As digital payment adoption accelerates across APAC markets, financial institutions face an escalating battle against increasingly sophisticated fraud schemes. Traditional rule-based detection systems, while reliable, struggle to identify complex fraud patterns that adapt faster than rule updates. For banks processing millions of transactions daily, even a 0.1% improvement in fraud detection accuracy can translate to millions in saved losses and enhanced customer trust. The challenge is particularly acute in APAC's diverse regulatory environment, where banks must balance fraud prevention with compliance across multiple jurisdictions. Simultaneously, they need to effectively combat fraud while delivering the fast, reliable experiences that customers demand.

Worldline has recently announced Cardlink Next, which is a robust and versatile platform designed to efficiently manage electronic payment transactions, with particular strength in processing credit card transactions quickly and securely. Building on Worldline's decades of payment processing expertise and serving over 80 banks across APAC, Cardlink Next represents the evolution from traditional payment processing to AI-powered financial intelligence. The newly introduced AI Fraud Detection module boosts its capabilities by adding AI-driven features that can detect intricate fraud patterns typically missed by conventional rules-based systems. This module integrates an AI model to analyze historical fraud data, assigning each transaction a fraud score that indicates the probability of fraud. Although the AI model provides a probability score, the new AI Fraud Detection module goes a step further by classifying transactions as fraud, suspected fraud, or legitimate based on configurable threshold settings. This enables issuing banks to implement more advanced, AI-driven fraud detection mechanisms within Cardlink Next.

The Cardlink Next solution is powered by IBM Z systems, noted for its scalability, reliability, and security features. Unlike other AI fraud detection modules running off-site with latency issues and SLA challenges. Worldline's Fraud Detection & Prevention module is co-located on IBM Z systems and benefits from Telum chip AI Accelerator and zIIP engines to enhance the solution's performance and latency. This powerful hardware foundation is complemented by Worldline's proprietary fraud detection pre-trained Al model, which provides a robust foundation designed to be fine-tuned or retrained using the bank's own historical transaction data, enabling it to adapt and recognize fraud patterns specific to their customer base, product offerings, and regional nuances—from account takeover schemes common in mature markets to card-not-present fraud prevalent in emerging digital economies across Southeast Asia. The new z17™ will have advanced features that leverages IBM z17™ Telum on chip acceleration and Spyre extendable technologies, to scale AI capabilities with speed. The IBM z17™ integrates advanced AI into hybrid cloud, optimizing performance, security and decision-making where critical data resides. With AI built in, IBM z17 drives innovation, powers new workloads and enhances productivity—all in a secure, reliable and resilient environment designed for your needs. IBM z17 with scalable AI capabilities enables more sophisticated AI models for advanced fraud detection, including using secure generative AI on your most trusted transactional data for real-time insights and applying multi-model AI with LLMs. IBMs enterprise focus on AI with IBM z17™ hardware and platform solutions allows building practical AI solutions with governance, transparency, proactive identification and mitigation of risks and potential damages with AI-powered security. Built on data security with the world's first series of quantum-safe systems.

Use case/benefits: Fraud Detection

By going beyond traditional rule-based and machine learning detection algorithms to include deep learning models—particularly deep neural networks (DNNs), recurrent neural networks (RNNs), and advanced large language models (LLMs)—we can more effectively identify complex fraud patterns in credit card transactions that traditional rule-based and classical ML methods often struggle with. IBM z17™ provides the necessary capacity for the scalability and sophistication required by AI models in Cardlink Next. It leverages the integrated AI Accelerator with scalable options, such as Spyre cards, to run Transformer models. New fraud patterns are detected by analyzing variations in transaction sequences and surrounding context. AI systems process both structured and unstructured data, handling variable numbers of transactions over short or long periods, and identifying both vertical and horizontal fraud patterns. The synergy of rule-based approaches with cutting-edge Transformer AI technology operating on IBM z17 ensures top-tier accuracy in fraud detection across 100% of credit card transactions—real-time and within SLAs.

Following are the advantages of using Cardlink Next for AI Fraud Detection:

Minimize Losses and Gain Trust: Real-time fraud detection and response. Advanced AI: Trained on synthetic credit card data from IBM Research using state-of-the-art Transformer model. Local deployment: With own AI model, data stays within the bank. Transparency: Transparent AI code base. Control: Fine-tune & re-train model as needed. Customizable: Can be customized for specific needs. Continuous Learning: Enhance model performance with new fraud data. Dynamic Authorization Check: Adjust responses based on thresholds. Customizable Fraud Thresholds: Global and card-type-level fraud thresholds. Balanced Detection: Optimize balance between fraud detection and precision. Full Control: Direct access to model training and testing.

The integration of AI within COBOL applications running on the z/OS system offers innovation while continuity of z/OS qualities of service, security, compliance, longevity, resilience, etc. Cardlink Next comes with a stock model that can be fine-tuned on banks data. IBM's concept of train anywhere with any open tools makes it an easy task for data scientist at the bank. A new version of the model is exported into ONNX interchange format and then deployed into CICS COBOL z/OS region along with Cardlink Next. The model is optimized for z17 zIIP processing to enhance cost efficiency. This integration simplifies AI interaction for CICS COBOL applications through standard commands, allowing seamless invocation of ONNX models. Additionally, it leverages the IBM z16/z17 Integrated Accelerator for improved AI performance, including features such as the IBM Deep Learning Compiler and server-side micro-batching. The solution also supports various models and provides robust lifecycle management for AI deployments.

Here are the list of functionalities that will enhance Cardlink leveraging IBM Z17 features. These features will work better with Z17 Telum and on chip acceleration (Spyre).

1. AI Features a. AI leveraged Fraud detection & Prevention b. AI Leveraged Advance Merchant Payments
2. Functional Enhancements a. Account Level Management b. Flexible Credentials c. Multiple Currency Cards now on Cardlink Next
3. Standard Integrations – Cardlink Next – Leveraging z/OS Connect a. Integration to other systems and improving connectivity
4. Data Access – z/DIH on z16/17 - platform to access Cardlink data for analytics/reporting/dashboard

The combination of IBM Z's reliability with modern AI/ML technology and the Worldline Cardlink value chain results in a powerful and efficient solution for banks and payment systems, enhancing operational effectiveness and customer service. Worldline's deep understanding of APAC's unique payment ecosystem—from supporting local payment methods to navigating complex cross-border regulations—distinguishes the company in the market. This regional expertise, combined with Worldline's global scale and extensive transaction processing experience, enables the delivery of AI solutions that are not just technically advanced but market-relevant and compliance-ready. It empowers banking institutions by offering enhanced security features for fraud detection and prevention, ensuring compliance and protection for both banks and their customers. With seamless integration into core banking systems, Cardlink supports various payment networks across channels like ATMs, POS systems, and online platforms, providing customers with diverse and convenient credit card payment options. Its scalability accommodates businesses of all sizes, facilitating growth and adapting to evolving market demands. Cardlink's efficient user interface and powerful capabilities play a critical role in streamlining transaction processing and reinforcing banks' payment processing infrastructure.

Worldline's Cardlink Next addresses the critical challenge facing APAC banks: detecting sophisticated fraud patterns that traditional systems miss while maintaining real-time transaction processing. By combining Worldline's proven payment expertise with IBM Z's AI acceleration, banks can reduce fraud losses while improving customer experience through faster, more accurate transaction approvals.