As insurers experiment with AI across underwriting and claims, a structural question emerges: can legacy core systems truly support AI-driven workflows, or does real transformation require AI-native architecture? This session explores the practical differences between retrofitting intelligence into existing stacks and building around modern, API-first, AI-ready systems. What implementation trade-offs matter? Where do integrations break down? And how should insurers evaluate long-term scalability? This discussion focuses on infrastructure decisions that shape the next decade of execution.