Deutsche Bank AG is currently testing artificial intelligence tools within the framework of scenarios for using this technology as a means of identifying potentially possible signs of misconduct based on studying the tone of telephone conversations of traders.

The giant of the German banking sector decided to get acquainted with Google Cloud machine learning. This work is carried out as part of a broader strategy to explore the possibilities of using artificial intelligence by a financial institution. Bernd Leukert, Director of Technology, Data and Innovation, told the media about this. The Bank has not yet announced when the artificial intelligence system will be integrated into the structure of its activities.

Currently, the financial institution is focused on forming its own understanding of how large language models can become effective tools in the framework of employee and customer support. The bank is also interested in studying the scenario of using machine intelligence, in which advanced technology is applied to help with coding.

At the current historical moment, financial institutions around the world are making efforts to increase the level of monitoring effectiveness, since in this case, rates are rising very quickly against the background of harsh measures by governments against violations. For example, the total amount of fees paid by banks for involvement in the practice of using unofficial communication channels, such as WhatsApp, has exceeded $2.5 billion since December 2021.

The monitoring tools currently used by Deutsche Bank have the ability to analyze the semantic content of what has been said, but at the same time, they are not able to respond to changes in tone and cannot detect some kind of hints of cynicism. The new system should understand the hidden meaning of what was said and navigate in situations where an emotionally neutral phrase in the context of a conversation means a description of some illegal actions, containing verbal symbols and hints that the interlocutors identify, but are not able to recognize third parties who are not aware of all the features of conversations between certain people. Also, the updated toolkit should identify any evidence that the information under discussion is secret, and communication through public communication channels is indicated exclusively in a superficial-conditional format without detailing details.

Bernd Leukert says that a tool that can understand the tonality and pay attention to it provides much better control compared to mechanisms that fix exceptionally suspicious words. He also stated that the financial institution has the potential to take management to a completely different level.

Deutsche Bank trains artificial intelligence using data arrays generated on the basis of existing conversations between traders so that the AI understands their language and communication norms. The software will direct analysts to the areas of interest to them, people will make final decisions about further actions.

Employees of Deutsche Bank’s innovation department are currently working to ensure that video surveillance systems comply with regulatory acts. Achieving this goal will help to avoid problems with confidentiality. Also, in this case, false positives will be minimized. Bernd Leikert says that it is extremely important to ensure fair banking algorithms, in which there is no hidden discrimination, and the results are correct and understandable.

Currently, the number of Deutsche Bank employees who are engaged in technology, data, and innovation is about 16 thousand people, about half of whom are engineers working on the code. According to the consulting company Evidence, this financial institution is one of the five largest lenders for recruitment using artificial intelligence.

At the same time, some lawyers say that using AI as a tool to identify unfair practices entails potential risks. Duncan Black, a financial services partner at Fieldfisher LLP in London, says that software aimed at finding laughter or a deal that is more profitable than the average values can ignore more important alarm signals. Also, according to him, there is a possibility that traders will start claiming that the charges against them are AI-generated content.

Duncan Black acknowledged the advantages of technologies that are able to detect patterns in huge amounts of data faster and more efficiently than people but noted that it is necessary to monitor what is used as a basis in this case.

The risks and disadvantages of AI are still being identified and eliminated, but at the same time, the use of machine intelligence in the banking sector is scaling up. Many lenders apply Google’s cloud technology to transform unstructured data, including text messages from traders and transcripts of phone calls.

Societe Generale SA uses AI to analyze traders’ communications across 30 channels, including mobile phone calls, instant messages, and social networks.

Ann Beaumont, a partner at the law firm Friedman Kaplan Seiler Adelman & Robbins LLP in New York, says that artificial intelligence systems can function as a lie detector and detect signs of suspicious behavior without having to prove their presence.

As we have reported earlier, Deutsche Bank AG Survives Slowdown in Trading Well.