Context: For decades, integrating new capabilities into agile teams, whether UX practices or emerging technologies, has followed a consistent paradigm: teams remain stable while methods adapt. Applied to AI, this approach produces fragmented tooling, governance uncertainty, and vibe coding: unstructured AI use that prevents cumulative organizational capability. We argue this integration paradigm, successful for disciplines like UX, may be inadequate for AI’s transformative scope.

Objective: We reframe AI integration as an organizational design problem: how should teams restructure around AI-Dev-Systems, configured environments of AI agents, workflows, and knowledge bases functioning as shared infrastructure?

Method: We propose a three-layer model extending platform engineering: (1) System Teams design AI-Dev-Systems; (2) Platform Teams maintain them; (3) Product Teams develop software within them.

Results: Developers shift from producers to stewards monitoring AI outputs. Iteration cycles compress, paradoxically elevating product thinking: when implementation ceases to be the bottleneck, what to build matters more than how. XP practices persist but transform into governance mechanisms for AI-generated work.

Conclusion: Teams should restructure around AI-Dev-Systems rather than adopt AI tools, our three-layer model offers a structured provocation to reframe the discourse.