内容来源：https://aibusiness.com/agentic-ai/anthropic-finance-agents-threaten-established-service-providers

内容总结：

AI公司Anthropic瞄准金融行业，发布十款智能体模板冲击华尔街

由谷歌云赞助

为帮助用户快速切入生成式AI应用，专家建议首先关注那些能改善人类信息体验的领域。

AI公司Anthropic正加速进军企业市场，并重点布局金融行业。本周，该公司一口气推出十款可直接运行的智能体（Agent）模板，专门针对金融服务业中最耗时的工作环节，包括制作路演手册（Pitchbook）、筛选尽职调查文件以及月末结账等。这些模板以插件形式嵌入其Claude Cowork和Claude Code产品中。据Anthropic介绍，每个模板都是一个参考架构，包含技能（指令与领域知识）、连接器（数据访问接口）和子智能体（由其他Claude版本驱动）。金融公司可根据自身建模规范、风控策略和审批流程对智能体进行定制调整。

与此同时，Anthropic本周还与富达国民信息服务公司（Fidelity National Information Services）成立了一家16亿美元的合资企业，用于开发金融犯罪软件；另与多家华尔街公司联合投资15亿美元，向企业销售其AI工具。

一系列动作表明，Anthropic正从单纯的模型提供商向服务特定行业的AI平台提供商转型。今年早些时候，其推出的Claude Cowork插件曾震惊法律界，被部分人士视为对初级法律岗位和法律信息提供商的威胁。今年4月，网络安全行业也对Claude Mythos识别和利用安全漏洞的能力感到担忧。如今，这家AI公司再次出手，打造支持研究、客户覆盖以及财务运营的智能体。

一方面，自2021年成立以来持续亏损的Anthropic，随着今年预计的IPO临近，似乎正步入盈利轨道。“Anthropic在企业业务上的表现非常强劲，大多数人都认为这令人印象深刻。”高德纳（Gartner）分析师汤姆·科肖（Tom Coshow）表示。

在瞄准企业市场的过程中，Anthropic不仅争夺软件服务商在AI技术领域的市场份额和机遇，还瞄准了IT系统陈旧的老牌金融公司。借助这些新智能体，Anthropic直接与彭博（Bloomberg）、FactSet等初级分析师用于制作比较分析的传统数据系统，以及印孚瑟斯（Infosys）、埃森哲（Accenture）等传统咨询服务商展开竞争。

伊利诺伊大学芝加哥分校数据科学与AI战略副校长迈克尔·贝内特（Michael Bennett）指出，Anthropic正瞄准美国最赚钱的行业之一。金融业掌握敏感数据并管理着特殊的客户关系。企业必须决定是否让智能体访问这些敏感信息。然而，在金融领域，投资回报率同样至关重要。

“这是提升这些金融关系投资回报率的重要工具，”贝内特说，“仅凭速度提升、准备效率、顾问服务乃至发现新客户的能力，整个行业就必须进行深刻反思。”

此外，随着Anthropic的行业专用智能体开始威胁到那些拥有细分领域专业知识的中小型金融公司，这些企业可能面临艰难抉择。“企业是选择与一家专门为金融业构建AI智能体的公司合作更好，还是Anthropic的智能体真的能处理书中所有的边缘案例？”科肖反问道，“如果它们能做到，那对很多人来说将是巨大的威胁。”他还补充说，目前尚不清楚Anthropic的智能体是否需要大量调整和对接工作才能连接到客户数据。

“所有涉及AI智能体的问题，归根结底在于它们能否将所使用的数据置于具体情境中做出决策，以及它们的可靠性如何，”科肖进一步指出，“企业将拥有多少自主的AI能力？如果Anthropic包揽一切，他们是否还觉得自己拥有这些能力？这将是未来五年驱动企业的核心智能。”

贝内特表示，除了对小型供应商构成威胁、让老牌金融公司和新兴金融科技公司面临生存危机外，这些新智能体还可能侵蚀初级金融分析师和入门级分析师的岗位。“他们现在做的许多工作，未来将通过订阅服务来完成，”他说，“仅凭这一点，我们就应预见到该行业将受到重大影响。”他还补充道，教育工作者也将面临压力，需要思考在AI智能体日益普及的环境下，如何培养入门级金融从业者实现职业进阶。

