1.2. The Microsoft AI Platform Architecture

💡 First Principle: Different AI tasks require fundamentally different infrastructure—generating text needs transformer models, searching documents needs indexes, and analyzing images needs vision pipelines. Microsoft organizes these into separate pillars (Models, Agents, Tools, Search) so each can optimize for its workload while sharing security and governance. When an exam question asks "where does X live?", ask yourself: "Is this about generating content, searching content, analyzing content, or orchestrating actions?"

The diagram below shows how Microsoft Foundry organizes AI capabilities into four pillars. Understanding this architecture helps you navigate exam questions about where specific services live.

Azure AI Foundry Documentation