AI in Credit Management – How Will Working Practices Change?

Generative AI, as used by ChatGPT, is currently enormously popular, but this is only one of many AI‑based is approaches. In various industries it is assumed that artificial intelligence will fundamentally change people's way of working. In this analysis, Prof. Matthias Schumann explains how this will affect credit management specifically.

Already today, IT solutions in credit management automate processes such as the granting of credit lines and payment terms based on automatically generated ratings. Whether a human decision-maker intervenes in the process depends on the authorizations given to the IT solution and the legal framework such as data protection regulations. This development has also led to increased use of AI‑supported processes to predict the risk of default. These can make use of additional data sources such as social media or employee portals in order to increase the quality of the predictions.

The utilization of AI presents new possibilities, for example in the forecasting of payment dates on the basis of customer profiles. Such methods not only improve risk evaluation in relation to the customer but also enable more precise predictions of when payments will be made, which can be used to improve the planning of cash flow. The results obtained primarily serve to support decision-making and are less significant for the complete automation of processes, especially in scenarios of negative developments.

Automatic learning by AI systems faces challenges, especially when decisions made by humans are overruled. It is critical to know whether the decision is based on internally available data or external information and whether the decision can be considered as correct at the point in time it is made. These factors influence whether, and how, the system learns from such decisions. An example: If a customer is rejected by the system but is later accepted, there is a risk that this customer will default. In addition, it is not possible to find out whether a rejected customer would actually have been able to pay.

There will also be new dangers resulting from the fraudulent use of AI, for example in relation to dealing with company shells or questionable company liquidations. In order to meet these challenges, suitable instruments for the recognition of such activities need to be developed.

Ultimately, this is a question of the fundamental requirements for a successful AI solution. An essential prerequisite for this is the availability of sufficient training data. In order to avoid errors in the results, it is important, for example in a prediction of insolvency cases, to train using an approximately equally large number of "good" and "bad" customers. This necessity indicates that the primary application of AI-supported solutions is probably with business information agencies because these have the necessary large amounts of data available.

Finally, it can be said that AI-supported solutions in credit management will lead to improved support for decision-making and that automation will slightly increase. It is not, however, expected that there will be fundamental changes in this sector. The time gained through the use of AI will instead offer the opportunity to consider decisions more thoroughly and to react more agilely in an increasingly complex global commerce situation.