Artificial Intelligence to empower data-driven Private Banking

Today, information is a key differentiator, and financial services institutions are fully aware of this. Artificial Intelligence (AI) and Machine Learning (ML) will remain a hot topic in financial services. 85 of 100 interviewed decision-makers in our international, cross-industry study on AI rate the potential of Data & AI projects as high. Another study predicts a yearly 1 Trillion USD contribution of Data & AI in banking in the coming years.

With AI and ML to a competitive edge

Banks are facing increased competitive pressure from fintechs and neobanks. It has become clear that new digital, client-centric services are a necessity to serve as a value-add as well as to bolster customer retention and growth. This effect generates a lot of pressure on marketing and sales. The good news is that digitization has paved the way for data-centric ML solutions that help to cope with that pressure. Hence, many banks leverage AI technologies to gain a competitive edge.

We embarked on such an endeavour with VP Bank, headquartered in Liechtenstein, to support their client advisors with a decision-support system. Client-advisors, among other things, provide investment advice to their clients. Client advisors rely on investment recommendations from the research team and are often supported by dedicated product specialists and portfolio managers. This requires in-depth knowledge about financial markets and a good understanding of client needs (risk appetite etc.). As the bank acquires more clients and offers more products, complexity grows.

Our project aim was to support the client advisors in the recommendation process and to enhance targeting of customers. In general, such ML systems are not going to replace a human but rather act as decision-support to empower the human. Think of how calculators are instruments of decision-support systems, imagine being productive without them. Think of such investment recommenders as intelligent calculators.

We started the project with a discovery phase. Together with senior management from the bank, our team quickly identified the following as a high-potential business case: VP Bank has released investment ideas for many years, the main challenge was to target the right clients. Investment ideas have to fit to the clients portfolio strategy and pass compliance regulations. In the picture below one can see fictitious examples of investment ideas, as they can be found in our recommender system.

Data science teams from Zühlke and VP Bank domain experts working closely together

With the investment story (“Alexa, protect me”), the full potential of the investment recommender unfolds, as can be derived from the picture below.

On the top left we see a description of the investment idea with the option to download a PDF with a more detailed description of it, ready to send out to the customer.

On the bottom left we see all client portfolios, compliance-filtered and ranked according to the match of the portfolio with the current investment idea. The main technical challenge in this part was to model and predict how well a client’s portfolio and potential client interests match with a particular investment idea. Our data science team built and evaluated machine learning models, closely working together with the domain experts from the bank.

Building up trust as a crucial element within the process

Additionally, we managed to build a compliance engine, which automatically filters the clients which are not compliant towards the investment idea. This process automation frees the client advisors from a tedious and repetitive task. The client advisors can now contact the best matching clients with this investment idea. In the top centre we see a reasoning for the investment recommendation that can be used for client communication.

It was very important to build up trust for this ML solution. We addressed this challenge by involving the client advisors in the project as early as possible, as well as by providing recommendation reasoning in a transparent, user-friendly front-end. Furthermore, investment recommendations show both rule-based results, which were derived from current practice, as well as the ML recommendation, so the client advisors can compare the different portfolio matching scores on the spot.

Another important feature is the list of investment instruments that are contained in the investment idea. It is at the client advisors discretion to make the final decision of recommending each instrument. To round off the middle section, the “VP Sustainability Score” of the current customer portfolio can be seen.

On the right-hand side, the client advisors can see the portfolio of the client: an asset-class overview, the top three holdings and the latest transactions enable the client advisors to have an informed and professional customer interaction. It is very important to make sure this intelligent solution improves itself: With the two buttons on the bottom-right we added a possibility for the client advisors to provide feedback regarding the recommended matches. This knowledge is stored and curated. This enables to further improve the solution and to become more robust to different real world edge cases.

Limited window to leverage Machine Learning

According to the Gartner CFO and Finance Executive Conference 2021, companies will have a limited window to truly leverage ML for competitive advantage. Most companies are prepared to invest in Data & AI, but they need to also bet on truly transformative projects rather than just modernizing existing processes for long-term ROI impact.

We at Zühlke believe in a balanced strategy of investment in Data & AI to instil trust and to ensure growth. If you are interested in learning how we support businesses to define this strategy and help with the implementation of both bold transformative and steady modernization solutions, reach out to us.