# The Deployment Gap ## Access to AI is easy. Deployment is hard. Any company can buy access to a strong AI model. Few companies can put that model into daily work. The scarce item is not the model. The scarce item is the link between the model and the company systems, data, permissions and people. > [!tip] The core claim > The constraint moved from access to deployment. A vendor that owns the deployment step can earn money from every model, because the model is a supply input. Four blockers cause the gap: - Company knowledge sits in documents and inboxes. - Work crosses systems that do not talk to each other. - Security teams need control before AI touches real work. - Executives need a measured result, not a demo. ## Why it matters If the claim is true, value moves from the model layer to the layer between model and company. If the claim is false, model vendors absorb this layer. The diligence task is to test which case is real. See [[Incumbent Bundling Risk]]. ## Related - [[AI Deployment Layer MOC]] - [[Pilot-to-Production Failure Rates]] - [[AI Agents Stack]] - [[AI Eats Services Not Software]]