It is not easy to define what exactly makes a project labelled as AI; but as a point of reference, most experts perceive that these tools usually have some kind of self-learning function (machine learning, natural language processing). They are not yet widely used internationally, and the applications that we see at the client level are often based on simple analytical algorithms and do not yet provide an authentic, frictionless customer experience.
According to Szabolcs Pinter, the last decade has brought about some progress in the proliferation of connected devices, data storage and transmission technologies, cloud-based services, and the “smartening up” of algorithms, the so-called machine learning. It is then no surprise that the field is also attracting increasing attention from bank decision-makers and that more and more advanced self-learning solutions are being deployed.
There is consensus among experts that the domestic market has caught up well on an international level and that there is a growing development in our country as well. It is also true that, despite the often-voiced prejudices, the Central and Eastern European region as a whole is not doing badly regarding the development of banking AI systems compared to Western countries. However, there is of course still room for improvement for the truly cutting-edge technologies.
The successful catch-up is also reflected in the range of services offered by banks: competitive electronic channels, new payment solutions and mobile banking applications are available in our country also. Comparing this with the range of services offered by primarily digital banks, it seems that the latter offer a much cleaner product portfolio, but traditional banks are able to compensate for this disadvantage with trust capital and more complex services,” says Gabor Stren.
According to Peter Faykiss, if we are talking about the domestic market, it is better to first split it into vendor and in-house development solutions:
Despite the cautiousness and initial mistrust, there is now a clear consensus that AI has huge business potential and will be used in an increasing number of fields of application in the near future. It is already widely used, with the following areas currently being dominated by AI-based tools:
Improving the customer experience
The most common use is the improvement of the banking customer experience, mostly based on cognitive service technologies. These solutions mimic human cognition and use complex tools to achieve a level of efficiency and automation. Such tools include software developed to analyze spoken language, face and voice recognition systems, emotion recognition, sentiment analysis (using text analytics to determine the emotional value of a given content – text, image, sound).
The first generation of applications includes chatbots and other solutions that support customer relationship management – or acquisition. In many places, chatbots are still rudimentary, providing a somewhat “piecemeal” experience, responding in panels.
EXPERTS SAY THAT WE ARE NOT NECESSARILY LAGGING BEHIND INTERNATIONALLY, AND IT IS WORTH NOTING THAT THE COMPLEXITY OF OUR LANGUAGE AND THE SIZE OF THE MARKET HAVE SOMEWHAT HINDERED ANY MAJOR BREAKTHROUGH.
One of the strongest international trends is the spread of Cognitive Search, with almost all innovative financial institutions using it or planning to use it. This technology has many tangible benefits for customer services and internal operations. The essence of Cognitive Search is that it allows users to search for digital content from multiple sources. It goes beyond keyword matching techniques and returns the complete answers in a document. The technology can understand users’ needs and can offer them content that enhances the customer experience.
Other important trends include the proliferation of translation automation, multi-lingual customer journey solutions, intelligent customer scoring (next best action recommendation) and full profiling solutions.
Background processes and day-to-day work
Deep understanding can be improved, of course, and the functionality of the tools will be extended accordingly, so that they can solve more complex problems. This partly requires AI to process voice more efficiently and interpret text more accurately, but it also requires front-end processes to be effectively integrated with back-office systems.
ACCORDING TO GABOR STREN, THIS IS KEY, AS BACKEND SYSTEMS NEED TO KEEP PACE WITH DEVELOPMENTS: IT IS CRUCIAL TO WHAT EXTENT AUTOMATION CAN BE APPLIED TO A CUSTOMER ACQUISITION OR ACCOUNT OPENING, FOR EXAMPLE.
Successful automation of backend processes will allow individual processes, such as an identity check for an account opening, to be carried out with much less interaction.
In the wake of practices intensified during the coronavirus outbreak (remote banking and digital channels), many new types of photo-based documents, ID cards and critical data sets have been created, which could also be identified, classified and interpreted using AI tools. This is an area that definitely deserves more attention in the future.
Artificial intelligence can also support everyday work effectively. There are many examples of this: we can already think of collaboration applications, which are becoming more and more prominent with the spread of home office work, and which also have AI extensions that make everyday use much easier (translator, subtitler, transcriber for spoken language). Typical RPA solutions can also provide significant support: they are automation software that are used
WHEN A PROCESS AND ITS OUTCOME ARE RELATIVELY WELL-DEFINED, BUT NOT SO TRIVIAL, AND A LARGE NUMBER OF PROBLEMS NEED TO BE SOLVED.
Security and fraud prevention
A less glaring, but all the more important area of AI developments in banking is fraud prevention. These developments mostly help banks by detecting the less algorithmic attacks. These tools actually filter anomalies, which can be various phishing or social engineering methods. As Gabor Stren puts it, detecting these incidents is no longer a manual task, due to their large number and constantly evolving methods.
