Interview Of The Week

Interview Of The Week: Alexander Vedyakhin, Sberbank

Alexander Vedyakhin is the First Deputy Chairman of the Executive Board of Sberbank, Russia’s largest bank. He began his career at Sberbank in 1999 as a senior cashier-controller, rising through the ranks to his current position, which includes supervision of the bank’s retail business, corporate and investment business, wealth management division, sales network and problem assets department of the bank. He also oversees the Department for the Development of Artificial Intelligence and Machine Learning, Artificial Intelligence Laboratory, Center for Partnership Funding and Special Projects and the ESG Business Unit. Vedyakhin, who has a PhD in economics, recently spoke to The Innovator about how Sberbank is scaling AI and earning $2.7 billion in operating income from AI in 2021.

Q: How is Sberbank using AI?

AV: We are addressing the demand of our clients who would like to have every product online and nearly for free. This is unachievable without AI. We have around 300 products available for our 100 million retail customers and we personalize the product offer for every one of them. On average, each customer uses 50 products a year. That means 3.11 х 1057 possible product offers! This is too big of a number to calculate. AI is the only way we could offer this number of customized services online. To create a product and a service at a low price for customers you need to have a product with a really low cost for the bank. It is a real challenge as product development is expensive, but it is possible to introduce vastly improved services at a low cost with AI. Let me give you an example.  When I joined the bank more than 20 years ago calculating loan risk for retail clients took two to three weeks. Around four or five years ago we hired special underwriters who would decide within two to three days. Now it takes two minutes, and we lowered the price because no employee is involved – only algorithms. Also, thanks to the AI the amount of a pre-approved loan is based on a customer’s income so they can use it immediately for a mortgage or another type of purchase. It is the same as a wallet. Customers can use cash in their wallet or use credit in their wallet. The only difference is that it is online. Clients are more likely to take loans if they are immediately available and personalized.  Only AI can give this type of result. There is no other choice. The change is challenging for management, but the business as well as the CEO must understand it and then the risk management, compliance and other support people need to make the proper amendments in the banking system.

Q: How much of targeted net profit will come from AI-based businesses in 2021 and how quickly does the bank see this percentage growing?

AV: In 2021 we are projecting about 200 billion rubles of operating income from AI, this equals about $2.7 billion. In 2022 it should be at least 260 billion rubles or $3.2 billion, so it is quite significant. For every 15 rubles we invest in AI we yield around 100 rubles of profit.  I would say it is a rather good investment.

Q:  Most traditional businesses are having trouble scaling AI and using the technology to achieve top-line growth.  How did Sberbank manage to do it?

AV: We started in 2017.  We have a special meeting every quarter and every department head in the bank gets an AI transformation rating – who is good, who is bad, who is making money with the technology and who is not. All the bank executives must pass AI exams. Not only the executives of the AI-dedicated departments but other departments as well. Everyone must enroll in the special corporate university program. Everybody – including executive board members – must learn to code. They are asked to choose some special issue that needs to be resolved. Once they choose it, we tell them ‘Well, guys, now do it yourself. Sit down and just code.’ They say they can’t do it, but they need to understand how to do it, they need to learn how to code by themselves. This is a real challenge for our top executives. 

We have also implemented rules around how AI is developed in-house. Imagine that you have a lot of AI developers. Some are using Google Cloud, some are using Amazon Cloud and other people are working on desktop systems. This doesn’t work. You need to have one platform and every data scientist must work on this platform.  If one of our data scientists writes really good code for a chatbot then all of our divisions must be able to use that code. We are using at least 20 chatbots across divisions but the core AI engines are the same. We also have a special platform for scoring customers and every department can use it. These platforms are uniform, but anyone can submit a change that will improve this platform for the whole company. It is like an internal open-source system.

Q: How is the implementation of AI impacting staffing?

AV: Scaling AI requires an increase in the number of data scientists. We have initiated a special program for this at the top university so we can get the best students to work for Sber. It is a big challenge. We need to compete not just with Russian IT companies like Yandex, but also with companies like Google. There are not enough graduating data scientists to meet demand. To solve this issue, we have singled out employees with mathematical backgrounds, but no data science training, and we are enrolling them into special courses to get them up to speed.

