Wawrzyniec Hyska

Director for Financial Sector

Will Artificial Intelligence Revolutionise Banking? AI: Its Potential and its Challenges

We have been hearing about the revolution to be brought about by artificial intelligence for 60 years. Today we have access to a greater amount of data than ever, with 80% of all data generated within the last 2 years! We have access to more powerful graphic processors that allow us to process images 60 times faster than traditional processors.

Machine learning is developing by leaps and bounds. It seems that we are on the verge of a breakthrough, but one question yet to be answered is: When will AI begin to generate real business benefits?Artificial Intelligence is an integral part of the strategic plans for many banks for the upcoming years. Debates at the 2017 Meeting of Leaders of Banking and Insurance was centered around technological changes to the face of banking. This article summarizes the presented information.

Definition of AI

Artificial intelligence (AI) as broadly understood is the ability of machines to learn and make decisions. It is necessary to clearly distinguish narrow AI, designed to perform a single specific task, from general AI, the capability of adapting to new circumstances and performing any intellectual task that a human would be able to handle.

In March 2016, AlphaGo, software developed by Google DeepMind, beat one of the most experienced Go players in the world, Lee Sedol. As indicated during the discussion by Dariusz Piotrowski (Managing Director, Dell EMC), this is a sign of a new stage in AI development, as the Google algorithm exhibited intuition during the match, making moves that a human player would not have made, even a neophyte. Using a strategy contrary to adopted practices and tradition, the program kept disrupting the experienced player’s rhythm.

Business Application

In business, AI is used in various fields as voice assistants such as Siri or Alexa, which respond to voice commands, robotics, e.g. Kiva by Amazon - mobile robotic fulfillment systems (more information about this topic can be found in the box below), or personalisation.

In banking, AI is currently used primarily in robotics and process automation. Banks are also eager to test chatbots aimed to improve customer service. Banks find various applications for data science mechanisms (behavioural analytics) and machine learning mechanisms in the field of fraud detection. These technologies allow prompt and automatic detection of behavioural anomalies, client login from a new country, browser or device, for example, or actions which are inconsistent with his usual behavior, allowing immediate reaction to emerging threats.

AI in Numbers – Statistics Regarding Artificial Intelligence

  • - 2016 investments in AI in Poland amounted to approximately EUR 11 million. Investments of other European countries were commensurate
  • - Interest in AI keeps growing. In 2016, the number of press articles devoted to artificial intelligence was twice higher than in 2015 and nearly four times higher than in 2014
  • - The world currently generates 2.2 exabytes, i.e. 2.2 billion gigabytes, of data per day
  • - Application of Kiva robots by Amazon to navigate around the warehouse and complete orders reduced the “order placement to dispatch” cycle from 60-75 minutes to 15 minutes and improved warehouse capacity by 50 percent without the need to invest in additional space. The annual return on investment amounted to nearly 40 percent

Data based on: The AI revolution. How artificial intelligence will change business in Poland, McKinsey & Company Report, Forbes Poland, 2017.

AI is Fuelled by Data

Large data sets constitute the foundation for the implementation of artificial intelligence. Jim Marous illustrates this problem in the form of the Maslow’s pyramid: with data collection featured as its foundation, the basis for all other activity. The next layers of the pyramid are data processing, data governance, segmentation, and then, at its apex, artificial intelligence. It is the large quantities of data that provide the algorithm with the material necessary for learning.

At this stage, a key issue for banking is data acquisition and structuring.

Challenges

Data acquisition, collection and structuring are not the only challenges faced by banks that want to introduce AI solutions. Kamil Gustaw (Head of the Banking Product Research and Development Centre, Citi Handlowy) believes that it will be difficult to explain to the regulator how artificial intelligence works and what will be its exact scope of operation – since by design, the machine will learn independently based upon the data to which it is exposed, identifying or projecting new patterns of behaviour For this reason even the creator of the AI product cannot specify exactly how AI decisions will be made at the outset.

There is also the question as to whether customers are ready to contact a chatbot, for example, with such a sensitive matters as their money, and whether such technology will have a deterrent effect on customers.

Customer in the Age of AI

Marcin Kotarba, Head of the Department of Analyses and Strategy, Deutsche Bank Polska, claims that customer response to AI depends largely on the segment – premium customers will most likely be attached to traditional customer service and expect interpersonal contact. For millennials, on the other hand, it is the result that counts: they want to quickly obtain the necessary information, and whether this information comes from a human or bot is irrelevant to them. What’s more, innovative technological solutions may be welcomed by them, adding an attractive novelty value.

In the end, this dilemma comes down to the issue of Customer Experience. Paradoxically, we use technology to make contact between the client and bank feel more human. We provide personalised offers, ensure coherent and relevant communication via every channel, streamline processes... Technology is a means to achieve goals, i.e. customer satisfaction, rather than a goal unto itself. AI solutions implemented as art for art’s sake do not make much sense.

Recommendations

The basis for successful AI implementations is the development of a business justification, which starts with a realistic assessment identifying areas in which AI may be beneficial. The second important condition is implementation in an area of strategic importance and a strong conviction about its significance – both criteria met in the implementation of Kiva robots by Amazon in the case described above.

In today’s world banks are becoming technological companies. As such, they must recognize that innovative technological solutions are developed when there is a real, cross-field dialogue. This is the case with tribes at ING Bank Śląski, which boldly adapts the agile methodology. In such organisations, IT specialists are not disconnected from business, but rather are in constant dialogue with it and propose new solutions on their own. That is why the move away from silo-based divisions is a prerequisite for the successful adoption of modern solutions, including artificial intelligence.