中文翻译：

由谷歌云赞助
选择你的首批生成式AI应用场景
要开始使用生成式AI，首先应聚焦于那些能改善人类与信息交互体验的领域。

该供应商的新智能体有望入驻大型华尔街公司，对中型服务商构成威胁，并开始取代初级金融岗位。

随着其持续积极进军企业市场并构建智能体化AI工具，Anthropic如今瞄准了一个重要行业：金融。

这家AI实验室本周早些时候推出了十个即用型智能体模板，针对金融服务业中最耗时的环节，如制作项目推介书、筛查尽职调查文件以及月末结账。这些模板以插件形式集成在Claude Cowork和Claude Code中。Anthropic表示，每个模板都是一种参考架构，包含技能（指令和领域知识）、连接器（数据访问接口）以及子智能体（由其他Claude版本驱动）。该AI实验室称，金融公司可根据自身的建模规范、风险策略和审批流程对任何智能体进行调整。

同样在本周，Anthropic与富达国民信息服务公司达成16亿美元的合资企业，用于开发金融犯罪软件；并与华尔街其他公司组建了另一家15亿美元的合资企业，向企业销售其AI工具。

金融服务业智能体及其他金融合资项目进一步表明，Anthropic已开始从模型提供商转型为服务特定行业的AI平台提供商。今年早些时候，Claude因其推出的Claude Cowork插件震惊了法律行业，一些人认为这对初级法律岗位和法律信息提供商构成了威胁。4月，网络安全行业也开始担忧Claude Mythos识别和利用安全漏洞的能力。

如今，这家AI模型制造商再次出手，这次创建了支持研究、客户覆盖以及财务与运营的智能体。

一方面，Anthropic自2021年成立以来一直处于亏损状态，但随着今年预期中的IPO临近，它似乎正走上盈利轨道。

“Anthropic在进军企业业务方面做得非常出色，”Gartner分析师Tom Coshow表示，“大多数人都会同意这令人印象深刻。”

在瞄准企业市场时，该供应商还追求了软件服务提供商在AI技术中看到的关注度和机会，以及拥有过时IT系统的老牌金融服务公司。通过这些新智能体，Anthropic与Bloomberg和FactSet等传统数据系统（初级分析师用于整理比较分析）以及Infosys和Accenture等传统咨询服务展开竞争。

伊利诺伊大学芝加哥分校负责数据科学与AI战略的副校长助理Michael Bennett表示，通过金融服务智能体，Anthropic瞄准了美国最赚钱的行业之一。

金融行业持有敏感数据并管理着特殊的客户关系。Bennett称，企业必须决定是否允许智能体访问这些敏感信息。然而，在金融领域，投资回报率同样重要。

“这是提高这些（金融）关系投资回报率的重要工具，”Bennett表示，“如果仅仅是为了提升速度、准备、咨询甚至发现新客户的速度，行业内部将需要进行大量反思。”

此外，随着Anthropic的特定领域智能体如今对在这些领域拥有专业知识的较小金融公司构成威胁，这些公司很可能需要做出艰难的决定。

“企业是选择与专门为金融业构建AI智能体的公司合作更好，还是Anthropic的智能体真的知道如何处理书中所有边缘案例？”Coshow问道，“如果它们能做到，那对许多人来说将是一个巨大威胁。”他补充说，目前也不清楚Anthropic的智能体是否需要调整以及大量工作来连接数据。

“涉及AI智能体的一切，核心在于它们是否能够对用于决策的数据进行情境化处理，以及它们的可靠性如何，”Coshow继续说道，“它们将拥有多少自己的AI？如果Anthropic为它们包办一切，它们是否感觉自己拥有AI？这将是五年后驱动企业的智能。”

