Mobile banking is widely accepted because it is not bound to space or time, like a traditional brick-and-mortar branch, and allows customers to bank from anywhere at any time at no additional cost. This gives the customer freedom and immediate gratification, which, in the digital economy, has become the baseline customer expectation.
From the bank’s point of view, on the one hand, online banking reduces traditional operational costs of a brick-and-mortar branch but, on the other hand, it also reduces opportunities for interaction between customers and bank employees. This implies that building strong relationships with the customer and customer satisfaction are no longer the sole responsibility of bank employees, but also of digital front ends. In addition, we know that providing personalised services positively influences customer satisfaction, which further affects brand loyalty. Therefore, we need to ensure that the level of customer satisfaction in the digital world is at par with a BGL BNP Paribas branch and that customers benefit from the same quality of service no matter which channel – digital or brick-and-mortar – they choose for their banking operations. Online banking systems also bring in new types of challenges such as fraudulent transactions, protection of digital information, etc., that need to be addressed appropriately.
Data is the new oil which powers the artificial intelligence engine.
So how can we address the challenges presented by digitisation and offer customers a great personalised experience even in the digital world? With artificial intelligence (AI).
At BGL BNP Paribas, we realised that data is the new oil which powers the artificial intelligence engine. AI can help us build the next generation of intelligent, safe and robust digital services. To this end, the BGL BNP Paribas Datalab was set up three years ago under the leadership of chief data officer Marc Aguilar and chief transformation officer Fabrice Cucchi. In addition, since the technology behind AI algorithms is not trivial and has a steep learning curve, to cut the time to production, we collaborated with research teams to bring in expert knowledge. Together, they are building AI systems for services like product recommendation or automated credit approval in less than 30 minutes, and have already deployed AI-based solutions for fraud detection and Know Your Transaction using modern deployment technologies like Gitlab and containers.
The quality of AI solutions completely depends on how well the problem is defined, on the quality of data available and on how this data is presented to the algorithm (feature engineering). Thankfully, our digital services create massive amounts of data that feed the data-hungry algorithms, and data engineers ensure the data quality. In addition, to define the problems, we promote a culture of frequent atomic communication between data engineers, data scientists and business experts. The term “atomic” means small yet full of energy and information.
Another important aspect of AI is that it is computationally intensive. Therefore, we need to make sure that we have sufficient computing resources in training, testing and production environments. To address this matter, we are adopting modern self-scaling deployment infrastructure like containers with a container-orchestration system like Kubernetes.
It is essential to take data science out of silos to develop its enormous potential.
It’s not yet clear to what extent organisations are aware of the impact that new forms of data exploitation through AI could have on their activities, or whether they give them the priority they deserve. Many are still trying to understand the issues of AI.
The banking sector is still much more concerned with the constraints imposed by regulatory provisions than by the benefits that these new technologies could bring. It is essential to take data science out of silos to develop its enormous potential.
Many talk about possible applications of AI, some make pilots or prototypes. But that is not where the complexity lies. The challenge is to cover the entire value chain and to combine systems that have been in place for up to 40 years with new generation technologies. It is important to be aware that, for this type of project, most of the effort is focused on legacy systems, not on new AI technologies.