Q: What about middle management and entry level employees who are being replaced by AI?

AV: In 2019 we started to tell our employees that it is highly probable that their jobs will be replaced by AI and that they had around three years to prepare and retrain. Not everyone could or would want to learn how to code but there are other ways they can transition, for example train AI to recognize images. Image tagging is also a way for the bank to provide jobs for people in areas with high unemployment. The former head of our underwriting department understood her job was going away and became responsible for image tagging. She created an internal startup for tagging that is competing with Yandex and others and started to provide this service internally and externally, with a goal of making this division net profitable by 2023.  That said, there are some employees that are not willing to transition. They can move to jobs at other banks in Russia that are still offering services to their customers in the traditional way.

Q: What is Sberbank doing to ensure that AI is being used ethically?

AV: We do our best to make sure that the AI we deploy is ethical. It is why we have adopted our own AI Ethics Corporate Principles. We have a special department in charge of validating AI models before they are deployed to ensure that the data and the algorithms are not biased. 

While we have created rules, we know that these need to evolve along with the technology.

Q: How does Sberbank’s AI transformation fit in with the bank’s overall strategy?

AV: We are now transitioning to becoming a technology company and an ecosystem that offers much more than banking. The Sber ecosystem includes more than 65 different companies which offer a variety of services needed in everyday life such as food delivery, taxi services, video-on-demand services, streaming audio, online healthcare services, delivery of drugs and an app that lets you choose and buy cars online. We have tens of millions of active customers of our ecosystem services. Our core digital product, Sberbank Online, has 70 million MAU [monthly active users.]  All ecosystem services have one digital technological platform at the core, using Big Data and AI.

Q: What advice would you give to other companies that are trying to scale AI?

AV: Our AI transformation is based on seven pillars. The first is strategy and management.  The CEO and the board must start and manage the AI transformation process. It is a long, difficult way to go but there is no other way. It needs to go down from the top. If you just put a couple of people in place and set some targets it won’t work.  The CEO must really be hands-on.

The second pillar is data. Without good data there can be no good AI. If you put garbage in, you will get garbage out.

The third pillar is the creation of models that allow for quick testing and validation and establishing special procedures to improve and revise them over time. With our model framework it is possible to develop an AI model in the morning and get it into production the same evening, providing it is not high risk.

The fourth is infrastructure: identify what you will use and make sure it is uniformly applied.

The fifth is process. Our bank has approximately 2000 defined processes and our goal, by the end of this year, is to embed AI in more than half of them.

The sixth is people and culture.  My advice is to make necessary staffing changes and retrain existing employees to ensure that your company is competitive.

The seventh is research and development. Make sure that what you are doing is cutting edge. The Chief Scientific Advisor of our AI Institute is Jürgen Schmidhuber, one of the most respected computer scientists in the world. His lab’s work in the early 1990s is credited with revolutionizing machine learning and AI.

These enablers were identified at the start of Sberbank’s AI transformation program, now they are being used to build our AI Maturity Index, the methodology Sber uses to measure the dynamics once a quarter. The results are presented either to the bank’s CEO or at the bank’s board meeting.

Finally, keep in mind that the AI transformation does not take place overnight. It has taken us three years to generate significant revenues with AI and it requireda big push from the CEO and the management board.

This article is content that would normally only be available to subscribers. Sign up for a four-week free trial to see what you have been missing

About the author

Jennifer L. Schenker

Jennifer L. Schenker, an award-winning journalist, has been covering the global tech industry from Europe since 1985, working full-time, at various points in her career for the Wall Street Journal Europe, Time Magazine, International Herald Tribune, Red Herring and BusinessWeek. She is currently the editor-in-chief of The Innovator, an English-language global publication about the digital transformation of business. Jennifer was voted one of the 50 most inspiring women in technology in Europe in 2015 and 2016 and was named by Forbes Magazine in 2018 as one of the 30 women leaders disrupting tech in France. She has been a World Economic Forum Tech Pioneers judge for 20 years. She lives in Paris and has dual U.S. and French citizenship.