Big Data

Big Data and AI in Banks: Trend or Real Tool?

Pinterest LinkedIn Tumblr

How and why do banks introduce new technologies

When do banks start implementing artificial intelligence and Big Data? What are the prerequisites for such projects?

– If you follow the hype or technological fashion, the project is usually doomed to failure. Therefore, in the banking sector, only technologies are introduced that are recommended or required by the regulator, or that have a commercial effect.

Our bank is also a commercial organization, we are ready to consider and implement any technologies that will lead to their further monetization and improvement of customer service. Each technology is audited for financial viability. If it gets a positive assessment, we start a pilot implementation. 

We focus on market practice and ready-made cases, since it is easier to assess the feasibility for solutions already accepted by the market and approved by the market.

Sometimes the reason for introducing new solutions is the requirements of internal banking processes, for example, if we see some kind of gap that can be closed by this or that technology.

What does your AIDA division do?

– AIDA is an abbreviation that stands for Artificial Intelligence and Data Analytics Office. The name speaks of the division’s specialization: artificial intelligence technologies and data analysis. The group is engaged in data analytics, advanced analytics, machine learning, deep learning. The AIDA division gathers all the bank’s expertise in these two areas.

Since URALSIB is the same commercial organization as other banks, the innovation division is focused on obtaining economic benefits. The introduction of new technologies is not an end in itself, we strive to improve business processes and increase profits.

Do you already have practical results of the department’s activities?

– AIDA creates tools and implements methods for monetizing Big Data, extracting commercial value from a huge array of data that forms the bank’s front-end systems. The target function of the department is to maximize the value of data.

For more than twenty years, banks have been developing expertise in risk assessment and analysis of CRM (Client Relationship Management) – this is the study of clients, their transactional activity, profiles and other parameters in order to increase the marginality of client activities.

In addition, in any bank there are many other areas where the commercial value of data can be found. 

Wherever there are numbers, comparisons can be made: “more”, “slower”, “unprofitable”, “more profitable”, respectively, mathematics and optimization are applicable there.

There is a huge reservoir of data and directions that increase the efficiency of business processes. These include cash management, transactional analytics, financial monitoring analytics.

The most active development of mathematical analytics was in the retail segment, but today we are also developing data analytics for small and medium-sized businesses. This potentially brings huge monetization opportunities.

What are the AIDA Performance Indicators?

– The monetary effect of the work is compared with the costs of maintaining the unit, and they pay off many times over.

AIDA helps business expansion, more intensive, quality growth. As a result of the division’s work, the client base is growing, the quality of clients and their margins are growing.

A good effect is observed in the area of ​​risk assessment. The better the risk management and segmentation of borrowers into reliable and prone to default is established, the better the bank’s performance. It depends on AIDA: the fewer bad borrowers we “let in the bank”, the more profit will be. 

In addition, URALSIB business units are implementing recommendation systems based on artificial intelligence methods. Customers can already see them in their RBS (remote banking) systems. Users do not even realize that a complex mathematical model is working “under the hood”.

AI initiatives and applications

How can you characterize the initiatives related to the national concept for the development of artificial intelligence?

– Today, many are developing artificial intelligence – manufacturers and developers of instruments, scientific institutions, research institutes. But there is no unambiguous understanding of the term AI yet. Once the terminology is settled, it becomes clear whether it needs to be developed at the national level.

Many people divide artificial intelligence into weak and strong. In my opinion, a strong artificial intelligence is a Terminator that is capable of emotions, self-awareness, judgments and thoughts, and a weak artificial intelligence is the ability to automate certain tasks.

Weak AI can listen to audio and generate text, verbalize the results, extract information from an audio or video signal. Such automation is at the level of duplication or replacement of a real employee.

Today we use technologies that are able to understand human speech and respond to the interlocutor, recognize numbers and letters, enter them in the necessary columns and columns, and I tend to call all this weak artificial intelligence. Such developments will develop, because we are only at the beginning of the path.

But today such solutions are not widely used in the financial sector. What is the reason for this?

– It is not enough for the technologies to be developed. The business must figure out how to monetize them. And such opportunities are not always available. Ranking all kinds of initiatives, the bank must come to the conclusion about the high investment attractiveness of projects with AI. So far, I cannot say that in terms of payback, these technologies are being implemented as a priority. Many in the market talk about this, launch pilots, but it often turns out that it is possible to introduce another service, not related to artificial intelligence, and get more margin.