The tool works well when “you see very little of it” and when it only notifies users in the most urgent cases. However, even if everything is running smoothly, , the human factor still cannot be ignored. Interventions that require human judgement are specifically assessed by the AI and reported to staff.
Security engineering is most relevant in IT systems where many users are involved and complex data sets are in circulation. This can be the case for corporate mail systems and collaboration platforms, which are exposed to increased security risks as home office becomes more common.
Bank branches of the future
Artificial intelligence could also reshape the role of the branch network: as digitalization advances, branches are trying to find their place in the market, and it is not yet clear what their future role will be. There are already pilots in our country and technologies are already being deployed in other countries based on making the data generated when a customer visits a branch. A lot can be revealed by the customer’s emotions, their location in the customer area, the amount of space they use, etc. Of course, security is also an important consideration, so AI must also filter out suspicious elements.
OF COURSE, IN THE CASES MENTIONED ABOVE, THIS TECHNOLOGY USES HIGHLY SENSITIVE DATA, THE LEGAL ENVIRONMENT IS NOT YET FULLY IN PLACE, AND THE USE OF THE INFORMATION ALWAYS REQUIRES THE CONSENT OF THE CUSTOMER.
A serious problem is that in a significant number of financial institutions, the data needed for AI-based analytics is not available or is not available in an appropriate way. This can be due to outdated systems and the limitations of existing processes and human resource capabilities. Szabolcs Pinter believes that much could be done to improve this by modernizing data storage methodologies and introducing new big data analytics infrastructures.
Time dimension is also very important, in which there is still room for improvement, according to Peter Faykiss. Banks today often do not have a whole year to deliver projects. They have to cope with market conditions that are too turbulent for that (just think of the changed conditions due to the coronavirus epidemic).
THE TIME IT TAKES TO DEVELOP A SOLUTION, OR THE ABILITY TO MAKE PROCESSES FASTER AND MORE STREAMLINED, IS CLEARLY AN IMPORTANT COMPETITIVE FACTOR.
A good example of this could be an online onboarding process, the continuity of which can be a crucial factor in customer acquisition.
Fraud prevention and anti-money laundering software often casts too wide a net, and financial institutions are often unable to undertake the self-learning process needed to ensure that in the future everyday transactions do not get caught up. According to Szabolcs Pinter, this could temporarily lead to such a level of customer complaints and consequent loss of customers that financial institutions are generally reluctant to adopt these solutions. The low effectiveness of anti-money laundering software may also be behind the revelation of hidden oligarchic assets in several Western banks following the sanctions imposed upon Russia. It is reasonable to assume that the AI applications designed to support anti-money laundering protocols were under-utilised or, if they were properly used, they suffered from a number of shortcomings.
The application and development of artificial intelligence can be effectively supported even by cloud technology, and local stakeholders can develop solutions that are tailored to their own objectives and scale: most international banks operating in the country rely on the capabilities of their parent banks or use cloud-based systems.
But the KBC group’s robust, high-performance computer could be an important factor even at the level of the national economy. The project is likely to create intense competition between banks, and Peter Faykiss says that sooner or later most financial institutions will have to start building up the expertise to exploit the potential of AI.
PROCESSING OUR LANGUAGE OPENS UP MANY POSSIBILITIES FOR CREATING AI-BASED APPLICATIONS AND, OF COURSE, THE STORAGE OF SENSITIVE DATA ON AN IN-HOUSE DEVELOPED COMPUTER IS A LESS RISKY SOLUTION.
The supercomputer, built in collaboration with ITM and SambaNova Systems, aims to create a language model that will be able to interpret text in our language, and process and manage huge amounts of data. This will speed up backend and frontend processes and automate more and more repetitive workflows. The tool, with its high-level linguistic analysis functionality, could give a huge boost to the development of AI in the whole country.
According to Peter Faykiss, properly trained human resources are of paramount importance for such projects, and not only for the development team – as it is a great advantage if the bank has experts who can select the right vendor for the project with a good technological sense. Setting up a team is not easy, of course,
BECAUSE THERE IS A HUGE DEMAND FOR THE RIGHT EXPERTISE BOTH IN OUR COUNTRY AND ABROAD.
Artificial intelligence solutions can be demonstrated not only by commercial banks, but also by the National Bank. The central bank has managed to implement a machine learning chatbot into its operations, ahead of many domestic banks, working with a fintech vendor. According to Peter Faykiss, this would not have been possible without a digital capability pool closely linked to human resources, which is being continuously developed within the National Bank. He pointed out that there is a wide range of possible applications, and that solutions using various machine learning neural networks can support a wide range of functions, such as helping economic forecasting, processing various documents and supporting supervisory and customer service work.
Image source: Getty Images