Have you ever wondered about a strange charge on your credit card statement or a cryptic transaction on your bank statement? To find out where and what you spent the money on, you may have to contact your bank to get information about the merchant or the purpose of the transaction. A positive customer experience looks and feels different.

That is why accuracy is paramount when it comes to payment data. From transparency to recommendations and predictions, it is critical to creating personalized experiences and greater satisfaction. To achieve this, financial services companies rely on high-quality, comprehensive payment data.

Payment enrichment refers to the process of enriching transactions in order to increase the quality and information content of payment data and provide personalized recommendations. It is a subset of data enrichment, which has been widely used in marketing and retail since the 1990s to improve customer profiles and enable more targeted marketing campaigns. With the advent of big data and advanced analytics technologies, fintechs and financial service providers, including SIX, have begun to develop and use payment enrichment to make transaction data self-explanatory. To do this, they combine their own payment data with external data sources and use various techniques such as pattern recognition to derive the meaning of the transaction. This allows them to generate new content such as the category (e.g., transportation, groceries), unique merchant names, logos, and geographic information about the merchant, i.e., the location of the store. This enriched data not only improves the customer experience, but also makes AI applications more efficient and optimizes key processes such as chargebacks.

Imagine a bank account that collects your customers’ payment data and analyzes their spending habits. As a bank, you can identify spending patterns, such as frequent purchases at certain stores, regular payments for subscriptions, or recurring invoices. You can use this information to help your customers make better financial decisions. For example, if you notice that a customer frequently shops at a particular grocery store, you can offer a credit card with cash back rewards. Alternatively, you can offer intuitive services to help your customers better understand their personal finances. This can range from simple questions such as “How much did I save last month?” to more complex questions, such as identifying unusual transactions, predicting future grocery spending, or providing personalized recommendations to optimize savings. All of this is possible when high-quality payment data is combined with appropriate AI techniques to answer specific questions.

After the compulsory part comes the freestyle. It's like juggling. First, the juggler must master the basic techniques and movements to keep the balls or clubs safely in the air. Only then can they move on to freestyle and try out creative, more complex tricks and combinations. The situation is similar for the application of AI in open finance, especially in the context of the emergence of multibanking for private clients in Switzerland. Here, the need for high-quality and consistent payment data will increase – regardless of the house bank. The real trick is to combine the various payment data from different banks and information from external sources in such a way that it can be used for new applications to offer personalized financial services.

Like juggling, managing the quality of payment data is not a one-time process, but an ongoing one. Only by constantly checking and adjusting the data can it be made available in a quality that not only meets today’s requirements, but also tomorrow’s.

In payments, AI tools are already essential for card fraud detection. They analyze vast amounts of transaction data in real time to identify unusual spending patterns that indicate potential fraud. They can also be used to answer simple customer questions, such as those in a FAQ catalog.

For direct debit processing, payment processors today primarily use AI to analyze aggregated numbers and metrics from transaction data and to make predictions. In the future, however, there is potential for more predictive applications, such as fraud alerts and error prevention.

An AI that learns with each new data point and updates its “knowledge” on the fly is called online learning. However, this process is very complex and difficult to scale. In addition, malicious actors can manipulate the feedback loop, in which what is learned is fed back into the system to influence its future behavior. In practice, therefore, AIs are often static (they do not learn continuously) and require human oversight for updates and evolution.

Other important aspects of payment data processing include complying with data protection laws (e.g., GDPR), implementing security protocols, dealing with the challenges of ensuring data security, and gaining and maintaining customer confidence. The use of AI in the Swiss financial services sector is also accompanied by the regulatory expectations of FINMA. In this context, financial service providers must meet and document criteria such as reliability, equal treatment, transparency, and accountability, as well as clarify issues such as governance and accountability. Accountability, in particular, has been a persistent and unsolved problem for researchers for many years. One reason is that many AI models, especially complex ones like deep learning networks, are considered “black boxes.” These models make decisions based on internal processes that are difficult for humans to understand. Another problem is the so-called disagreement problem, where different explanatory methods provide different and sometimes contradictory explanations. This makes it difficult to find a consistent and reliable explanation for a model’s decisions. In addition, many of these methods are susceptible to manipulation, which further reduces the trustworthiness of the explanations.

Ethical considerations also come into play when using data, such as avoiding discrimination and ensuring fairness.

All of these factors can hinder the rapid adoption of AI. Future developments and innovations are therefore of great importance, especially approaches that promote security, transparency, and customer trust in AI-based decisions. One such approach is the integration of AI into blockchain networks. 

By using cryptography, blockchain makes management of and access to AI-generated data more secure and more in line with regulatory expectations. It verifies data sources and monitors their quality, ensuring data transparency and traceability. Smart contracts regulate the deployment and use of AI models, making the ecosystem fair and efficient.

The combination of AI and blockchain is promising, but still in the development and experimentation phase. Another exciting area is the combination of quantum computing with AI and blockchain. Quantum computers may be able to perform complex calculations and data analysis at speeds far beyond the capabilities of classical computers. Imagine if the “textbook” of AI were not only expanded, but transformed by the immense computing power of quantum computers. This could lead to even more accurate and efficient AI models that can process and analyze unimaginable quantities of data in real time. The synergy of quantum computing, AI, and blockchain could revolutionize the future of financial services – like a textbook to which we add a chapter with an entirely new dimension of knowledge.