Miller closes with the platform question: how do you build AI systems that scale? The answer is mostly not about the AI itself — it is about the infrastructure around it. Can your data pipeline handle 100x more data without redesign? Can your model serving infrastructure handle 100x more requests? Can your monitoring systems still give you visibility at 100x scale? Can your human review process scale with the volume, or does it break above a certain threshold? The builders who think through these questions during initial design make different architectural choices than those who optimise only for the immediate use case. Scalable AI platforms are not more expensive to build initially — but they require thinking about scale during the design phase, not after the system is already straining under growth.