Machine learning in Finance

In the beginning of 2020 Revolut raised USD 500 M at Round D funding, valuing the company at USD 5.5 B. N26 was valued at EUR 3.5 USD in 2019 and announced further investment rounds in 2021, expecting it to go above that level. Deutsche Bank’s valuation during most of 2020 has been bellow USD 20 B. Why is it that one of the old European banks, founded in 1870 with strong presence in several continents, over 80 000 employees and total assets of over USD 1.6 T in their balance sheet has just 4-5 times the value of new challenger banks with much limited operations and assets under management?

The answer like anything in tech lies in the future. While traditional banking has been relying on slow, gradual and controlled service development and go-to-market strategy, the tech sector operates with a different philosophy. Doing more with less, utilizing mainly technology and less “hard” assets, expanding fast, everywhere, in every way possible is the way it goes. Whether traditional banks realize the threat is beside the point – this change is coming, and it is coming fast.

So what does this have to do with machine learning? For one, challenger banks and financial institutions are keen on using AI and machine learning to handle key tasks, that were done in different ways before. Here are three of these:

  • Managing fraud detection. While traditional approaches have been to employ rule-based systems (e.g. entering 2 wrong pin’s is okay, but after the 3rd attempt your card is blocked), a new approach could be to look into the user behaviour and make a decision on whether to accept a transaction based on the behaviour patterns seen before. If you for example typically purchase groceries in a small supermarket in rural Germany, what are the chances that one day after you will purchase a shirt from a shop in Brazil? The rule-based approach does have its advantages, but has its limitations in cases of complex attacks. The AI route is not an easy one either. For example there have been numerous report from Revolut users that their accounts have been blocked without clear reason why and the customer support is very slow to respond to their requests (weeks slow). The reasoning for delays can perhaps be found in AI money laundering procedures, which indicates machine learning is necessary, but the implementation and management are key for good user experience.
  • Customer service. Monzo, founded in 2015, has managed to reach over 1 in 20 adults in UK in just 5 years, with total of over 3 M UK Customers. Considering the short life of the bank, the accounts were likely acquired from incumbent banks already well established. In 2019 Monzo stated that (at that time) they believe chatbots are overhyped and a combination of Machine learning and a human are a good way to address communication. When a customer writes a complain, AI scans system data and generates possible responses to the support team. In addition they do a weekly analysis of the complaints to decide on features to implement next. It is a perfect way to learn from the problems you face and make life easier for your customers.
  • Decide on which chocolate bars to order. In all seriousness, there was a post in n26 magazine, that tried to solve the problem of which chocolate bars to order based on the available data. You have four types of bars and consumption information, but is it enough? Perhaps weather information will help in that decision, perhaps you found out that rainy Mondays bring an interesting pattern of high consumption of one specific type of bars. Or perhaps you will find some other interesting pattern, relationship or correlation. In any case the conclusion is simple – use AI to continuously learn from your data and create a better product for your users – both external and internal.

To summarize, we see machine learning and AI as a major market opportunity and a major threat for the traditional financial industry. We believe it is here to stay and will slowly (or not so slowly) transform the industry function by function, service by service, market by market, leading to a new market dynamics and environment in just a few year-time. The future is here and now, the question is how you can be part of it.