For developers transitioning from relational databases to document databases like MongoDB, the use of "joins" is viewed as an anti-pattern due to potential performance issues and architectural fragility. Instead of relying on joins, MongoDB advocates for denormalization, where data that is accessed together is stored together in the same document to reduce query latency and simplify logic. This approach is part of a larger architectural pattern known as Command Query Responsibility Segregation (CQRS), which separates the command (write) and query (read) models, enabling efficient real-time data handling through event-driven architectures. MongoDB Atlas Stream Processing further supports this by providing a managed service that allows for real-time stream processing and continuous data materialization, minimizing the operational burden and enhancing application performance at scale. This shift in data handling is crucial for adapting to agent-mediated commerce, where AI agents autonomously make purchasing decisions, requiring brands to make their products AI-friendly through technologies like MongoDB Atlas to remain competitive in the evolving e-commerce landscape.