式版本控制系统Git。在Fusion套件中工作的开发者还可使用来自多家厂商的命令行界面。
甲骨文将利用这些工具创建的应用称为“Fusion代理应用”(Fusion Agentic Applications)。这类应用由专业化的代理系统组成,能够进行推理、协调并采取行动,在Fusion工作流程中执行任务。
甲骨文云应用开发高级副总裁Natalia Rachelson表示:“我们现在需要真正平等地对待每一位用户。无论你是想采用无代码方式,还是一名希望在专业环境中编码的专业开发者,我们都将提供工具和插件,让你能够留在熟悉的专业环境中,构建可在Fusion中运行的代理应用。”
不过,分析师对此持谨慎乐观态度。KramerERP创始人兼分析师Robert Kramer认为,面向开发者的新工具是对甲骨文代理平台的一次适度更新。“这不仅仅是又一个代理或开发工具,但在客户展示出实际成果之前,我不会将其描述为重大变革,”Kramer指出,“ERP是代理AI既能创造价值也伴随风险的领域。AI代理总结发票或建议催收行动相对容易,但要让代理在维护安全、审批和审计追踪的同时更新财务记录、释放订单、启动付款或更改供应计划则困难得多。”
Kramer补充说,该代理平台的价值不仅限于ERP,还延伸至人力资本管理、供应链管理和客户体验(CX)领域。在受治理的环境中跨更广泛的业务流程协调多个代理的能力至关重要。其他潜在优势包括加快代理应用投入生产的速度,以及使业务用户、开发者和合作伙伴能够利用一系列无代码、低代码和高级AI编码工具更轻松地协作。
但Kramer也提出了疑问:“这些工具在甲骨文之外的环境效果如何?企业如何管理多代理故障?业务用户在构建生产应用时应拥有多少控制权?现有的安全和审批机制有所帮助,但无法解决数据质量差、流程归属不清或责任不明的问题。甲骨文还需要在运行时成本、集成成本和可衡量的客户成果方面提供更多清晰度。”
甲骨文此次更新其代理AI平台之际,这家拥有49年历史的科技巨头正急剧转向人工智能,立志成为向生成式AI实验室OpenAI等客户提供AI数据中心算力的主要供应商。甲骨文与OpenAI合作的“星际之门”(Stargate)项目,在软银和MGX的巨额融资支持下,成为其未来AI战略的巅峰之作。该项目最初于2025年提出,计划投资5000亿美元,横跨美国四个州及多个国家,但已在部分区域遭遇阻力,特别是在其主要基地得克萨斯州,涉及用水和电网压力等问题。
与此同时,甲骨文为建造AI数据中心已承担超过180亿美元的债务,这让许多观察者担忧该供应商所承受的风险水平,以及其能否迅速实现预期的建设收入以履行偿债义务。甲骨文本月表示,向基于AI的商业模式转型以及内部部署AI系统,是其今年裁员约2.1万人(约占员工总数13%)的主要原因之一。
中文翻译:
由谷歌云赞助
选择您的首个生成式AI应用场景
要开始使用生成式AI,首先应聚焦于那些能够改善人类与信息交互体验的领域。
这家超大规模云服务商正在持续推进其智能体平台的建设,同时不断深化自身的人工智能能力。
甲骨文于周二发布了新功能,旨在帮助企业开发者为甲骨文Fusion应用程序构建、协调和管理AI智能体。
Fusion应用套件涵盖财务、供应链管理、制造、人力资源、销售、客户服务与体验以及市场营销等平台。
这些面向开发者的无代码和传统人工编程(即专业代码)工具,是对2025年3月推出的甲骨文Fusion应用AI智能体工作室中已内嵌编程功能的补充,后者主要面向业务用户。
借助甲骨文所称的“AI原生构建者体验更新”,开发者可以使用来自第三方供应商的最流行的AI编程模型和智能体,包括OpenAI Codex、Claude Code、微软VS Code、谷歌Antigravity以及开源分布式版本控制系统Git。在Fusion套件中工作的开发者也可以使用来自不同供应商的命令行界面。
甲骨文将使用这些工具创建的应用程序称为“Fusion智能体应用”。这些应用是由专业智能体组成的系统,它们能够推理、协调并采取行动,在Fusion工作流中执行任务。
“我们现在强调的是,必须平等对待每一位用户。无论你是想用无代码方式,还是作为专业开发者想在自己的专业环境中编写代码,都可以。”甲骨文云应用开发高级副总裁纳塔莉亚·拉克尔森表示。“专业开发环境已经进化了。开发者们在其中感到舒适。那就留在那里,但我们会为你提供工具和插件,让你能够留在那个环境中,构建一个能在Fusion中运行的智能体应用。”
KramerERP的创始人兼分析师罗伯特·克莱默认为,这些面向开发者的工具是甲骨文智能体平台的一次适度更新。
“这不仅仅是多了一个智能体或开发工具,但在客户拿出实际成果之前,我不会将其描述为重大变革。”他说。“企业资源计划(ERP)正是智能体AI发挥价值同时也是伴随风险的地方。让一个智能体去汇总发票或建议催收行动相对容易。但要让它更新财务记录、释放订单、发起付款或更改供应计划,同时维护安全性、审批流程和审计追踪,难度就大多了。”
但克莱默补充说,智能体平台的价值不仅限于ERP,还延伸至人力资本管理、供应链管理和客户体验管理;在受控环境中跨更广泛的业务流程协调多个智能体的能力至关重要。他表示,其他潜在的好处包括加快智能体应用投入生产的速度,并让业务用户、开发者和合作伙伴更容易地使用一系列无代码、低代码和高级AI编程工具协同工作。
“问题在于,这套方案在甲骨文生态系统之外表现如何,企业如何管理多智能体故障,以及业务用户在构建生产应用时应该拥有多少控制权。”克莱默继续说道。“现有的安全措施和审批流程有所帮助,但它们解决不了数据质量差、流程所有权不明确或问责机制缺失的问题。甲骨文还需要在运行时成本、集成成本以及可衡量的客户成果方面提供更清晰的说明。”
这家供应商对其智能体AI平台的最新更新,正值这家拥有49年历史的科技巨头正大力向AI领域转型,目标是成为生成式AI实验室OpenAI等客户所需AI数据中心算力的主要供应商。
甲骨文与OpenAI合作的“星际之门”项目,获得了软银和MGX的重大融资,是甲骨文未来AI战略的顶点。该项目最初于2025年提出,耗资5000亿美元,横跨美国四个州和多个国家,但在一些地方遇到了阻力,尤其是在其核心所在地得克萨斯州,问题涉及水资源使用和电网压力等。
此外,甲骨文为建造AI数据中心而承担的超过180亿美元债务,也让许多观察人士担心这家供应商所承受的风险水平,以及它是否能足够快地实现其预期的收入以偿还债务。
甲骨文本人也将其向AI业务模式的转变以及内部AI系统的部署,列为今年裁员约21000人(约占其员工总数的13%)的主要原因。
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.
The hyperscaler continues to build out its agentic platform as it deepens its AI capabilities.
Oracle on Tuesday released new capabilities to help enterprise developers build, coordinate and govern AI agents for Oracle Fusion Applications.
The Fusion application suite includes platforms for finance, supply chain management, manufacturing, HR, sales, customer service and experience and marketing.
The developer-focused no-code and pro-code -- traditional human programming -- tools complement the coding capabilities, primarily aimed at business users, already embedded in the Oracle AI Agent Studio for Fusion Applications, introduced in March 2025.
With what Oracle calls the AI-native builder experience update, developers can use the most popular AI coding models and agents from third-party vendors. These include OpenAI Codex, Claude Code, Microsoft VS Code, Google Antigravity and Git, the open source distributed version control system. Developers working in the Fusion suite can also use command line interfaces from various vendors.
