AI in Banking: From Proof of Concept to Industrialisation | IBM Consulting

When does AI stop being an experiment inside a bank? David Sebaoun, Partner at IBM Consulting in charge of data, AI and cloud, on what changes when artificial intelligence moves from proof of concept to industrialised banking processes. Recorded at Big Data & AI Paris 2026. What happens to a core banking process once AI runs part of it — and what has to be in place before anyone trusts the result. The conversation covers KYC and client onboarding, agentic AI and the division of work between agents and people, the shift in the human role from execution to decision, and what AI governance actually has to prove: transparency on training data, explainability of a credit recommendation, and measurement of model health under the EU AI Act and its Article 50. Thirteen minutes with someone who does this daily for large banking groups. Interview by David Perron, Senior Advisor AI, Aequitas Consultus. English subtitles available. Chapitres 00:00 Introduction 00:54 David Sebaoun, Partner at IBM Consulting 02:12 From isolated use cases to entire processes 02:57 What industrialisation changes for people and skills 04:45 Human on the loop, human over the loop 06:27 KYC: why the process is complex 07:28 Cognitive deconstruction into atomic steps 08:03 The agents: documents, sanctions lists, scoring 09:53 Adoption and change management 10:20 The AI Act, Article 50 and the register 11:05 AI governance: operational and reputational control 11:58 Transparency, explainability and model health 12:46 Closing