For renowned credit institutions like DKB, reputation is an important economic factor. It also plays an important role in the loan granting process. The DKB wants to ensure that it grants loans to trustworthy companies that operate responsibly.

The DKB has recognized that there is still technology potential to improve the credit decision-making process. A DKB bank advisor often has to decide in a short space of time whether a corporate customer will receive a loan. This is the only way the customer experience is right.

But how can publicly available information about a company be included in the credit decision in the shortest possible time?

The bank advisor needs a software application to assist him. It works like this: The bank employee enters the company name into the software application. The application searches millions of pieces of information about the company (e.g. through comments, news articles or posts) from publicly accessible sources such as Twitter, Facebook or t-online. The application then does a sentiment analysis using AI models (e.g. BERT, related to GPT) and gives the bank advisor an indication:

• “70% Negative”
• “20% Positive”
• “10% ambivalent”

At this point it is important to mention that the results of the sentiment analysis are only used as a guide. Nuances such as irony and exaggeration cannot be captured. This step is mainly about recognizing publicly known patterns of a company. Example: The sentiment analysis gives a predominantly negative indication. The bank advisor then uses the sources provided by the software for more detailed research and learns more about the background to the publicly shared news. In this way, the DKB - in conjunction with all other available information - can minimize its risk when granting loans.

This result is just part of the information that can be included in the credit decision. The software application accelerates this step within the risk assessment. The bank advisor does not have to search through all the sources independently. He can incorporate the generated results into his analysis and build on them. The better a bank advisor knows his customers, the better he can advise them.

Doesn’t the implementation sound simple? Yes, we think so too. But that was not always so. To be more precise, the implementation effort has only been very low since December 2022. Why? In December 2022, ChatGPT was released. And here too it, or BERT (the sister of GPT), finds its use. BERT stands for “Bidirectional Encoder Representations from Transformers” - a machine learning technique and pretraining of Transformer-based Natural Language Processing models (NLP models) developed by Google and introduced in 2018.

But take another step back: Why was it so difficult to technically implement this use case pre-GPT?

What we need here: A model that reliably indicates whether a text is written positively, negatively or neutrally.

“Conventionally” the case would have been processed using a natural language processing approach, i.e. completely “old school” with Python and the NLTK program library.

Alternatively, you would have had to train your own model very laboriously. These are exactly the approaches that require an extremely large amount of data, time and AI experts. The results in our short validation were anything but promising. You would probably have quickly rejected the case or fallen into a costly trap with a lot of (self-)training effort.

But luckily there is BERT. This gave us very promising results out of the box in just a few minutes. It was clear: the case worked and the effort would be low. What remains is: putting together, implementing, prompt engineering and delivery - all things that SIDESTREAM is very good at.