Oracle terms apps created with these tools Fusion Agentic Applications. The apps are systems of specialized agents that reason, coordinate and take action, executing work within Fusion workflows.
“We are now saying we really need to treat everybody equally. Do you want to do no-code or are you a professional developer and want to code in your professional environment?” said Natalia Rachelson, senior vice president of cloud applications development at Oracle. “The professional environments have evolved. They're comfortable there. Stay there, but we're going to give you tools and plugins such that you can stay in that environment and build an agentic application that will run in Fusion.”
Robert Kramer, founder and analyst at KramerERP, said he views the developer-focused tools as a moderate update to Oracle’s agentic platform.
“It is more than another agent or development tool, but I would not describe it as a major change until customers show results,” he said. “ERP is where agentic AI becomes valuable and also risky. It is relatively easy for an agent to summarize an invoice or recommend a collections action. It is harder to let that agent update financial records, release an order, initiate a payment or change a supply plan while maintaining security, approvals and an audit trail.”
But Kramer added that the agentic platform’s value extends beyond ERP into human capital management, supply chain management and CX; the ability to coordinate multiple agents across broader business processes within a governed environment is important. Other potential benefits, he said, include moving agentic applications into production faster and making it easier for business users, developers and partners to work together using an array of no-code, low-code and advanced AI coding tools.
“The questions are how well this works outside Oracle, how enterprises manage multi-agent failures and how much control business users should have when building production applications,” Kramer continued. “Existing security and approvals help, but they do not solve poor data quality, unclear process ownership or accountability. Oracle also needs to provide more clarity around runtime costs, integration costs and measurable customer outcomes.”
The vendor’s latest updates to its agentic AI platform have come as the 49-year-old tech giant is pivoting sharply toward AI, with ambitions to become a major supplier of AI data-center-generated compute for customers such as generative AI lab OpenAI.
The Oracle-OpenAI Stargate project, with major financing from SoftBank and MGX, is the capstone of Oracle’s AI strategy for the future. Originally proposed in 2025 as a $500 billion project across four states and several countries, the undertaking has encountered turbulence in some places, particularly in its anchor state, Texas, over issues such as water use and grid stress.
Also, the more than $18 billion in debt Oracle has taken on to build AI data centers has many observers concerned about the level of risk the vendor has accepted and about whether it will realize the revenue it anticipates from the buildout quickly enough to meet repayment obligations.
Oracle itself cited the move toward an AI-based business model and internal deployment of AI systems as major reasons for laying off some 21,000 employees, or about 13% of its workforce, over the course of this year.
文章标题:甲骨文将代理式人工智能工具的重点聚焦于Fusion应用开发者。
文章链接:https://news.qimuai.cn/?post=4564
本站文章均为原创,未经授权请勿用于任何商业用途