On June 5th, the Luxembourg Analytics Summit, entitled "AI: Separating the hype from reality" will bring together over 150 business users, analysts and data scientists to discuss analytics and Artificial Intelligence, its opportunities and its threats. Prior to the event, we met with experts Marc Aguilar (Chief Data Officer, BGL BNP Paribas), Rein Bryssinck (Sales Director, SAS) and Michel Philippens (Head of Customer Solutions, SAS) to address some of the opportunities and potential revolutions brought to the banking industry, among others, thanks to the advent of Artificial Intelligence.

"Before even talking about Artificial Intelligence, it is essential to highlight that it is first and foremost digital which changes the way we do business," starts Marc Aguilar. The CDO defines digital as scalable and states that it will impact every single industry, from entertainment to banking. "Digital brings us the raw material we so deeply need: data. It therefore is a main component and enabler of the development of Artificial Intelligence. Through the combination of data and algorithms, AI will definitely define the business models of the future," he explains. And thanks to today's ability to easily store, power and move the data, its growth will be exponential.


Providing customized and informed services

On the impact of AI on the banking industry, Marc Aguilar underlines that financial actors will be able to optimize operations and operational costs, but also improve fraud detection. He adds: "Processes are going to be much smoother in the sense that AI will help with the crosschecking of information". According to the BGL BNP Paribas digital expert, the use of AI will allow to strengthen the KYC – Know Your Customer – process as it helps understand how customers interact with the bank. "We can imagine that credit scoring won't be based only on current liquidity and salaries, but AI will tell the bank how the client evolves in time, how he paid back previous debts, etc," highlights Mr. Aguilar, meaning that the way to create products and services for specific clients will be much more accurate, thanks to AI. As a matter of fact, each product and service could potentially be customized, as more and more data and insights will be integrated – coming from different sources. "As you can, through AI, have a much better insight at the needs of the customer, banks could end up with one-to-one project design processes, which are different for every single customer," adds the Chief Data Officer. As an example, loans and rates, could be defined with the use of precise information and calculated individually for every customer. They would eventually depend on the risk the bank takes on this very client. "AI brings faster and better services, which will be much more personalized and better adapted to the specific needs of customers. There will be a tremendous acceleration in innovation for companies and sectors using the technology," resumes Michel Philippens.

Moreover, banks are requested by law to store data and have more than 20+ years of experience processing these data, making them the perfect candidate to fully benefit from Artificial Intelligence. The SAS expert Michel Philippens, explains: "Information-rich companies like banks and insurers depend on their capability to drive operational decisions based on data and algorithms. These decisions could be accepting a credit, filing an insurance claim, etc. If you have algorithms that can take precise decisions, processes will be automated". This automation will then lead to customer satisfaction because of the short amount of time it takes. Therefore, the competitive advantage of companies that have a lot of micro-operational decisions to take will depend on their ability to leverage data, algorithms and Artificial Intelligence. Yet, the way they collect and use data is now regulated by the GPDR – Global Data Protection Regulation – which clearly gives the control back to the user. In this respect, as Marc Aguilar underlines it, "new business models could be created. For instance, a bank could offer to a customer the creation of his profile, which he could use for selected merchants and whenever he wants. The traditional model will completely be transformed, with the customer retaining the control of his data while still trusting his bank." As an example, following the implementation of the PSD2 directive earlier this year, third parties are now authorized to provide instant payment APIs, disrupting the existing banking business models. Therefore, the traditional players could soon envisage to sell risk rather than actual credit cards, creating new business models and resulting in a complete change of paradigm.



Ethics at the center of Artificial Intelligence

Another challenge addressed by Marc Aguilar, Rein Bryssinck and Michel Philippens was re-training algorithms in an ever-changing environment. "It means self-learning, and putting in production the re-learning of algorithms in real-time, or at least detecting that it has be re-trained," adds Mr. Aguilar. "That is what we call the control and governance of AI because, by definition, it is a self-learning model," highlights then Rein Bryssinck, who also raises the following question: how do you control over time, whether it is still making decisions in the way you originally planned it to work? Also, who makes the eventual decision? As Michel Philippens underlines it, it is still prohibited by law in many countries to make automated decisions based on algorithms which can have adverse effects. He adds: "in many circumstances, algorithms will just support Humans in making a decision. As a matter of fact, if a transaction which may be fraudulent is flagged, it will still end up at the desk of an expert who will have to verify whether it is actually fraud or not". Hence a big necessity to deal with change management as many collaborators are involved in these new real-time and IT projects, and need to know how to cope with these new technologies.

"Making it happen successfully requires a collaboration between several departments of the company: IT, obviously, but also data scientists who build algorithms, business managers, legal experts and the operational staff," explains Mr. Philippens, who continued: "communication is key, and discussions about the role of the algorithms in the decision making process must clearly be defined. Yet, it represents a big challenge for most organizations, which are most of the time still split in siloes". According to Marc Aguilar, different skills are needed to approach these challenges of ethics and control of the algorithms.


Separating hype from reality

Threats are real as technology can always be used in a negative way. First, anonymity doesn't exist anymore, and according to Marc Aguilar, several countries have been tracking their citizens to ensure political sustainability but have also built a credit model based on how they behave in society. "It brings out a social issue as it is now possible to track anybody, thanks to Big Data, and these pieces of information could very well be exploited in the wrong way".

On the idea that robots might take over Humans, the SAS Experts remind us that they are simply trained to perform tasks and still miss creativity and innovation. Yet, according to Marc Aguilar, the combination of intelligence, storage, processing power and the speed of network could definitely mean that Humans might lose control over machines. For instance, in case of a house fire people are able to quickly leave the room while the computer – and therefore the Artificial Intelligence – can't. "In the future, computers will be connected and communicate, meaning that AI could transfer, hence the need to be cautious when building systems we might lose control of in the years to come. We are not there yet, reassures the CDO, but there will be a point in time where it could become uncontrollable". Today, algorithms and robots already outperform Humans on certain tasks, notably in the field of healthcare and especially when it comes to spot fractures, tumors, etc. "It means a tremendous progress and opens the way for a personalized and customized medicine, which might, in the end, cure diseases that are incurable today". Marc Aguilar adds: "In the future AI will decode DNA and provide a greater number of information, spot the probability of getting cancer, and so on…" The experts also discussed the fact that AI could help reduce or even eliminate the number of car accidents, thanks to autonomous driving. "AI is an extreme accelerator of innovation," concluded the three experts.


Participants of the Luxembourg Analytics Summit will learn more about Artificial Intelligence and its promises in several industries, from banking to healthcare, while also foreseeing a brighter and innovative future through actual use cases and trends. More information: www.analyticssummit.lu


Interview by Alexandre Keilmann

Publié le 25 mai 2018