New Trends in Banking Technologies: Psychoanalytics and Emotional Assessment

How do banks use video analytics and emotional assessment technologies?

– Video analytics can be used as a way to extract information from a video stream, for example, from surveillance cameras. In China, methods for detecting emotional states are already being used: if video analytics systems detect the nervousness of a client who has entered a bank office for a loan, this data is taken into account in scoring models. So AI helps to suspect fraud or a high probability that a person will not be able to repay the loan.

And in Russia no one has yet proposed a really working model of intellectual psychoanalysis?

– There are attempts. We are developing psychoanalytic services, detecting and determining the psychotype of bank clients, extracting psychometric parameters. Much attention is paid to the issues of digitizing psychotypes, although so far in the form of experiments.

What interesting projects are your developers doing right now?

– For example, we are piloting a speech analytics module. It avoids the overhead associated with recording audio by converting it into text. We try to make sure that the speech analytics module picks up intonation. Emotions recognition, emotion recognition is the direction of artificial intelligence technologies. This is useful in contact centers, especially in collection agencies, where nervous interlocutors meet on the other end of the line.

What problems arise when introducing innovations

Are all innovations easy to implement? What are the main reasons for the difficulties?

– The main reasons lie in the psychological plane, so it is difficult to introduce any cardinal innovations. A more efficient business process means that a computer can perform better than a human. People who are less effective as a result resist.

We purposefully popularize the direction of artificial intelligence, show the advantages of new technologies on pilots, try to convey to the management the position that innovations should be dealt with and show their benefits to employees.

To what extent can new technologies replace humans? Do the bank’s clients still need live communication?

– There are banks that operate without offices, I think this is an effective business model. But the segment of banks, which contain numerous additional offices, is also quite large, in the near future its volume will remain.

Our bank is not going into digital communication yet. We work both online and in the physical world. Additional offices are in demand, a significant part of our clients interact with the bank only through them.

In terms of technical issues, do you face the problem of data cleansing, initial data collection?

– Dirty data is a problem for any business. But what is considered clean and what is dirty is a philosophical question. Most often, you have to work with the data that you have. Sometimes there is not enough information, there are many gaps in the data, there are anomalies and outliers.

In the real world, everyone faces this. To improve the quality of the data, we control the “garbage”, try to clean the data, and develop this direction.

Do banks have a need for some kind of external data that could significantly improve the quality of work in the financial market?

– We actively use external data, we buy them. The external data market is already well developed. We even use data that the Federal Tax Service publishes on its website. They enrich our machine learning models, make predictions more accurate, and make management decisions better.

In-house development or external solutions: what banks choose

In-house development is popular in many large companies, including banks. Why is this so?

– In banks, we deal with banking secrets – commercial information that cannot be transferred outside the organization. You can work with it only using bank funds, banking equipment, specialists must have the appropriate competence and admission.

In those areas where there are open APIs, we use different possibilities, but when data cannot be transferred, we develop only internally.

To what extent is the external market for solutions for the banking sector sustainable? Do banks have a need to take something from outside, not using in house where there are no internal restrictions?

– There are a lot of offers, sometimes you can get confused about the possibilities that the products of certain vendors provide. There are no obstacles to their use. However, there are many open-source products on the market, and they are quite developed and functional, so you don’t need to buy expensive commercial data analytics products. The only question is the availability of competencies to work with them. We develop them today – we buy them on the labor market.

Also, banks are not very fond of using public clouds, although some are already moving in this direction. So far, we keep almost everything on internal servers, based on information security requirements.

But is there already an interest in external solutions?

“ Analytics cloud services are very attractive, powerful and functional. Instead of maintaining and configuring your hardware, supporting the required versions, it is easier to use a ready-made solution, for which there is enough Internet and expertise. Many of the administrative competencies would be unnecessary if we could use cloud-based analytics services.

Sooner or later, many banks will come to cloud services. Most likely, this will happen when the issue of information security is resolved. For example, using blockchain or some other technology.

Banks and digitalization: three main trends

  1. In the financial sector, new technologies are introduced only if they have commercial value or are required by regulators.
  2. The main areas of application of Big Data and artificial intelligence are risk assessment and work with client profiles to increase margins.
  3. Banks are ready to switch to cloud services if information security issues are resolved.

Feature Image Credit