With the emergence of locally deployable LLMs, enterprise AI strategies are at a turning point. This session explores practical approaches to in-house AI development that leverage legacy systems without APIs and tacit organizational knowledge while keeping proprietary data on-premises. Rather than replicating Western models, we present a phased implementation that respects Japanese corporate culture, avoiding departmental silos to achieve enterprise-wide optimization. We'll share real-world examples of practical solutions for breaking free from cloud dependency and securing data sovereignty.