Something strange is happening in enterprise AI. The newest, most capable models are getting beaten — in practical business outcomes — by systems built on decade-old infrastructure. SAP's autonomous enterprise initiative generated $2.7 billion in customer value in a single quarter. Not from the newest foundation model. From context. Specifically: 7.3 million data fields of proprietary business context that no startup can replicate. This isn't a SAP commercial. It's a map for where the actual leverage is. The Capability Gap Is Closing The gap between the best foundation model and the second-best has never been smaller. GPT-5, Claude Opus, Gemini Ultra — they're all within a rounding error of each other on capability benchmarks. For commodity tasks — summarization, code generation, basic analysis — capability is essentially solved. Any of them works. The differentiation has moved somewhere else. That somewhere else is context. Specifically: context that competitors can't easily acquire.…