Clearstream Head Warns of Limitations of Simple AI

As many firms look to embrace artificial intelligence (AI) and robotics to automate processes, others have taken a more cautious approach.

Berthold Kracke, head of global operations at Clearstream and CEO of Clearstream banking AG, says robotics is useful in improving efficiency and automation, but can only be applied in limited use cases. Robotics is considered a subset of AI technology but operates with a simple rules-based methodology. It is this simple functionality that makes it useful for only basic computations or tasks.

“A robot is very much rules-based,” Kracke says. “To understand the limitations of robots from a logical perspective, it can help you with an ‘if-then’ solution. You must know exactly or be able to define what the robot does. [In terms of] deviation, it cannot do anything else.”

Often these technologies are applied to a simple task to automate parts of the workflow. In these cases, Kracke says, it is much more effective to revamp or restructure the process entirely to achieve straight-through processing rather than applying robotics to small areas of the system.

Another drawback is the level of maintenance and cost required to run the machines. This often relates to cases where a tech engineer may need to continuously upgrade the robots and input coded instructions to be able to manage new tasks or functionality.

“Robots must be used carefully,” Kracke says. “The downside is that if you end up with a huge number of robots, you must still be able to operate them efficiently and of course you need to have some kind of monitoring. [For example] if there are system upgrades, maybe the robot was trained to look into a certain field, maybe in one application the format of the field has changed and the robot must learn this, so having robots also require having some maintenance.”

However, robotics has shown promise in one area for the post-trade services firm: data extraction.

For financial markets firms that have been around for decades, one of the core challenges they face is data management, particularly for those struggling with siloed legacy systems located in multiple parts of the globe. According to Kracke, robotics and AI applications in some cases can offer a quick fix for extracting data from one system and moving it to another. This can also be potentially useful in sourcing unstructured data, which can then be manually formatted in a Microsoft Excel sheet for a client or for internal consumption.

Intelligent artificial intelligence

More sophisticated AI technologies that go beyond the simple rules-based structure have greater potential to transform workflows and resolve complex challenges, Kracke says.

Clearstream has just begun exploring intelligent algorithms that are based on statistical learning, often referred to as machine learning (ML). In one use case, the firm took 3.5 million unstructured Swift messages and developed an AI model that would allocate the messages and instructions from clients to the correct working groups and teams. Kracke says that to date the model has achieved an STP rate of 98%.

As another example, Clearstream has also developed a digital AI assistant called Albert that uses natural language processing to read documents and attachments from clients, which are then forwarded to internal team members. Albert sits on the team members’ inbox, it identifies keywords in the email or documents and triggers a workflow, which is then handed over to the individual handling the task.

“This is the first step of this kind of workflow and it is very promising,” Kracke says. “By means of optical character recognition [technology that recognizes text characters] and natural language processing, data are extracted and the workflow is initiated by an AI client and then monitored.” 

In the future, the firm is going to identify new uses cases where AI can be applied, Kracke says: “If you utilize artificial intelligence, you must be very careful that you do it in an efficient and maintainable, scalable manner. Therefore, we have set up an IT center of excellence for automation and we are currently investigating which kind of tool landscape we want to establish.”