Bennett表示，除了对小型供应商构成的危险，以及老牌金融公司和新兴金融科技公司面临的潜在生存危机，新智能体还可能削弱初级金融助理和初级分析师的岗位。

“他们现在所做的许多工作未来将订阅服务所覆盖，”他说，“仅凭这一点，我们就应预期该行业将受到重大影响。”

他补充说，教育工作者也将面临压力，需要思考如何在AI智能体迅速普及的环境中培训初级金融从业者，使其能够取得进步。

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 vendor’s new agents could find a home in big Wall Street firms, threaten mid-sized service providers and start to push entry-level finance jobs aside.
As it continues to aggressively push into the enterprise market and build agentic AI tools, Anthropic is now targeting a major industry: finance.
The AI lab introduced ten ready-to-run agent templates earlier this week, targeting what it said is the most time-consuming work in financial services, such as building pitchbooks, screening due diligence files and closing the books at the end of the month. The templates ship as a plugin in Claude Cowork and Claude Code. Each template is a reference architecture that includes skills (instructions and domain knowledge), connectors (access to data) and subagents (powered by other Claude versions), Anthropic said. The AI lab said that financial firms can adapt any of the agents to their own modeling conventions, risk policies and approval flows.
Also this week, Anthropic entered into a $1.6 billion joint venture with Fidelity National Information Services for financial crime software and another $1.5 billion joint venture with other Wall Street companies to sell its AI tools to businesses.
The financial services agents and other financial ventures are further signs that Anthropic has begun to evolve from a model provider into an AI platform provider serving specific industries. Earlier this year, the Claude stunned the legal industry when it introduced its Claude Cowork plugins, which some saw as a threat to entry-level legal jobs and legal information providers. In April, the cybersecurity industry also became worried about Claude Mythos’ ability to identify and exploit security weaknesses.
Now, the AI model maker is at it again, this time creating agents to support research, client coverage, and finance and operations.
On the one hand, Anthropic appears to be on track to profitability after losing money since its 2021 founding, as it approaches an expected IPO this year.
“Anthropic has done a really strong job of going after enterprise business,” said Tom Coshow, an analyst at Gartner. “Most people would agree that it’s been impressive.”
In targeting the enterprise market, the vendor has also pursued the mindshare and opportunities that software service providers saw in AI technology, as well as with older financial services firms with dated IT systems. With these new agents, Anthropic competes against traditional data systems like Bloomberg and FactSet, which junior analysts use to pull together comparative analyses, as well as legacy consulting services such as Infosys and Accenture.
With the financial service agents, Anthropic is going after one of the most lucrative industries in the country, said Michael Bennett, associate vice chancellor for data science and AI strategy at University of Illinois Chicago.
The finance industry holds sensitive data and manages special client relationships. Enterprises will have to decide whether to grant agents access to that sensitive information, Bennett said. However, ROI is also important in finance.
“This is a major tool for increasing ROI on those [finance] relationships,” Bennett said. “If only for the increase in speed, in preparation, in advising and actually even finding new clients, there’s going to have to be a lot of soul searching in the industry.”
Moreover, with Anthropic's domain-specific agents now threatening smaller finance companies with expertise in these areas, those firms will likely need to make challenging decisions.
“Would enterprises be better off partnering with a company that builds AI agents for finance, or do Anthropic agents really know how to handle all the edge cases out of the book?” Coshow said. “If they do, that's a very big threat to a lot of people.” It’s also unclear if the Anthropic agents require tweaks and significant work to connect to data, he added.
“Everything involving AI agents is about whether or not they can contextualize the data that they're using to make a decision about what to do and how reliable they are,” Coshow added. “How much of their AI are they going to own, and do they feel like they own it if Anthropic is doing everything for them?” he continued. “This is the intelligence that is going to drive enterprises in five years.”
In addition to the dangers to smaller vendors and the potential existential crisis facing older financial firms and newer fintech vendors, the new agents could erode the roles of junior finance associates and entry-level analysts, Bennett said.
“The work that they do now is going to be covered in many instances by a subscription,” he said. “We should expect a significant impact in the industry, if only for that reason.”
He added that educators will also face pressure to figure out how to train entry-level finance workers to advance in an environment in which AI agents are proliferating.