BNP PARIBAS & ARTEFACT at AI for Finance by Artefact – Transforming commercial banking with AI

Pierre Ruhlmann highlighted how the bank structures its AI initiatives around value. The program is built on three pillars: tribes that deliver customer journeys, an AI Factory for scaling solutions, and a center of expertise for assessing business value and promoting AI literacy. This organizational framework has been implemented over the past four years, marking the end of the first phase.

Pierre emphasized that BNP Paribas measures value in multiple dimensions, not just financial returns. While cost savings and efficiency are crucial, the bank also prioritizes customer satisfaction and employee engagement. The bank uses Net Promoter Scores (NPS) to track how AI features impact customer experiences. Employee engagement is equally important, ensuring that staff feel the benefits of AI, which aligns with the bank’s dual focus on customers and employees.

Pierre shared examples of how the bank scales its AI solutions, particularly in machine learning for scoring and intelligent document processing. Though BNP Paribas has made significant investments in these areas, they are less mature in generative AI projects. Two MVPs, “Gary” and “Genius Bar,” are under development, focusing on internal knowledge management and customer service improvements. Both projects require further optimization before they can be fully industrialized.

The Gary system is designed to help employees with product checks and procedures. While 90% of the workforce has adopted the system, implementation challenges have arisen, such as inconsistencies in the knowledge base across various procedures. Addressing these inconsistencies is critical before the solution can be scaled further. Gary represents a crucial step toward improving internal processes with AI.

Genius Bar, the second key MVP, aims to enhance customer service by replacing outdated chatbots with a generative AI solution. With over a million anticipated interactions in the next two years, industrializing this solution requires a cautious approach. Red-teaming exercises have been critical in identifying security and reliability issues, which have been patched to ensure smooth scaling of the system.

Pierre emphasized AI adoption, using Microsoft Copilot as an example. A trial with 20 employees showed that despite initial interest, half stopped using the tool. This led to a Champion Program and targeted training. The bank is now scaling Copilot cautiously to ensure a good return on investment.

Adoption challenges also include fostering a critical mindset among employees to ensure AI outputs are interpreted correctly. Given the regulated nature of banking, employees need to be cautious about relying solely on AI-generated solutions. Pierre also addressed the fear that AI might take jobs away, reassuring employees that AI will augment, not replace, their work. He highlighted the use of AB testing to compare the value of tasks done with and without AI, helping to promote trust and understanding of AI’s potential within the organization.