Sberbank, Russia’s largest lender, has stepped up its investment in digital transformation to drive business efficiency.
Sberbank is investing more than 450 billion roubles in IT in the coming three years, as it continues its push to develop domestic technology solutions.
Technological transformation allowed Sberbank to increase the efficiency and reliability of IT systems, replacing foreign software with our own developments and innovations, which we already offer to the market.
Sberbank CEO Herman Gref has placed artificial intelligence and digital transformation at the heart of the bank’s new three-year strategy. The bank expects profits from artificial intelligence (AI) in 2024-26 to be about 800 billion roubles.
Sberbank has launched the multimodal neural network GigaChat, weaved in all Sber smart devices. On top of this, Sberbank has significantly progressed in the training of neural network, which communicates with the user, generates images and writes program code on a more advanced level.
The number of GigaChat users already reached 2.6 mn people. Its generative model Kandinsky 3.0 now creates more photorealistic images, detailed artistic paintings and has become one of the most popular in the world by the number of requests that exceeded 65 million.
“Our goal to improve efficiency across all processes is unaltered, in particular, by leveraging AI. In 2023, economic result from AI integration reached RUB350 bn. For the first time, our cost-to-income ratio came at below 30 percent, enhancing foundation for further business transformation and innovations,” Herman Gref said.
Sberbank Group’s Strategy 2026 outlines a new human-centered business model aimed at unlocking human potential by means of the next generation AI technology.
The direction towards the development of its own platform made it possible to replace IT solutions from the leading foreign vendors. Sberbank does not reveal the name of the foreign technology partners.
Generative AI GigaChat, a service that can communicate with users in a dialogue format, write code and create images, surpassed ChatGPT-3.5 in the quality of answers in Russian and English language based on MMLU (Massive Multitask Language Understanding) exam results, Sberbank said.
In addition, the GigaChat API software is available to business for their own solutions and process optimization.
Number of GigaChat unique users has already reached 2.6 million.
In order to save developers’ time for product creation and improvement, an AI assistant GigaCode is used. It accelerates the process of writing code, offering the most likely and relevant options for continuing the code in real-time.
GigaCode helps Sber developers write more than of the 20 percent code every day.
In 2023, Sber offered to users a generative model Kandinsky 3.0 that creates photorealistic images and generates full-fledged artistic paintings. The model understands requests in 101 languages. Kandinsky 3.0 is integrated into GigaChat and can generate images from a dialogue. Kandinsky Video is the first Russian model for video creation based on text description.
At the beginning of 2023, business users of SberJazz video conference service got access to the function of automated results summary of video calls, implemented on the basis of the Gigachat neural network model.
GigaChat in SberJazz will automatically prepare an informative summary of a video call: key topics of the meeting, tasks and notes. As of now, 55 tools and components of Platform V are registered in the Russian software registry.
The number of monthly users of virtual assistant Salut rose to 23.0 mn in SberBank Online and other Sber digital surfaces. Sales of Sber smart devices rose by 2.4x in 2023, reaching 1.7 million. The number of devices sold amounted to 3.0 mn from the start of sales launch.
The number of technology implementations and changes increased by 44 percent to 2.3 million in 2023, while the impact of technology changes on incidents decreased by 2x.
Financial effect from AI technology implementation exceeded RUB350 billion in 2023, while for the last three years financial effect exceeded RUB800 billion.
In 2023, Sber, together with the AI Laboratory and the Government of Moscow, developed and applied diagnostic assistant model in all adult medical clinics in Moscow.
Sberbank said the share of retail product sales through digital channels increased by 6.0 pp in 2023 to reach 67 percent. The share of consumer loans issuance through digital channels increased to 93.7 percent (+4.4 pp-for 12M 2023). Sberbank has 108.5 million active retail clients and 3.2 mn companies.
The number of monthly users of virtual assistant Salut rose to 23.0 mn in SberBank Online and other Sber digital surfaces.
Sberbank has reported a record annual profit of 1.5 trillion roubles ($16.3 billion) for 2023, a more than five-fold increase on the previous year.
Decisions on 60 percent of corporate loans issued by Sberbank will be made using artificial intelligence (AI) by the end of this year, Herman Gref said.
Herman Gref has placed artificial intelligence and digital transformation at the heart of the bank’s strategy.
Sberbank is investing more than 450 billion roubles ($5 billion) in IT in 2024-2026 and expects profit generated from AI in that period to amount to about 800 billion roubles.
Sberbank’s lending decisions on consumer credit, mortgages and car loans are all being made by AI, and Herman Gref said the figure for corporate lending was 40 percent.
“The figure should amount to 60 percent by the end of the year and by end-2026 the figure should reach 80 percent,” Herman Gref said at a festival in Sochi.
Rajani Baburajan
