Experts of Capgemini Consultancy Company believe that automated systems cost banks 50–90% cheaper than the labor of hired employees. Besides, artificial intelligence improves quality and speed of customer care, especially when analyzing the creditworthiness of a customer.

Advantages of artificial intelligence for banking

Customer care is the main area where artificial intelligence solutions are used. They are chatbots, round-the clock customer support, analysis of transactions and loans.

Advantages of using AI systems in the financial field include:

  • automation of routine processes;
  • increase of customer service rate;
  • reduction of costs related to the solution of standard tasks;
  • improving of accuracy in big data processing;
  • improving of service quality.

As an example, the activity of the Russian Sberbank can be mentioned, which actively uses new technologies and implements AI in all business areas. Specifically, the bank has automated the work of the contact center for corporate customers. Now, chatbot Anna answers their questions and the service rate has increased by 50% (in basic themes).

It does not mean that artificial intelligence will take away workplaces: bots are designed to do routine and repetitive work, whereas ban employees will have more complicated and creative tasks.

Taking into account how actively AI is integrated in banking, the Fintech Association has taken on developments in the field of artificial intelligence and big data. For example, big data is applied for credit analysis – scoring.

AI Conference: Artificial intelligence issues credits 1

Artificial intelligence and scoring

Credit scoring is the assessment of customer creditworthiness and striving to pay off debts. Conclusions are based on a variety of data: total income, credit history, transaction analysis, and even employment history.

In fact, scoring is a mathematical model that relies on statistical methods and tracking of big data volume. Artificial intelligence and big data help to resolve such a task quickly and efficiently.

Uplift models for Binbank

These technologies help to assess debts and conduct credit analysis of customers: now solutions are implemented in the work of the Russian banks. For example, Binbank was one of the first to use the machine learning technology to analyze delinquencies in retail.

Experts of Binbank build self-learning uplift models that predict the response of customers to debt recovery. It allows optimizing the process and thinking over how to improve communication with separate loaners. Specifically, uplift models detect customers calling to whom and reminding of repayments is useless, form lists of customers that make repayments themselves (calls are disturbing them in vain), and lists of people that require reminders. The project uses Python programming language to analyze data.

The Gini coefficient (index), a universal scoring metrics, is used to assess the accuracy of predictions. On the stage of testing self-learning AI models at Binbank, the main coefficient has grown from 65 to 88%. Earlier, the bank used logistic regression models.

AI issues credits at Sberbank

Back in late 2017, the management of Sberbank announced plans to issue most of credits to private individuals basing on the solutions of artificial intelligence. It was reported by the senior vice-president Alexander Vedyakhin. According to him, the AI system makes the decision about crediting within one minute.

At the same time, a parallel testing is conducted with the participation of people that also make decisions on loan issuing. It is required to evaluate the efficiency of AI performance. According to representatives of Sberbank, all issues related to crediting will go to the area of responsibility of AI with time.

Meanwhile, the results of artificial intelligence performance are satisfactory: the level of loan delinquencies has decreased (as compared with the period when an employee only was responsible for decision-making).

Besides, the management of Sberbank is against mass dismissals: they will send employees earlier involved in underwriting to professional retraining (including data scientists).

Smart scoring systems

LIME has developed its own innovative scoring system. It stands out for unusually high speed of operation: the system analyses 10,000 characteristics of a customer within several seconds. Specialists of LIME state that the model was designed for full automation of credit processes. Moreover, it can increase income in terms of provided loans.

The distinctive feature of the system is that it uses not only the credit history but also behavioral data: for how long the customer fills in the form, thinks over answers, how much time separate actions take and everything in total. Specifically, it helps to detect fraud schemes on early stages.

The Belarusan startup GiniMachine came up with a similar development. Their system provides a solution for credit scoring based on machine learning. It develops analytical models, calculates credit scores, and analyzes risks of specific loaners. According to developers, this algorithm can save weeks of human work. GiniMachine system can also solve other business tasks connected with forecasting.

Conclusions

Artificial intelligence in credit scoring can save time and general expenditures of the bank. For example, AI helped Binbank to improve the accuracy of the Gini index. In such a way, decisions about loan issue are taken more efficiently and the bank does not lose money on dishonest loaners.

Alexander Vedyakhin of Sberbank believes that the AI technology designed for crediting of both private individuals and enterprises will become popular at large financial institutions in the coming years. Among other things, this direction is actively developed as part of the Russian Digital Economy program.