Customers might not yet notice it, but Germany’s banks are slowly waking up to artificial intelligence. They hope thinking computers can help them sort through regulatory jungles, fight money laundering or even help analyze customers. But for now, they’re just getting comfortable with the technology.
“In the next few years, the increase is likely to be driven mainly by efficiency improvements, and later new products and services could have a greater effect,” says Christian Kirschniak, a data expert and partner at PwC. The consultancy believes AI’s contribution could be similar to that of the first computer revolution and contribute $2 trillion (€1.73 trillion) to global GDP by 2030, in part by relieving financial sector employees of tedious, monotonous tasks that traditional computers can’t perform.
Take regulations: The constant flow of new ones requires timely evaluations of new laws and guidelines. “An AI system can analyze any number of documents in a short time and mark problematic passages in them,” says Hendrik Rees, an AI expert at PwC. The system can also monitor outward communications and sound an alarm if it detects a misstep.
The idea has already sparked Andreas Totok to declare war on stupid paperwork in the financial sector. He’s leading the AI charge at an IT service provider for Germany’s state-run Sparkassen savings banks. Mr. Totok recently used AI to read hand-written forms and then enter the information in the right database for a Sparkasse credit unit.
“In order to evaluate them automatically, we used a neural network that we trained in handwriting recognition. The system now has a recognition rate of more than 98 percent and continues to learn,” he says.
Meanwhile, at Commerzbank, the country’s No. 2 listed bank, AI may not only recognize texts in the future, but also write them. Its Main-Incubator division and the start-up Retresco are developing a prototype. The idea: An AI system that can evaluate and write about quarterly results of listed companies.
“We were able to show that machine learning offers real added value, especially in the area of highly quantitative analyses, and can support analysts,” says Michael Spitz, managing director of Main-Incubator. The technology has to be evaluated for reliability but Retresco is already using a similar system for sports reporting.
Speed is particularly important in customer identification, known in financial circles as “Know Your Customer” (KYC), and in money-laundering prevention. Banks must avoid banking with criminals while protecting against money laundering and terrorist financing. Failure to do so can lead to high fines.
You paid too much for that muffin
AI can quickly create a risk assessment of new customers, for both private and corporate customers. When it comes to loans, assessing creditworthiness as accurately as possible reduces the credit default risk.
Financial startup Fintecsystems is working on a digital self-disclosure that scans a client’s online bank account and scores their creditworthiness. Correctly evaluating purchases and income is vital to the system, which relies on about 10,000 rules, but will likely expand via a self-learning loop. “We feed an AI system with manually classified sales, from which the machine derives rules and learns,” explains Tobias Ruland, head of product development at Fintechsystems.
Intelligent analysis systems are already quite common in payment transactions. “For example, banks and credit card companies already ask customers if a credit card has been used in different countries within a short period of time,” says Markus Hamprecht, head of Financial Services at Accenture. “AI allows further analysis of usage behavior, making unusual transactions more noticeable.”
Fraugster, another startup, shows how such analyses can be used for online payments. The company handles risk management for payment service providers such as Ingenico and Six and, according to its own statements, hedges payments with an annual volume of €35 billion.
Checks and balances
Instead of fixed rules, founder Max Laemmle relies on a self-learning system: “Just reducing the fraud rate is not enough, you have to reduce the proportion of payments that are wrongly rejected at the same time,” Mr. Laemmle says. “Within 15 milliseconds, up to 2,500 data points are checked. If, for example, the location of the IP address does not match the address, it’s suspicious.”
E-mail addresses also provide information. Fraudsters often use e-mail addresses with random combinations of letters, while serious users often include their name in the address.
TeamBank is testing a completely different application. The cooperative credit specialist, which brokers loans under the easyCredit brand, wants AI to KI assist in resource planning.
“When we place advertisements, the workload in the area of credit processing and customer service increases,” a spokesperson for TeamBank says. An intelligent forecast provides information on how individual advertising formats could affect staffing requirements.
Whether bank or fintech, almost every company in the financial sector deals with artificial intelligence. But to keep up with the competition, most banks will have to do more, says Mr. Kirschniak, the PwC partner.
Mr. Totok from FI-SP also sees a need for investment. “We now have to take money in hand for development and accept that not everything works immediately,” he says. “We will not be intimidated by startups and pioneers from abroad.”
Katharina Schneider is a correspondent in the finance section of Handelsblatt based in Frankfurt. Andrew Bulkeley adapted this story into English for Handelsblatt Global. To contact the author: email@example.com