How to make use of the potential of artificial intelligence in banking? Part 2

AI & Machine Learning

It's no surprise that good preparation for large projects is the key to success. The same principle applies in the case of preparing banks for implementation of systems based on artificial intelligence (AI). Large financial institutions require remarkable precision in implementing changes, and each "half-baked" performance at the stage of planning may result in several million in losses for the bank. Therefore, when investing in the improvement of financial systems and using Artificial Intelligence in banking, it's worth remembering a few basic things:

Analysis and standardization of data for implementing Artificial Intelligence in Banking

If there is a sector where availability of large amounts of data is not a problem, it's certainly banking. In fact, companies such as Oracle and Accenture created entire departments helping with storage and management of bank data. For many years, financial institutions have been gathering huge amounts of information on our preferences, consumer behavior, credit history or family situation.

Therefore, the challenge lays not in acquiring an adequately large database, but in enabling the use of these data in building machine-learning systems while preserving the restrictive regulatory limitations related to the safety of client data protection. There are plenty of these - apart from standards concerning only bank security measures, financial institutions must observe a number of regulations concerning the use of personal data, particularly sensitive data.

Another challenge in this context - perhaps an even greater one - is proper analysis of available data that are not only fragmentary but also located in various, often incompatible systems. This happens mainly due to frequent mergers and acquisitions in the banking sector, as well as relative undercapitalization of back-office departments. To solve this problem, banks should start building AI systems with a small set of complex data and add subsequent ones, thus creating a universal record of each client.

Bearing the above issues in mind, before the actual implementation of Artificial Intelligence in banking sector, it should invest and focus its efforts on organization of data available in the company.

Substantive support in the processes of implementation of AI systems

Implementation of artificial intelligence algorithms in organizations such as banks is extremely difficult and requires expert knowledge. Substantive knowledge with regard to data analysis is necessary already at the stage of collecting information. Certainly, when making the decision about implementation of AI systems in a bank, it is necessary to obtain the help of specialists involved in the creation of algorithms based on machine learning. This would both help detect potential hazards in the implementation at the early stage of the works and will enable efficient identification - and then execution - of goals and priorities of the organization.

Test and learn

If data analysis starts to bring results, there comes the time to inspect it. It is necessary to conduct A/B tests (comparative tests) in order to check the accuracy of results and obtain high level of trust towards the system.

For example, if data are used to extrapolate trends and predict behaviors, the data analyst should verify each variable in regression analysis and test the expected results against a known set of data. Omission of this step may result in creation of a system that would generate incorrect forecasts, which could, for obvious reasons, expose the bank to serious financial losses.

Product for internal use

Determination of the initial use-case for systems based on artificial intelligence in banking may sometimes be very difficult. An extraordinary popularity is currently enjoyed, for example, by virtual assistants (bots), which not only help effectively serve thousands of clients a day, but also limit the banks' costs and allow them to open to new markets relatively inexpensively. Unfortunately, even in the case of products that may at first glance only yield profits to the company, there is a risk of sustaining losses worth several million. Such was the case of the Tay bot - a system prepared by Microsoft tasked with responding to twitter posts and client inquiries. Despite the fact that the product substantially performed its entrusted tasks, sometimes the bot became offensive and its answers - racist. Dissatisfaction of users caused the system to be temporarily withdrawn from commercial use. Taking the above into consideration, it's better to start using AI systems with internal projects, which can also yield immediate income, but at a much smaller risk of endangering the reputation of a financial institution, for which the trust of its clients is one of the key factors of success.

Implementing Artificial Intelligence in banking sector. Success stories of the 3 largest American banks

JPMorgan Chase

The first spectacular example of use of AI systems by JPMorgan Chase is the Contract Intelligence system (COiN), created specifically for the needs of the bank. The program was designed to "analyze legal documents and extract important data points and clauses". Certainly, in the long run, the system will bring millions in savings to the bank. Manual review of 12,000 credit agreements usually requires approximately 360,000 manhours. The results of initial implementation of this technology showed that the same number of agreements can be verified in several seconds only. COiN certainly has a broad potential and the company examines the ways of implementation of this powerful tool in other fields of operations.

Another example of implementation of artificial intelligence at JPMorgan Chase is the so-called Emerging Opportunities Engine, created in 2015. It's tasked with analysis and identification of clients best prepared for obtaining capital by issuance of shares. This technology turned out to be effective on capital markets and is currently expanded onto other areas, including debt capital markets.

In total, the bank invested in 2017 more than USD 9.5 billion in various kinds of technologies, 3 billion of which were allocated to new initiatives, such as AI.

Wells Fargo

In its pursuit of utilizing the possibilities of artificial intelligence and contributing to strengthening of its organizational structure, Wells Fargo announced in February establishment of a new team called Artificial Intelligence Enterprise Solutions. Steve Ellis, EVP of the bank, was appointed to manage the new team. The effects of the activities of the newly created body appeared very quickly. In April, the company began a pilot program with the use of a chatbot based on machine learning algorithms. This virtual assistant establishes relations with users through Facebook Messenger and is already able to provide accurate information on the client's account, as well as perform simple actions, such as resetting a password. After being tested by 700 company employees (a good example of initial internal testing!), the system will be introduced for all clients of the bank, many of whom have been performing financial operations through Facebook Messenger already since 2009.

Bank of America

Bank of America was one of the first financial institutions which provided mobile banking to its clients 10 years ago. Last year, the bank introduced Erica to the market, a virtual assistant that was regarded in 2017 as the most significant innovation in payment and financial services. As a result of using predictive analytics, Erica works as a financial advisor for over 45 million clients of Bank of America. Integration of the intelligent assistant with advanced mobile banking solutions is intended to reduce burdens related to routine transactions, so as to allow client service centers to solve more complicated cases faster.

2017 was the second most lucrative period for Bank of America, which informed that it spent USD 3 billion on technological progress in this year only. The company is on a good way to achieving subsequent records and constantly increasing its presence in the finance industry.

These 3 presented examples obviously do not exhaust the catalogue of uses of AI systems by the banking sector. Let's not limit ourselves only to the United States - a good example of gaining benefits thanks to virtual agents based on machine learning systems is China Merchants Bank. This Chinese commercial bank uses a bot in the popular application WeChat to handle 1.5 to 2 million inquiries a day. In order to handle such a quantity of work without using AI systems, the same bank would need to employ more than 7,000 people. Another spectacular example is the use of artificial intelligence mechanisms by one of the Australian banks. Currently, it is experimenting with an independent, intelligent virtual assistant, the primary task of which is to listen in on conversations of bank employees about loans. If a bank employee forgets about something or makes a mistake, the bot automatically engages in the conversation.

With a good basis for implementation of systems based on artificial intelligence, as well as a set of relevant tools and substantive support in creating algorithms, every financial institution is able to achieve real benefits from the use of AI. It's only a matter of time until systems of this type become a determinant of the market position of banks and a key element of competitive advantage.

Want to know how we can help you in implementing Artificial Intelligence in banking sector? Contact Us.