“Banking Is Necessary, but Banks Are Not”

This quote by Bill Gates has been recently proven to be correct by the pandemic that accelerated digitalization of all bank processes.

How can AI and ML implementation help any bank to stand out from the future competitors?

Every year AI and ML solutions bring in more and more of the income in retail banking. Without algorithms, which tell us what and when to offer to the client, we will lose clients to those banks who have such algorithms. And it will be the battle of ML-solutions. Those banks with more data and better data mining get better model accuracy, the best model accuracy gives more clients, more clients give more data and so on. Therefore, timely work with ML-solutions and building the right architecture and processes for the rapid implementation and testing of such solutions give to banks an opportunity to survive in competitive environment.

Where do you see the future role of ML in redefining retail banking field?

In 1994 Bill Gates said “Banking is necessary, but banks are not”, and COVID-19 confirmed it. COVID-19 accelerated shift to digital without bouncing back because it saved clients time and reduced banking costs. ML will continue the trend of reducing costs and improving client experience. For the client the key to the choice of the bank will be how we save his time (in case of equal service costs). In limit we should build zero-steps processes performed automatically without client participation. To do this we should predict the client’s needs, their next best action, and act for them, becoming «banking for a client». And ML-solutions will be basis for it.

How does the next-generation banking automation look like and what potential does it hold?

ML-solution are playing more and more important role in banking. But banks are not an AI-native companies (like FAANG), and by the time AI became popular banks already had had their established automated banking systems. In this moment banks (and other non ITs companies) had faced difficulties in implementing ML-solutions in their processes – long, expensive and sometimes impossible. There are three key differences between the introduction of ML-solutions and the classical DevOps of new features:

1. To provide an environment where the model can be executed and delivered
2. To deliver current data to the model
3. Model results need to be interpreted and delivered to the consumer

Each of the above-mentioned tasks are solved by its own *-Ops: ModelOps, DataOps and DevOps. Gartner identified XOps = DataOps + ModelOps + DevOps as one of the top trends in data and analytics for 2021. That is why the next-generation banking automation will look XOps methodology and banks will start to become AI-native like FAANG.

---

Alexander ULYANOV, Head of Retail MLOps at Sberbank is a senior executive with more than 15 years of applied analytics and change management in banking with a broad experience in artificial intelligence and machine learning solution development and implementation in production. Launched innovative AI-solution on detection clients’ customer journeys in retail banking, which improved campaigning’s response and clients’ experience due to appropriate time and goal of communication. Realised MLOps transformation for launching high load real-time recommendations in retail banking.