Ant International is attempting to redefine the architecture of global commerce by transitioning from traditional fintech services to a fully "agentic" financial ecosystem. By deploying an AI-native, end-to-end stack covering payments, accounts, foreign exchange (FX), and treasury, the company is positioning itself to move beyond human-intermediated transactions toward a future where AI agents manage complex cross-border workflows. This strategic pivot, announced at the company's VOYAGE merchant event in Shanghai, introduces nearly 100 products designed to automate the lifecycle of global business operations through natural-language interaction and proprietary foundation models.
Ant International Deploying AI-Native Financial Stack
The company's latest product upgrade centers on two proprietary foundation models designed to handle the high-velocity data requirements of global finance. The first, the Antom 3-in-1 Transformer, is a payment foundation model with over 10 billion parameters that processes 90 trillion data points annually. It is engineered to simultaneously analyze sequential, tabular, and graph data to optimize payment success rates and enhance fraud detection. Complementing this is FalconTST, an 8.5 billion-parameter time-series transformer model used for FX and liquidity forecasting. Ant International claims FalconTST has achieved over 93% accuracy in forecasting, which has reportedly cut corporate FX hedging and allocation costs by 30% to 60% in real-world applications.
To support these models, Ant International is introducing the Alipay+ Agentic Mobile Protocol (AMP), an open protocol designed to enable payments via AI agents across digital wallets and smart devices. The protocol aims to reduce the steps required to link a payment agent to a wallet by 50% and supports "nano-grade" agent-to-agent (A2A) settlements as small as $0.000001. Furthermore, the company is utilizing WhaleRTP, a blockchain-based wholesale settlement platform, to manage liquidity. In 2025, WhaleRTP processed 45% of Ant International's cross-border volume, which the company states reduced working capital requirements by 60% and increased interest income by 23%.
The suite is organized into four business pillars: Alipay+, Antom, WorldFirst, and Bettr. These pillars aim to provide "autopilots" for various workstreams, such as the Antom Autopilot, which manages the merchant payment lifecycle from onboarding to reconciliation. The company is also launching the world's first "Account for Agent" (AFA) through WorldFirst, which utilizes smart contracts and a "know-your-agent" (KYA) framework to allow businesses to interact with AI agents securely.
Scaling AI Integration Across Global Merchant Networks
The rollout of these AI-native solutions follows significant existing adoption within Ant International's merchant ecosystem. According to the company, 89.5% of clients using its main merchant payment service, Antom, have already deployed FinAI solutions within the last 12 months. Furthermore, 81.4% of payment tasks are now being completed with AI assistance. This high level of integration suggests that the move toward a full-stack AI architecture is a response to documented merchant demand for connected solutions that go beyond simple payment processing to address the complexities of global operations.
Ant International's infrastructure currently links 150 million global merchants to over 2 billion consumer accounts, facilitating daily average transactions exceeding 25 million across 210 markets. The new suite aims to capitalize on this scale by offering specialized tools like Antom SmartDispute, which the company claims can drive average dispute success rates from 24.5% to 41.8%. For SMEs, the WorldFirst brand is introducing "Wyn," a natural-language AI agent designed to manage account workflows, including opening accounts and applying for virtual or physical cards in 17 currencies via single commands.
Security remains a central component of this technological expansion. Ant International is implementing a two-layered security architecture that includes AgentSafePay, a product the company describes as the industry's only security offering that provides a fund-back guarantee against losses stemming from agent-specific risks, such as malicious prompt injection or intent misinterpretation. The company is also collaborating with Mastercard and Visa to develop interoperable agent-security standards through its KYA framework. These products are scheduled for a global rollout throughout the fall and winter of 2026.
Key Takeaways
- Ant International has launched an AI-native financial stack featuring two proprietary models: the 10-billion-parameter Antom 3-in-1 Transformer and the 8.5-billion-parameter FalconTST for FX forecasting.
- The company reported that 89.5% of Antom merchant clients have deployed FinAI solutions in the past year, with 81.4% of payment tasks aided by AI.
- The new ecosystem includes the Alipay+ Agentic Mobile Protocol (AMP) for AI-agent payments and WhaleRTP, a blockchain-based platform that processed 45% of the company's cross-border volume in 2025.
FinanceInsyte's Take
In our view, Ant International is not merely adding AI features to existing tools; it is attempting to build the foundational operating system for an "agentic" economy. By moving into the "Account for Agent" (AFA) space and developing protocols like AMP, the company is betting that the next era of financial transactions will be driven by autonomous software rather than human users. This is a high-stakes strategic shift that targets the very plumbing of global commerce.
The emphasis on proprietary foundation models—specifically those handling multi-modal data like the Antom 3-in-1 Transformer—suggests that Ant International recognizes that generic LLMs are insufficient for the precision required in high-frequency payments and FX management. If the company can successfully standardize the "know-your-agent" (KYA) framework and secure interoperability with major players like Visa and Mastercard, it could establish a dominant position in the infrastructure that governs how AI agents move value. This transition from a service provider to a protocol provider represents a significant escalation in their competitive positioning within the global fintech landscape.
Questions & Answers
How does the Antom 3-in-1 Transformer differ from standard large language models in a financial context?
Unlike standard LLMs, the Antom 3-in-1 Transformer is specifically designed to process three distinct types of core financial data: sequential (behavior over time), tabular (structured records), and graph (relationships between entities and devices). This multi-modal approach is intended to optimize payment success rates and fraud detection by analyzing complex relationships that a single-mode model might miss.
What specific financial efficiencies does the FalconTST model claim to deliver to corporate clients?
The FalconTST model, a time-series transformer designed for FX and liquidity forecasting, is reported to achieve over 93% accuracy in forecasting. In real-world applications, the company states this capability has enabled corporations to reduce their FX hedging and allocation costs by between 30% and 60%.
How is Ant International addressing the unique security risks associated with autonomous AI agents?
The company is implementing a two-layered security architecture that includes a "know-your-agent" (KYA) framework and a product called AgentSafePay. AgentSafePay is positioned as an industry-first solution that offers a 100% fund guarantee against losses caused by agent-specific risks, such as malicious prompt injection or the misinterpretation of intent.
What is the strategic purpose of the Alipay+ Agentic Mobile Protocol (AMP)?
AMP is an open payment protocol designed to enable seamless payments by AI agents across various digital interfaces, including smartphones and AR glasses. It aims to reduce the integration friction for payment agents by 50% and supports ultra-small, high-frequency "nano-grade" settlements between agents as low as $0.000001.
Source: Businesswire
