February 19, 2019 at 16:40Half the kingdom for AI: how much banks save on machine learning, neural networks and chat bots
Assessment of creditworthiness by customer profile on Facebook, robots for collecting debts and financial advice to investors, the fight against fraudsters and the fight against routine - banks need artificial intelligence in almost all areas. About how AI helps Sberbank, VTB, Tinkoff Bank and other financial organizations save billions of rubles - in a review of the Binary District.
According to the forecast of the research company Autonomous Next, by 2030, banks around the world will be able to reduce costs by 22% with the help of artificial intelligence technologies. Savings could reach $ 1 trillion.
Russian banks are already earning and saving considerable amounts with the help of AI. So, in 2017, Sberbank earned an additional $ 2-3 billion (the bank’s net profit for 2017 was about $ 11.6 billion) only through the use of AI and data analysis in managing risks and sales.
We selected seven tasks that banks solve with the help of artificial intelligence, and looked at how it benefits them.
Credit scoring is the most promising area for the implementation of AI. Its capabilities in this area were used by the majority of Russian banks polled by the Expert RA rating agency in 2018 (11 banks took part in the study: Tinkoff, Gazprombank, MTS Bank, Moscow Credit Bank, Russian Standard Bank, etc.).
In Sberbank, AI is already making 98% of decisions on granting loans to individuals. Credit risks are analyzed based on the user's “digital footprint”. According to the head of the bank, German Gref, this trace is already reaching 500 MB per day, and on its basis a "second digital" I "is formed, which" very accurately repeats our human "I".
It’s more difficult to evaluate credit risks with legal entities: here, AI can accept only 30% of extradition decisions.
The second popular area of application of AI in banks is collector robots. Sberbank was also a pioneer here: in 2016, he introduced a pilot project of his subsidiary Active BK. A year later, the effectiveness of the robot was almost a quarter (24%) higher than that of live operators: so more often debtors paid delays within two weeks after the car rang.
After that, AktivBK worked with 27 more banks (Otkrytie, Binbank, etc.), in 2017 this area brought the company about 25% of the total revenue. In the fall of 2018, VTB introduced a collector robot after three months of pilot operation .
Post Bank was one of the first to introduce biometric technologies in its branches in 2015. Now more than four thousand branches of the bank and 50 thousand stores of the bank's partners in POS business are equipped with a face recognition system. Two-factor authentication - by login / password and photo - is also necessary for bank employees to gain access to the CRM system and other business applications.
In 2016 and 2017, this saved Post Bank a total of 3 billion rubles: in 2016, the bank received 9.2 thousand fraudulent loan applications worth 1.5 billion rubles, in 2017 - about 10 thousand applications for same amount. The system helped to identify who received these applications. The results for 2018 have not yet been announced.
Alfa-Bank in 2018 was going to replace people with robots in thirty routine business processes. After the automation of the first seven processes, an annual savings of 20 million rubles were achieved. As a result, the bank planned to save up to 85 million rubles annually.
The Bank handed over to robots such operations as processing payments of legal entities and individuals, processing unidentified payments, parsing internal incoming mail, changing customer data on its application, editing credit agreements of individuals according to their applications, as well as posting contact financing and answering standard requests.
To work with robotic programs, Alfa-Bank used the Blueprism platform (a three-year license costs less than a million rubles). Each robot receives a virtual workstation, on which the Blueprism agent and the software necessary for work are installed. Further, the system is trained by a person familiar with the business process of the bank and with the technology of training robots. Prior to this, the operating staff was supposed to grow by 3.3%, but in the end the bank decided not to hire new employees.
Robo-advising is another area that Russian banks have become more interested in since last year. One of these advisor robots for its Tinkoff investment brokerage platform was launched by Tinkoff Bank in July 2018.
For the first month after the launch, according to the bank, 42 thousand people used the application . In total, 142 thousand investment portfolios were generated during this time. The average check for the purchase of assets with the help of a robot adviser was 60 thousand rubles and 1678 US dollars. Mostly, users purchased ruble-denominated securities.
Earlier, in 2016, similar projects were launched by Sberbank together with FinEx, AK Bars Bank and VTB24 (the latter joined VTB in 2018). At the same time, their robot advisor - the Right application - was created by Conomy.
Rosbank in 2018 found another way to use AI - for the development of the retail network. About this in a column for Future Banking said the deputy chairman of the bank Arno Denis. According to him, the bank used the technology of Marketing Logic, which specializes in geomarketing.
