Speaking with Delano as part of the series, Olivier Ramlot, the chief information officer at Banque Internationale à Luxembourg, elaborated on the behind-the-scenes efforts at Bil to actively integrate AI, especially generative AI, into enhancing banking services and compliance processes. This focus is aimed at improving customer experience, internal efficiency and regulatory adherence, while also navigating challenges such as talent acquisition and evolving regulations.

Kangkan Halder: In your view, how will the integration of AI, particularly generative AI, affect regulatory compliance and the evolution of banking services in Luxembourg?

Olivier Ramlot: It might not have been widely noticed, but AI has already started to change the way banking services work in [recent] years. Some services are already fully empowered by AI models allowing them to provide faster and better answers to the customers, or by simply improving internal processes. To illustrate, at Bil, card limit extensions are treated automatically by an AI. Nevertheless, we are only at the premises and a lot is arriving.

With a focus on regulatory compliance, a wide topic in constant evolution, GenAI supports officers to always be on top of their expertise while continuing to increase their efficiency. For instance, it can allow to deep-dive cross-functional topics, help to detect new trends and emerging risks, or simply support the teams by summarising complex subjects.

But even without GenAI, regulatory compliance is already evolving thanks to other ways of machine learning. Most major European banks are already using AI models to detect new fraud patterns with real-time monitoring. In addition, graph mining technology, supported by PSD II [ the EUpayment services directive], will improve money laundering detection by deepening the understanding of customer interactions.

It is assured AI will soon be a standard within the Luxembourgish banking market.

Could you share insights into the strategic plans of Bil for AI deployment in the coming years, and how you expect this will reflect broader trends in Luxembourg’s banking sector?

Our objective is to elevate AI as a cornerstone technology within our organisation. Our CIO, CTO and AI lead are spearheading this movement, beginning with the integration of Generative AI. Initially, we are focusing on internal enhancements, such as augmenting coding processes and AI-assisted document reviews.

Building on the success of our internal human resource chatbot, Laura, which already assists Bil’s employees in self-service tasks and query resolutions, we are poised to take our human-AI interaction to the next level. Moving beyond IBM Watson’s capabilities, we are exploring foundational large language models. This shift aims to harness our collective intelligence by tapping into the extensive knowledge embedded in our policies, procedures and wikis, thereby providing AI copilots to all employees.

Our ultimate ambition extends to enhancing customer experiences. We are concentrating on our mobile application and web banking platforms, with plans to introduce features that mirror the conversational ease found in tools like ChatGPT. This innovation is not just about improving customer interactions; it’s also about providing more sophisticated, AI-tuned tools to help customers better understand and manage their financial assets. By leveraging finely-tuned AI models, we intend to empower our customers with greater insights and capabilities in handling their financial portfolios.

We expect AI for compliance to be an even more scrutinised matter by regulators in the coming days.

Olivier Ramlot, chief information officer, Banque Internationale à Luxembourg

How is Bil utilising AI to streamline and enhance compliance processes, and what implications does this have for the banking industry in Luxembourg?

At Bil, before using AI for compliance processes, we decided to first settle strong foundations before investing in this topic.

Firstly, with our core banking system evolution, we focused on the migration and the stability of our current approaches. Secondly, we wanted to strengthen our XAI [explainable AI] and monitoring techniques to increase our compliance with the upcoming regulations. We expect AI for compliance to be an even more scrutinised matter by regulators in the coming days, such as the EU AI act, Luxembourg Financial Sector Supervisory Commission white paper on AI, amongst others.

Lastly, we desired to train compliance teams on the usage of AI. Going from the methodology to the legal and ethical considerations and responsibilities related to AI development. It is key that users have a good understanding of how AI works to avoid misconceptions on the topic, and how it will help them in their day-to-day activities.

However, the foundations are settled, and starting in 2024, we will focus on three major AI capabilities for the compliance sector that aim to support their day-to-day activities.

The first capability is to have an automatic and in-depth customer transaction classification. It will allow us to increase the efficiency of our know your transactions activities, as well as being a foundational brick for other AI topics.

The second one will be an AI model that will improve and refine fraud detection. The goal is to drastically reduce the false positives (wrong alerts) that high-jack the attention of our inspection team, but also detect new fraud patterns that can come, leading to always having a low false negatives rate (missed alerts).

Last but not least, an automatic customer profiling, based on their transaction history, we will support the ‘know your customer’ team, to rightfully pinpoint the few medium and low profile that require remediation.

All these capabilities are here to support a sector that regularly has new regulatory requirements leading to a need to constantly improve and refine our processes. 

What specific opportunities and challenges do Bil, and other Luxembourg banks, face in adopting and implementing AI or generative AI technologies?

The primary challenge lies in elevating awareness and upskilling our staff in the realm of GenAI. This endeavour requires robust technological leadership coupled with the empowerment of experts who can coach and guide other team members. Our leadership is committed to progressing further in this domain. By grasping the potential of AI, we are poised to make smart moves and excel in this era of digital augmentation.

The second hurdle is specific to the Luxembourg market, which is currently experiencing difficulties in expanding its talent pool amidst increasing technological diversity. The growing need to recruit data scientists, AI engineers and AI educated legal/risk management experts could potentially create an imbalance in the market, particularly between larger and smaller players.

Our third challenge involves striking a balance between AI initiatives and the existing regulatory framework, as well as managing our risk appetite. The EU AI Act, still in the process of finalisation, already indicates the need for financial institutions to work closely with regulatory bodies such as the [Luxembourg financial regulator] CSSF and the national commission for data protection, CNPD. It is crucial to ensure that our decisions align with upcoming regulations and do not become counterproductive in the foreseeable future.