AI agents streamline cross-border business payments

Cross-border payments are undergoing a transformation driven by AI agents, as businesses compete to develop the necessary infrastructure for automating transactions across international borders. This change is compelling merchants and financial networks to rethink how payments integrate into operations, shifting them from a final step to an authorised, embedded action within broader workflows.
The 2026 Singapore FinTech Festival featured Chinese fintech provider Lianlian DigiTech introducing three products aimed at addressing this challenge. Lianlian XWorker serves as an AI workforce platform that coordinates specialised agents to execute business tasks. Lianlian XPay links agent-initiated payment requests to their execution, while Lianlian XCard, a Visa-linked business card, provides spending controls and AI-driven capabilities. These tools can operate independently or together, with XWorker functioning as an open platform.
Cross-border vendors currently face a disjointed process when managing workflows. Tasks such as market research, supplier payments, and order fulfilment often require switching between multiple tools, each with distinct subscription requirements and operational rules. Jerry Zhou, chief architect of Lianlian’s AI business unit, observed that while AI agents already exist, the burden of integrating and coordinating them falls on humans. “There are already plenty of AI agents out there, but switching between them, coordinating and communicating creates enormous costs,” Zhou said.
Traditional commerce treats payments as a post-task action, but agent-driven trade demands they be authorised within the task itself. For example, an agent sourcing suppliers should initiate payments according to predefined rules without awaiting human approval. According to S&P Global’s 2026 Voice of the Enterprise Merchant Study, 65% of merchants are exploring partnerships with payment providers to support agentic commerce, while two-thirds anticipate AI agents will manage at least 10% of e-commerce transactions within three years.
China Leads with Domestic AI Payments
China’s domestic market has advanced further in this area. Alipay processed over 300 million agent-initiated transactions via AI Pay in 2026, and WeChat Pay has introduced an AI-specific card. However, cross-border business-to-business agentic payments remain in early development stages.
Lianlian’s strategy focuses on workflow orchestration. XWorker utilises RouterBrain.ai, its AI computing infrastructure, to direct tasks to appropriate models while maintaining context across different tools. Budgets, transaction limits, and approval rules are established in advance, enabling programmable spending rather than manual processing. Zhou emphasised that cost efficiency must not overshadow achieving business objectives: “Cost efficiency is a consideration, but not at the expense of performance, Zhou said.”
The Agent Wallet enforces these rules by defining which business entity an agent represents, its spending limits, and permitted use cases. For instance, an agent handling supplier payments operates only within its delegated authority, preventing overages or unauthorised transfers.
A major challenge remains establishing trust in AI agents. Traditional KYC and KYB procedures do not apply to AI entities, which lack legal standing. Lianlian’s solution combines Know Your Agent (KYA) verification with security controls. The first layer confirms the agent’s identity and the business it represents: “If the underlying entity cannot be trusted, you certainly cannot trust its agent,” Zhou said. The second layer applies programmable rules to restrict spending based on predefined conditions.
Auditable records track agent identity, authorisation scope, approval steps, and transaction execution, ensuring compliance and accountability—particularly important for cross-border transactions where disputes may arise. Lianlian and Visa completed a B2B agentic transaction in July 2026, followed by another in October using an agent-specific payment token.
Trust and Standards Hold Back Global Adoption
The lack of common standards presents another obstacle. Lianlian intends to support Visa TAP, UnionPay APOP and Mastercard’s protocols through compatibility adapters rather than commit to a single framework. “There is no standard at the moment. Agents registered in the different card networks’ trusted agent registries cannot yet recognise one another,” Zhou acknowledged.
Human confirmation therefore remains necessary in many situations, particularly for larger transactions. Nonetheless, adoption is expected to begin with frequent, low-value payments, such as token purchases, API calls and small advertising expenditures. Mastercard’s August 2026 Signals report, “Encoding Trust”, similarly identified verifiable intent, agent identity, and clear permission frameworks as necessary components.