The system developed by this company uses machine learning. She estimates the potential of the place for the new branch by 250 variables, which are divided into three groups. The first group - geo-characteristics (distance to the center, to the metro, price per square meter, etc.), the second - traffic (the number of ground transportation routes in different radii from the location) and the third - objects (the presence of a number of shopping centers, business centers, houses and banks).
By analyzing all of these parameters in the next few years, the bank plans a “significant increase” in the financial performance of the branch network. (Now the bank has 350 branches).
Chatbots are one of the most effective ways to answer questions from employees and customers 24/7. According to the results of the R-Style Softlab survey conducted in 2017, every fifth bank (21%) in Russia and the CIS was ready to use bots, and most credit organizations planned to implement them in 2018.
One of the most successful examples in 2018 was the Alfa Bank bot, which he developed for his employees-users of salary projects. Prior to its implementation, bank operators daily received more than a hundred calls from colleagues with questions about the conditions and rules for opening current accounts. As a rule, these were standard questions. After they were handed over to the intelligent bot, the operators began to answer other questions 50 times faster.
In addition to chat bots, theoretically banks can use voice assistants. This is a more complex technology, there is only one working voice assistant in Russia - Yandex Alice. In December 2018, the head of Tinkoff Bank Oleg Tinkov announced that the bank plans to create such an assistant.
To learn more about how to use face recognition, neural networks and machine learning in different areas of the business, see the two-day AI for Business course . The course speakers from Microsoft, Nanosemantics and Home Credit Bank will tell you how to use different types of AI and what tools are available for this. The nearest intensive will take place on March 30-31.
For those who want to learn how to use machine learning for different tasks - AI School. It is intended for developers who have minimal Python skills. The nearest course is from March 2 to April 6.
How many banks save on the implementation of AI
According to the forecast of the research company Autonomous Next, by 2030, banks around the world will be able to reduce costs by 22% with the help of artificial intelligence technologies. Savings could reach $ 1 trillion.
Russian banks are already earning and saving considerable amounts with the help of AI. So, in 2017, Sberbank earned an additional $ 2-3 billion (the bank’s net profit for 2017 was about $ 11.6 billion) only through the use of AI and data analysis in managing risks and sales.
We selected seven tasks that banks solve with the help of artificial intelligence, and looked at how it benefits them.
What tasks AI helps to solve
1. Check the borrower
Credit scoring is the most promising area for the implementation of AI. Its capabilities in this area were used by the majority of Russian banks polled by the Expert RA rating agency in 2018 (11 banks took part in the study: Tinkoff, Gazprombank, MTS Bank, Moscow Credit Bank, Russian Standard Bank, etc.).
In Sberbank, AI is already making 98% of decisions on granting loans to individuals. Credit risks are analyzed based on the user's “digital footprint”. According to the head of the bank, German Gref, this trace is already reaching 500 MB per day, and on its basis a "second digital" I "is formed, which" very accurately repeats our human "I".
It’s more difficult to evaluate credit risks with legal entities: here, AI can accept only 30% of extradition decisions.
2. Knock out debts
The second popular area of application of AI in banks is collector robots. Sberbank was also a pioneer here: in 2016, he introduced a pilot project of his subsidiary Active BK. A year later, the effectiveness of the robot was almost a quarter (24%) higher than that of live operators: so more often debtors paid delays within two weeks after the car rang.
After that, AktivBK worked with 27 more banks (Otkrytie, Binbank, etc.), in 2017 this area brought the company about 25% of the total revenue. In the fall of 2018, VTB introduced a collector robot after three months of pilot operation .
“So far, it is effective for short periods of delay. The average talk time is one and a half minutes, which is comparable to a conversation with the operator. If an employee makes about 200 calls a day, then for the robot this number is practically unlimited, ” said Anatoly Pechatnikov, VTB Deputy Chairman of the VTB Board in an interview with the Izvestia newspaper.
3. Fight scammers
Post Bank was one of the first to introduce biometric technologies in its branches in 2015. Now more than four thousand branches of the bank and 50 thousand stores of the bank's partners in POS business are equipped with a face recognition system. Two-factor authentication - by login / password and photo - is also necessary for bank employees to gain access to the CRM system and other business applications.
