Frankfurt The shareholder letters from Jamie Dimon, head of the world’s largest bank JP Morgan, are something like a seismograph for the important issues in the financial world. In this year’s letter, the top manager dedicated an entire paragraph to the topic of artificial intelligence: “AI is an extraordinary and groundbreaking technology. AI and its underlying raw material, data, will be critical to the future success of our business,” writes Dimon.
Then follows a series of numbers: The institute employs 1000 data management staff, 900 data scientists and 600 machine learning programmers. Incidentally, Jamie Dimon announced: In the future, language models such as ChatGPT – applications that use artificial intelligence to understand and interact with natural language – will also support the bank’s employees.
The example of the largest bank in the world shows: Chatbots have arrived in the financial world. Driven by the hype surrounding ChatGPT, banks are trying to develop their own models of this type.
The importance of artificial intelligence is undisputed in the financial sector. The algorithms should develop new products, help customers and catch fraudsters – some of them are already doing this today. But many institutes still shy away from ChatGPT itself.
According to the Bloomberg news agency, Bank of America, Citigroup, Goldman Sachs and Wells Fargo have banned the tool. As a result, JP Morgan initially restricted the use of voice AI. In Germany, the two large private banks Deutsche Bank and Commerzbank are just as restrictive.
Artificial intelligence for financial investments: chatbots advise customers
Nonetheless, experts recognize that companies are approaching AI in a different way. “For the past few months, we’ve seen a trend that banks are making efforts to program their chatbots more intelligently,” says Michael Berns, director for AI at the auditing firm Pricewaterhouse-Coopers (PwC). He accompanies banks in their technological advances. “Banks can now buy better solutions for their own systems from external providers and thus have more resources to focus on their customers’ AI needs to concentrate.”
Commerzbank already uses methods that it classifies as trustworthy in dialog systems. Deutsche Bank is also working on a 3D avatar that customers can talk to. A seemingly necessary step: Berns assumes that customers are more likely to contact a chatbot than meet with an advisor.
>> Read here: Head of the BIS Innovation Center – “Models like ChatGPT could revolutionize the financial system”
The chat programs used by the two large private banks are significantly less complex than ChatGPT’s Large Language Model (LLM). “Companies don’t need this monster of parameters for their purposes,” says Berns. Because the banks’ chatbots don’t have to be able to chat about the weather. But very well about how to invest large sums of money or how to order a new credit card.
In investing, banks use an algorithm that Netflix also uses. With the streaming service, this suggests series and films to viewers that they might like based on their previous favorites. For banks, the program gives investment recommendations, such as buying government bonds, if they fit an investor’s portfolio. The AI informs bank advisors, for example if a security’s credit rating is downgraded or if analysts recommend selling a share. These algorithms are called “Next Best Offer”.
Regulators have provided clarity on AI use
A few years ago, German banks could not keep up with their international competitors when it came to AI. In spring 2020, Berns published a study in which the German financial sector was only slightly above the “rather underdeveloped” mark.
In the meantime, however, the German institutes have caught up. Commerzbank and Deutsche Bank, for example, both have a cooperation with Google Cloud, which offers, among other things, an AI platform.
In addition, the banks have recently been helped by the clearer set of rules by the bank supervisors, says Berns. “We welcome the fact that the EU AI regulation will create framework conditions for artificial intelligence in the course of the year,” says Markus Drösser, group leader in IT development at DZ Bank.
Since then, German banks have automated their internal processes using artificial intelligence. For example, when the computer automatically filters and processes data from documents. AI programs detect anomalies in data by being trained using models.
>> Read here: What ChatGPT can do
“The model training is often based on data that is marked by people,” explains a Commerzbank spokeswoman. “Then the model is applied to new data to generate results in the form of, for example, content, predictions, classifications or recommendations. “
Among other things, the programs can unmask potentially criminal activities or discover credit risks. Because money laundering is and remains a major problem for German banks. The software trawls through transactions and checks them for various criteria such as currency, amount, country of origin and destination. If the AI thinks a transaction is suspicious, it reports it to the account manager.
German financial sector wants to catch up when it comes to AI
The fact that the German financial sector was so far behind its international competitors is also due to a lack of know-how. That is why banks in this country are increasingly hiring data scientists or programmers who can drive their AI projects forward.
Deutsche Bank, for example, has set up an internal AI center to explore the extent to which AI can be used in all business areas. “Companies are benefiting a little from the waves of layoffs from the big tech companies, whose specialists are increasingly recruiting them,” says Berns.
More: Artificial intelligence is revolutionizing insurers