Cookie settings

Cookies are small text files stored on your device to identify you and can be used to remember user preferences and analyse traffic to further improve our website. We may share information about your use of our site with our social media, advertising and analytics partners. By clicking "Accept all cookies", you agree to the use of all cookies as described in our cookie statement or "Accept only essential cookies" to only use cookies that are necessary for the functioning of our site.

Read our cookie statement here.

You can choose to adjust your preferences at any time.

Wholesale Banking

Banks do not yet use Artificial intelligence (AI) for lending – at least not in large-scale corporate banking. But there will be no way of avoiding the widespread implementation of AI in the future, even in wholesale banking, as wholesale clients expect increasingly rapid response times from banks with as little active effort on their part as possible. The fact that banks involved in corporate banking can lend quickly is therefore a key competitive factor. The pressure to introduce faster, more streamlined and more efficient processes is increasing for cost reasons as well, not least because one day it will be possible even for wholesale clients to request and be granted loans via virtual platforms.

AI or machine learning is playing a key role in this. Retail banking already provides some guidance on how this might work, but this is comparatively much better developed, as lending decisions have been made based on automated processes with the help of statistical models (mostly logarithmic regression) for many years now. Admittedly, even in retail banking the use of AI is still very much in its infancy, despite the degree of automation being already much more advanced. But when it comes to loans for wholesale clients, credit analysis is still being carried out by humans in the traditional manner – using substantial resources and incurring considerable costs, in particular for customers.

Elsewhere, many banks have already started using AI-supported early warning systems. These applications use hundreds of newspaper articles and other sources of information to create daily sentiment analyses. Based on these, signals are then generated for the relevant account manager regarding potential issues on the borrower's side.

Challenges with AI

The challenge for banks, which explains their reluctance, is that they need large amounts of data and high numbers of cases to be able to develop meaningful models. This data is so limited that it is virtually unobtainable. There is even less data available for loan defaults, especially at individual bank level. This means that in the future there will be no choice but to purchase this data from external providers who have access to global default data.

Moreover, implementing technological solutions of this kind is time-consuming. Banks need to identify and acquire the necessary IT and database specialists in good time. It is also unclear how many and which of the platforms mentioned above will exist. While it would be desirable for platforms like this to be developed as a result of joint industry initiatives, it is likely that they will be established by vendors from outside the industry instead.

Despite the expenditure, corporate clients could benefit significantly from the introduction of AI. The bank would then allocate each company a maximum limit for immediate loan commitments in advance. AI will be the main requirement for this due to the need to generate and evaluate the "right" data. Banks will only be able to become part of future credit platforms like this if they meet this requirement. Companies would be able to satisfy their loan requirements faster than ever before – almost in real time.


Georg Hauser is managing director of credit risk management at ING in Frankfurt.