New AI Agent Gameplay in the Financial Sector | Lin Xiaozhen

CN
1 hour ago

179503737388796.jpg

179503737575531.jpg

At the beginning of the month during Hong Kong's FinTech Week, financial institutions, technology leaders, and FinTech companies gathered from various places. A stroll around the venue makes it clear that the most prominent keyword at the moment is "Agentic AI," and discussions have shifted from the desktop clerical work agents of the early "lobster" craze to specialized agents capable of replacing multiple financial professional roles. At the same time, financial institutions are struggling with how to manage their own agents and deal with external agents.

AI agents have deeply taken root

At FinTech Week, you can see FinTech solution providers everywhere developing AI agent tools for financial institutions. For example, a startup showcased an agent that processes loan applications, wherein basic tasks like collecting information from clients and preliminary reviews are fully automated. The AI agent can even call clients for follow-ups, easily handling Cantonese, English, and Mandarin. Even if clients are busy and need to reschedule, the AI agent can still manage the situation humanely, making it hard for clients to distinguish whether they are interacting with a real person or AI.

AI agent "intelligent organization" gradually taking shape

What everyone sees is an all-function AI agent, but in reality, it operates through collaborative agents, each responsible for specific procedures. One agent focuses on customer interaction while another checks credit databases, communicating with each other. For example, the agent responsible for checking credit information only shares the results and summarized reasons with the customer-facing agent, preventing the disclosure of sensitive information like approval criteria. Besides horizontal division of labor, there are also hierarchical relationships, similar to your office, where there are supervisors and subordinates. For instance, compliance is crucial for financial institutions, and the instructions from compliance agents often supersede those from other agents.

This division of labor and structure at the AI agent level aligns perfectly with the "intelligent organization" concept recently proposed by Baidu (9888) founder Li Yanhong at the "Baidu World 2026" conference. In simple terms, agents have their own areas of expertise that can enhance efficiency, and during collaboration, each has different data access permissions and even checks and balances mechanisms to prevent errors by agents. This is an important aspect of "AI agent governance" for enterprises.

Difficult to distinguish between human and agent troubles banks

AI has evolved beyond serving humans to serving other agents. A technology company has even launched a credit card specifically for agents, facilitating enterprises' procurement via AI, but issues have arisen. Traditional credit card transactions require signature verification, and online transactions have verification mechanisms. However, with agents also able to "swipe cards," can retailers and banks clarify whether a real person is using the card or it is being used by an agent? If the buyer's agent places a chaotic order, and the seller's agent accepts it, who is responsible? In the event of agent card misuse, who bears the liability? Moreover, if there are differing interpretations during interactions between agents, how should decisions be made? These pose new challenges for banks.

At the beginning of the month, the Monetary Authority published a discussion paper on AI agent governance in the banking sector, reminding financial institutions to "Know Your Agent" (KYA). However, it only provided general suggestions like formulating appropriate measures based on nature and risk and conducting regular reviews, while specific implementations still rely on the industry to explore gradually.

This article is reprinted from Sing Tao Headlines, with copyright belonging to the original author and original media.

免责声明:本文章仅代表作者个人观点,不代表本平台的立场和观点。本文章仅供信息分享,不构成对任何人的任何投资建议。用户与作者之间的任何争议,与本平台无关。如网页中刊载的文章或图片涉及侵权,请提供相关的权利证明和身份证明发送邮件到[email protected],本平台相关工作人员将会进行核查。

Share To
APP

X

Telegram

Facebook

Reddit

CopyLink