Visa and Lianlian Complete Live B2B Agentic Payment

WHAT HAPPENED

Visa and Lianlian DigiTech announced on 24 July that they had completed a live business-to-business transaction using LoopXPay, an AI agent developed by Lianlian.

The companies described it as the first live B2B agentic transaction in Greater China. That description is based on their announcement and should not be interpreted as an independently verified industry-wide record.

The transaction involved the purchase of a product sample from a supplier. LoopXPay handled multiple stages of the process within a single workflow, moving beyond product discovery or payment recommendations to execute a real commercial transaction.

Unlike a conventional checkout in which a buyer manually selects a product and confirms each step, the AI agent interpreted the purchasing requirement and performed several connected actions on behalf of the business.

HOW THE AI AGENT COMPLETED THE PURCHASE

According to Visa and Lianlian, LoopXPay first identified the purchasing requirement and recommended suitable suppliers.

It then compared available options, placed the order and executed the payment. The entire process was completed within one automated workflow rather than through separate procurement, ordering and payment systems.

However, the agent was not given unlimited authority.

The transaction operated within predefined spending controls and approval parameters established by the business. These controls define the boundaries within which the agent can act, including how much it can spend and which transactions require further approval.

This distinction is important. Agentic payments do not necessarily mean an AI system has unrestricted access to company funds. A commercial agent can automate approved activities while remaining subject to limits, permissions and human oversight.

For businesses, the practical value comes from connecting product discovery, supplier comparison, order creation and payment execution without removing financial controls.

VISA ADDS AN AGENT IDENTITY AND TRUST LAYER

LoopXPay has also been registered in Visa’s Agentic Directory.

The directory is designed to help participating businesses and merchants identify AI agents that have met Visa’s requirements. It supports Visa’s Trusted Agent Protocol, which provides identity, transparency and control mechanisms for agent-driven commerce.

These functions address a new payment question: when an automated system initiates a transaction, how can a merchant, payment provider or issuer determine that the agent is recognised and acting within authorised parameters?

In a conventional online payment, the transaction can be connected to a cardholder, device, account and established authentication process. Agent-led transactions introduce another participant—the AI agent—which may search, select and initiate the purchase.

Payment networks are therefore developing systems that can distinguish a recognised commercial agent from an unknown automated bot.

The Visa and Lianlian transaction shows how an agent registry, transaction permissions and payment credentials may operate together. It does not eliminate the need for merchant risk controls, authentication or transaction monitoring.

WHY THIS MATTERS FOR B2B PAYMENTS

Much of the early discussion around agentic commerce has focused on consumer shopping. This transaction applies the model to B2B procurement, where purchasing processes may involve supplier comparison, internal budgets, approval limits and accounting requirements.

Small and medium-sized businesses may not have dedicated procurement teams. Employees often spend time locating suppliers, comparing offers, preparing orders and arranging payment manually.

An AI agent could reduce that workload for routine and low-risk purchases. It may also help standardise procurement decisions by applying the same spending limits and approval rules across transactions.

For payment providers, the change means APIs may need to support more than transaction initiation. Agentic B2B payment workflows can require:

• Verified agent identities
• Merchant and supplier data
• Spending and transaction limits
• Approval rules
• Tokenised payment credentials
• Audit trails
• Exception handling
• Refund and dispute processes

Businesses also need records showing what the agent was instructed to do, which options it considered and why a particular payment was executed.

WHAT HAS NOT YET BEEN DISCLOSED

Visa and Lianlian did not disclose the transaction value, supplier name, payment credential, settlement route or the number of businesses currently using LoopXPay.

The announcement therefore demonstrates that a live controlled transaction was completed, but it does not establish the system’s transaction volume, commercial availability or performance at scale.

The companies have not announced a general launch date for every merchant or business in Greater China.

Any description of LoopXPay as universally available, fully autonomous or proven at large commercial scale would go beyond the published evidence.

WHAT COMES NEXT

Visa and Lianlian said they are exploring additional uses for AI agents across procurement, digital advertising optimisation and B2B platform payments.

These areas involve recurring purchasing decisions and large numbers of routine transactions, making them suitable for controlled automation.

The next stage will depend on whether businesses can integrate agent-led purchasing with their existing finance, compliance and approval systems.

The industry should also monitor how payment networks handle agent authentication, liability, consent, refunds and disputed transactions as AI agents become more involved in commercial payments.

The milestone suggests that agentic commerce is moving from demonstrations toward live B2B payment workflows. Its broader impact will depend on adoption, interoperability and whether businesses trust agents to operate safely within clearly defined financial boundaries.