The decision between buying a tool and building one internally has been complicated by advances in artificial intelligence, which have lowered build costs while making purchasing easier. Historically, product managers decided, but AI has shifted the trade-offs to factors like maintenance ownership, system integration, and user experience. A significant majority of enterprise AI use cases are now purchased, yet AI-powered coding agents make building internal tools appear deceptively simple. This "vibe coding" can lead to unmaintainable systems that become products needing budgets and dedicated owners, a cost often overlooked in initial build cases.Purchasing a tool means inheriting a vendor's roadmap, data model, and release schedule, and potentially their organizational structure as well. These purchased systems may not align with a company's specific users or workflows, leading to inefficiencies. Users experience tools not in isolation but through the gaps between them, often involving constant application switching and retyping data, which consumes significant work time. Complexity is a constant, and the choice is merely where it resides: in the company's codebase, a vendor's roadmap, or burdened onto users.Many products are now accessed through host surfaces like Teams or ChatGPT, meaning the design must conform to the host's constraints. Replacing existing systems also incurs costs, including the loss of user expertise, hidden rules embedded in old software, and the "not invented here" bias that favors rebuilding over adopting external solutions. AI coding agents can distort estimates of development time and create a false sense of speed.The core principle is to buy commodities and build differentiators. AI models themselves are becoming commodities, but the surrounding user experience, which is harder to replicate, remains a differentiator. Companies should build internally when proprietary data is a key advantage or when regulations necessitate custom solutions. However, most organizations lack the specialized skills required for production AI applications, making buying the more practical choice for many needs.