Machine learning transforms the Swiss financial center

Artificial intelligence enables banks and insurers to offer new personalized services and products (istock-foto)

18.05.2021

Swiss banks and insurers have integrated a range of machine learning applications into their daily business over the past year. This growing digitalization has not only resulted in slashed costs. The continued adoption of innovative machine learning tools, such as the offering of ultra-personalized financial services, has the potential to boost the country’s financial sector.

The use of machine learning (ML) within the financial industry currently focuses on automating processes or processing data. ML is for instance used by banks in risk and fraud assessment, where machines rapidly process large amounts of data. “The tool alerts the compliance officers if any suspicious activity is detected. This enables them to focus on the complex cases and undertake additional assessments whenever needed,” explains Pascal Wyss, Head Artificial Intelligence at Zurich-based Technology, Innovation & Management (TI&M).

Machine learning technology is also used for automated compliance checks. “Supervisors turn to such tools to carry out checks on persons, to make sure clients are not politically exposed or have links to countries that are blacklisted” by the Financial Action Task Force, Wyss adds. The autonomous onboarding of new clients is another field where applications using ML has become common practice. “It is nowadays possible to open a bank account in five minutes at any time of the day thanks to artificial intelligence (AI),” Wyss notes. (See fact box below on the difference between AI and ML.) And there are numerous other ML applications at hand or being developed likely to disrupt the financial sector in the years to come.

 

The artificial intelligence revolution

“Banks and insurers have actually just started playing with artificial intelligence and machine learning. This technology not only increases their productivity and results in cost savings, but also enables them to offer new personalized services and products,” says Wyss. “Vast amounts of research about AI already exists, but we have not yet witnessed a widespread adoption of the numerous tools at hand by the financial industry,” emphasizes Dr. Branka Hadji Misheva from the Zürcher Hochschule für angewandte Wissenschaften (ZHAW). “The ongoing Covid-19 pandemic has helped to speed up the process of digitalization due to the restrictions in place,” Dr. Jan-Alexander Posth, senior lecturer at ZHAW, notes.

 

Uptake of ML tools slowed down by legacy infrastructure, regulations…

There are several reasons behind the relatively slow uptake of the ML applications on the market by the financial sector. “The legacy IT infrastructure used by banks is one hurdle. These systems often lack the necessary flexibility and capacity to support the various computing requirements associated with running state-of-the-art ML algorithms and producing outputs in real time.”

Another key barrier to the swift adoption of ML tools is explainability. Bankers who will use the developed AI and ML systems must be able to interpret and understand how the system actually makes the decisions,” Hadji Misheva underlines. She leads the project “Towards Explainable Artificial Intelligence and Machine Learning in Credit Risk Management” that aims at resolving explainability issues through an application targeted to be released in early 2022. Stringent data protection laws such as the EU General Data Protection Regulation (GDPR) faced by Swiss banks also require the explainability of the ML software used. “As a direct consequence of the subprime crisis in 2007, regulators no longer accept black-box models. Banks must be able to explain how decisions are reached when a loan application, for example, is accepted or rejected by a ML tool,” Posth explains. 

 

Ultra-customized financial products hitting the market soon

The next phase within machine learning is likely to be the emergence of ultra-personalized financial products and services, Wyss forecasts.The service landscape will change as companies will be able to offer cross services. These will be based on the data they have collected about their clients: Neobanks (online banks) will for instance be able to offer personalized banking services where trading costs are optimized, and portfolios automatically designed to reflect the risk appetite of the client. Car manufacturers will provide perfect car insurances where you only are insured for your actual drive, as sensors will detect and analyze your movements…” Another area set to take off is the emergence of conversational interfaces in the form of chatbots, able to respond to complex queries. “Chatbots can for example calculate how much someone should inject into his or her supplementary pension scheme by analyzing earlier bank transfers made into it,” says Wyss, who currently develops such chatbot prototypes. The use of AI applications such as the ones just mentioned within the Swiss financial sector is likely to accelerate in the coming years, partially boosted the strong presence of fintech start-ups in the country.

Artificial intelligence versus machine learning

Artificial intelligence (AI) is a wide and fluid concept that has evolved over time. What started off with simple chess playing programs in the 1950s has developed into complex applications able to detect cancer, analyze the speed of traffic or take automated decisions for online credit applications. Machine learning (ML) is a subset of AI used in multiple applications. ML provides systems the ability to automatically learn new processes and concepts leading to an output, while only being fed with input.

Pascal Wyss, TI&M
Branka Hadji Misheva, ZHAW
Jan-Alexander Posth, ZHAW