Author: Zheng Jiayi, Wall Street Insights

On July 24, Visa and Lianlian Digital jointly announced that they had completed their first real B2B intelligent agent transaction executed by Lianlian's AI intelligent agent, LoopXPay.

This is also the first publicly disclosed transaction of this type in Greater China.

This transaction served a small business seller;

In terms of process, LoopXPay completed supplier sourcing, solution comparison, order placement, and payment within a single workflow, operating entirely within the enterprise's pre-set expenditure control and approval parameters.

LoopXPay's role goes beyond simply recommending suppliers; it directly connects procurement decisions with payment execution—enterprises only need to pre-set budgets, permissions, and approval conditions, and the intelligent agent can complete daily procurement within the authorized scope, while the enterprise retains necessary supervision and control.

LoopXPay's role extends beyond simply recommending suppliers; it directly connects procurement decisions with payment execution—enterprises only need to set budgets, permissions, and approval conditions in advance, and the intelligent agent can complete daily procurement within the authorized scope, while the enterprise retains necessary supervision and control.

For payment institutions, whether AI can complete payments is only the first step. After an AI agent initiates a transaction in the name of a company, new problems need to be addressed in the payment chain: who authorizes it, the scope of authorization, the trustworthiness of the transaction counterparty, and how to trace anomalies. As an important part of the cooperation between the two parties, LoopXPay has registered in Visa's Agentic Directory. This directory aims to help merchants and companies identify verified AI agents and, in conjunction with the Visa Trusted Agent Protocol, provide identity verification, transparency, and expenditure control mechanisms for agent transactions. This is equivalent to adding a layer of "identity registration" for AI agents entering the payment network: previously, payment institutions mainly verified cardholders, merchants, and devices; after an agent participates in a transaction, the verification scope is further expanded to include the software entity that performs tasks on behalf of the company. However, this mechanism is still in its early implementation stage. The information released by both parties has not yet disclosed the transaction amount, product category, specific payment method, or the manual confirmation steps required from demand identification to final payment. Therefore, the outside world cannot currently assess its level of automation and commercial replication costs. Especially in B2B scenarios, procurement often involves multiple stages such as contract terms, goods acceptance, invoice processing, and cross-border compliance. A successful single transaction proves that the intelligent agent can enter the procurement and payment process, but it does not equate to its ability to handle complex enterprise procurement. Visa and Lianlian revealed that their next plan is to expand their cooperation to areas such as procurement, digital advertising optimization, and B2B platform payments. Lianlian Digital stated that its capabilities for building intelligent agent businesses will cover identity authentication, transaction authorization, smart payments, and global fund settlement. From the perspective of the evolution of the payment industry, AI agents previously focused more on customer service, marketing, and operations; now, they are beginning to acquire the ability to execute orders and allocate funds. The questions that payment infrastructure needs to answer have shifted from "how to enable AI to make payments" to "how to verify who it is, how much it can spend, and who is responsible for its actions." When AI begins to spend money on behalf of businesses, identity, authorization, and accountability mechanisms will become more important than the payment action itself.