XTransfer Highly Commended For AI Fraud Detection Excellence

XTransfer, the world’s leading B2B cross-border trade payment platform, has been highly commended in the category of Best In-house Use of AI in Fraud and Financial Crime Detection at the 8th Regulation Asia Awards for Excellence 2025. XTransfer is notably the only B2B cross-border payments provider to be recognised at this year’s ceremony.

“We were impressed by XTransfer’s innovative application of a custom-built LLM to address the specific fraud and AML challenges in B2B cross-border trade,” said a judge on the Regulation Asia Awards panel. “The solution, TradePilot, leverages a unique, proprietary dataset to automate compliance for the SME sector, demonstrating XTransfer’s clear vision for making global trade safer and more accessible.”

Processing over US$12 billion in payments each month, XTransfer is applying AI not for trend value, but to solve real operational challenges while strengthening financial integrity. During a recent industry keynote, Yanfang Liu, Co-founder and CTO of XTransfer, detailed how the company is redefining B2B cross-border financial services with AI and why XTransfer built its own TradePilot model from the ground up.

Transforming Document Processing with Multimodal AI

Cross-border trade generates a vast array of non-standardised documents, such as business licenses, commercial invoices, customs forms, and logistics bills, each of which varies by country and even by region.

"For example, in many developing countries, there are numerous types of personal identification documents that differ by state," Liu noted. 

Previously, traditional OCR systems could recognise only about 5% of these documents, forcing teams to rely heavily on manual data entry. The turning point came with multimodal AI models, which can comprehend both the spatial layout and the semantic meaning of complex documents.

“With multimodal large models, we’ve automated around 50% of document recognition for non-standardised certificates and trade documents,” said Liu. “That’s a tenfold leap in efficiency, significantly cutting manual workload while preserving accuracy.”

Reinventing Risk Management with AI Intelligence

Risk control lies at the heart of XTransfer’s operations, verifying if clients are engaging in legitimate trade and maintaining consistency across information, logistics, and fund flows.

Traditional rule-based systems and machine learning models demanded extensive human correlation of fragmented data sources.

AI has changed that equation. XTransfer now uses AI-driven data collection and multi-source cross-verification to identify hidden relationships across the supply chain.

“In recognising connections between buyers and sellers, such as upstream-downstream or component relationships, our large models can now make more precise judgments than human analysts,” Liu explained.

After more than a year of deployment, this AI-assisted system has cut risk management personnel costs by over 20%, with expectations of reaching 50% savings within the next two years.

Building AI-Native Customer Operations

Beyond risk control, XTransfer is embedding AI directly into its customer-facing ecosystem. The company has built a multi-agent AI operations system in which different “agents” handle specialised tasks, from marketing strategy and creative generation to performance evaluation. “Supervisor agents” orchestrate task allocation and quality control.

This system integrates DeepSeek as the reasoning engine and connects with XTransfer’s proprietary knowledge base. The company’s AI employee product, launched in 2023, has seen rapid adoption among clients, while its smart customer service solution has achieved a 100% improvement in issue resolution rates.

Why XTransfer Built Its Own TradePilot Model

While XTransfer leverages advanced third-party models like Alibaba’s Qwen and DeepSeek, it has also developed its in-house TradePilot model, now in its multimodal 2.0 version.

Liu explained the motivation, “General-purpose large models have immense internet knowledge, but their understanding of niche financial domains can be shallow. By training a domain-specific model, we achieve superior performance on trade finance tasks and ensure data security, which is especially critical under evolving global regulatory scrutiny.”

In document recognition scenarios, Liu said, “TradePilot outperforms even ChatGPT and Google’s latest models within XTransfer’s specialised verticals, proving that targeted domain adaptation often beats general-purpose intelligence.”

A Balanced Path Toward AI Transformation

XTransfer’s approach reflects a deliberate balance between innovation and prudence. In high-stakes risk management, AI augments, not replaces, human oversight. In customer engagement and operations, automation takes a more central role. The company’s results underscore how AI can revolutionise legacy financial systems without compromising compliance, accuracy, or trust.

As Liu summarised, “The key is knowing what AI can do and what it shouldn’t. That understanding shapes how we continue evolving AI for global trade.”