S2 E4. Build vs Buy in the GenAI Era: Inside Belfius’ Data & AI Strategy - The Data Playbook Podcast with Kris Peeters & Hannes Heylen

Belfius Insurance’s Head of Data & AI, Hannes Heylen shares how his team scaled GenAI - from a fraud detection flywheel to “Nestor,” a claims copilot that speeds summaries, completeness and coverage checks. We unpack AI agents in the claims flow, build-vs-buy decisions, and why content/data governance drives LLM quality. Plus: a pragmatic delivery mantra - make it work, then right, then cheap - for CIOs, CDOs and Heads of Data.What you’ll learnHow to pick first AI cases that prove €ROI (fraud models)Designing a claims copilot: summarization, completeness & coverage checksWhere AI agents fit (GenAI + ML + humans) across the claims flowBuild vs. buy in 2025: foundation models, vendor flexibility, cost controlContent/data governance as the make-or-break for LLM apps“First make it work, then right, then cheap”: an AI operating model for CIO/CDOGuest: Hannes Heylen, Head of Data & AI, Belfius Insurance🌐 More at ⁠www.dataminded.com⁠Chapters:00:00 Why AI now in financial services06:30 GenAI’s impact on text-heavy insurance processes18:40 AI agents across claims31:00 Governance > model tweaks38:00 Fraud detection: the € case41:30 Claims copilot (“Nestor”) & lab-to-prod55:00 Lessons for CIOs/CDOsTopics: ROI-first use cases • Claims automation • AI agents (GenAI + ML + human-in-the-loop) • Governance • Vendor flexibility & costs