GraphQL offers a way to address the issue of overfetching by allowing clients to request exactly the data they need. However, after years of practical use in enterprise settings, the author, John James, argues that the problem GraphQL aims to solve is often already handled by existing Backend for Frontend (BFF) architectures. In many cases, BFFs aggregate data from multiple REST services and tailor responses to meet UI needs, effectively eliminating overfetching without needing to shift to GraphQL. When GraphQL is layered on top of REST APIs, it merely shifts the overfetching issue rather than resolving it. Implementation of GraphQL is more complex than REST. Developers must define schemas, types, resolvers, and data sources, which adds significant overhead to the development process. In a fast-paced enterprise environment, where speed is critical, these extra steps can hinder productivity. Observability also suffers; GraphQL’s status codes can lead to confusion, as a 200 response might indicate partial failures. The caching system, while technically impressive, becomes cumbersome, requiring manual configuration to work effectively. Moreover, the necessity for unique IDs for each object complicates integration with many existing APIs that don't return IDs, increasing the burden on the BFF to manage extra logic. Other practical challenges include difficulties with binary file handling and a steeper learning curve for developers used to REST. Error management in GraphQL is more convoluted, introducing layers of complexity that can lead to confusion and slow down troubleshooting. Overall, the author concludes that while GraphQL has its merits, it often introduces more problems than it solves in typical enterprise environments, making it a niche solution rather than a universal one.