Software ate the world. Federation will eat embeddings (www.gnanaguru.com)

A recent article emphasizes the importance of federating data access for AI applications instead of creating an extensive parallel infrastructure. It argues that many enterprises mistakenly build separate AI-specific data systems, like vector databases and embedding pipelines, when they can utilize existing data sources such as CRMs and billing systems. By leveraging AI agents with the ability to query these systems directly via the Model Context Protocol (MCP), organizations can more efficiently obtain real-time insights without the need for extensive data transformation and storage, ultimately saving time and reducing technical debt. This approach is particularly significant for the AI/ML community as it shifts the focus from centralization to real-time data federation, allowing companies to avoid unnecessary infrastructure investments. The article outlines three emerging AI agent architecture patterns—simple federation, federation with ephemeral compute, and agentic AI with long-term memory—that enhance AI capabilities while minimizing redundancy. As organizations strive to remain competitive, this paradigm encourages them to start with existing technology and capabilities instead of building complex systems that may become obsolete as the landscape continues to evolve.