In 2016 and 2017, this saved Post Bank a total of 3 billion rubles: in 2016, the bank received 9.2 thousand fraudulent loan applications worth 1.5 billion rubles, in 2017 - about 10 thousand applications for same amount. The system helped to identify who received these applications. The results for 2018 have not yet been announced.
4. Rid of routine work
Alfa-Bank in 2018 was going to replace people with robots in thirty routine business processes. After the automation of the first seven processes, an annual savings of 20 million rubles were achieved. As a result, the bank planned to save up to 85 million rubles annually.
The Bank handed over to robots such operations as processing payments of legal entities and individuals, processing unidentified payments, parsing internal incoming mail, changing customer data on its application, editing credit agreements of individuals according to their applications, as well as posting contact financing and answering standard requests.
To work with robotic programs, Alfa-Bank used the Blueprism platform (a three-year license costs less than a million rubles). Each robot receives a virtual workstation, on which the Blueprism agent and the software necessary for work are installed. Further, the system is trained by a person familiar with the business process of the bank and with the technology of training robots. Prior to this, the operating staff was supposed to grow by 3.3%, but in the end the bank decided not to hire new employees.
5. Help clients with investments
Robo-advising is another area that Russian banks have become more interested in since last year. One of these advisor robots for its Tinkoff investment brokerage platform was launched by Tinkoff Bank in July 2018.
“In just a few minutes, according to the set parameters, the robot adviser can assemble an investment portfolio balanced by industry and company, taking into account the available investment amounts, with the optimal risk-return ratio,” the release explained.
For the first month after the launch, according to the bank, 42 thousand people used the application . In total, 142 thousand investment portfolios were generated during this time. The average check for the purchase of assets with the help of a robot adviser was 60 thousand rubles and 1678 US dollars. Mostly, users purchased ruble-denominated securities.
Earlier, in 2016, similar projects were launched by Sberbank together with FinEx, AK Bars Bank and VTB24 (the latter joined VTB in 2018). At the same time, their robot advisor - the Right application - was created by Conomy.
6. Search for a place for new branches
Rosbank in 2018 found another way to use AI - for the development of the retail network. About this in a column for Future Banking said the deputy chairman of the bank Arno Denis. According to him, the bank used the technology of Marketing Logic, which specializes in geomarketing.
The system developed by this company uses machine learning. She estimates the potential of the place for the new branch by 250 variables, which are divided into three groups. The first group - geo-characteristics (distance to the center, to the metro, price per square meter, etc.), the second - traffic (the number of ground transportation routes in different radii from the location) and the third - objects (the presence of a number of shopping centers, business centers, houses and banks).
By analyzing all of these parameters in the next few years, the bank plans a “significant increase” in the financial performance of the branch network. (Now the bank has 350 branches).
7. Answer where the salary is clear and fast
Chatbots are one of the most effective ways to answer questions from employees and customers 24/7. According to the results of the R-Style Softlab survey conducted in 2017, every fifth bank (21%) in Russia and the CIS was ready to use bots, and most credit organizations planned to implement them in 2018.
One of the most successful examples in 2018 was the Alfa Bank bot, which he developed for his employees-users of salary projects. Prior to its implementation, bank operators daily received more than a hundred calls from colleagues with questions about the conditions and rules for opening current accounts. As a rule, these were standard questions. After they were handed over to the intelligent bot, the operators began to answer other questions 50 times faster.
In addition to chat bots, theoretically banks can use voice assistants. This is a more complex technology, there is only one working voice assistant in Russia - Yandex Alice. In December 2018, the head of Tinkoff Bank Oleg Tinkov announced that the bank plans to create such an assistant.
“So far, very modest, we decided to call“ Oleg ”. But maybe we’ll change it, maybe we’ll call Ivan, ”Tinkov explained.According to him, the assistant will help users in solving financial and everyday tasks - for example, transfer money or reserve a table in a restaurant. The voice of Oleg will not be the same as that of a businessman. Other banks of voice assistants do not plan to implement yet.
To learn more about how to use face recognition, neural networks and machine learning in different areas of the business, see the two-day AI for Business course . The course speakers from Microsoft, Nanosemantics and Home Credit Bank will tell you how to use different types of AI and what tools are available for this. The nearest intensive will take place on March 30-31.
For those who want to learn how to use machine learning for different tasks - AI School. It is intended for developers who have minimal Python skills. The nearest course is from March 2 to April 6.