1.1. Why "Apps and Agents," Not "Services"
💡 First Principle: AI-102 asked "which service does this task?" AI-103 asks "how do you assemble a production system?" The rename from "Azure AI Services" to Microsoft Foundry reflects a real architectural shift — the unit of work moved from a single API call to an orchestrated app or agent that grounds on data, calls tools, and is governed and observed end to end.
If you studied the older AI-102 exam, the instinct to reach for a dedicated service for every task — a Vision endpoint here, a Language endpoint there — is exactly the habit this exam tests against. The cost of carrying that habit is real: you'll pick a narrow service-by-service answer when the exam wants the platform-level one. Where AI-102 rewarded knowing which box to call, AI-103 rewards knowing how the boxes fit into a grounded, multi-step, governed solution.
The mental model: think of the difference between owning a pile of power tools versus running a workshop. AI-102 was the pile of tools — each excellent, each separate. Foundry is the workshop: the tools are still there, but now there's a shared workbench (projects), a single power supply and lock on the door (identity and governance), and a way to chain operations into a finished product (agents and orchestration).
⚠️ Common Misconception: "Foundry is just the new name for the Azure OpenAI resource." It isn't. Azure OpenAI capability is one tenant of Foundry, alongside the Agent Service, the model catalog, evaluations, and connections to other Azure services — all organized under projects. Treating Foundry as a renamed OpenAI resource will steer you wrong on any planning or resource-hierarchy question.
