Simply offering the latest model is no longer a competitive advantage. True value is created in the system built around the model—the system prompts, tools, and overall scaffolding. This 'harness' is what optimizes a model's performance for specific tasks and delivers a superior user experience.

Performance gains increasingly come from the "harness"—the surrounding system of tools, data connections, and agentic workflows—not the underlying model. Stanford's "meta-harness" concept shows a 6x performance gap on the same model, suggesting the product layer is where real innovation and competitive advantage now lie.

User stickiness for AI models is increasingly driven by the 'harness'—the custom prompts, workflows, and integrations built around a specific model. This ecosystem creates high switching costs, even when a competing model offers incrementally better performance.

Unlike traditional APIs, LLMs are hard to abstract away. Users develop a preference for a specific model's 'personality' and performance (e.g., GPT-4 vs. 3.5), making it difficult for applications to swap out the underlying model without user notice and pushback.

Switching AI vendors is difficult not because of data lock-in, but because of user expertise. The cost for a power user to learn a new tool is too high unless a competitor is at least twice as good, creating an "inertia grab" moat.

A model's native power does not automatically translate to effective organizational capability. Migration costs and ecosystem fit are paramount. Factors like existing developer tools (OpenAI vs. Claude APIs), identity controls, and data residency should weigh as heavily as raw performance benchmarks when selecting a flagship model.

Debate around Anthropic's Claude Tagg reveals a broader truth: as AI systems become deeply embedded with organizational context and permissions, high switching costs are an unavoidable consequence. This lock-in is not a product flaw but a signal of successful, high-value integration.

Data from fintech Mercury shows a startup's initial choice of AI platform (e.g., OpenAI vs. Anthropic) is a critical decision. This choice often dictates subsequent tool adoption and creates significant lock-in as workflows and knowledge bases are built around that initial platform.

Despite constant new model releases, enterprises don't frequently switch LLMs. Prompts and workflows become highly optimized for a specific model's behavior, creating significant switching costs. Performance gains of a new model must be substantial to justify this re-engineering effort.