2.1. Select the Appropriate Azure AI Services

💡 First Principle: Once you've identified the service family using input type (Phase 1), service selection becomes a feature-matching problem: which specific service within that family has the capabilities you need? Think of it like choosing a tool from a toolbox—you know you need a screwdriver (the family), but should it be a Phillips, flathead, or Torx (the specific service)?

What breaks without proper service selection: Using the wrong service wastes development time and money. Imagine building a custom ML model for invoice extraction when Document Intelligence's pre-built invoice model achieves 95%+ accuracy out of the box. Worse, using a generic OCR service when you need structured field extraction means you'll spend weeks writing parsing logic that Document Intelligence provides automatically.

Consider a real scenario: Your company processes thousands of customer support emails. You need sentiment analysis, key phrase extraction, and entity recognition. Do you build three separate integrations? No—Azure AI Language provides all three through a single API call. The exam tests whether you recognize that combining capabilities within a service family is almost always better than mixing multiple services.

Recall from Section 1.3 that input type determines service family. Azure AI services fall into two main categories: pre-built APIs for common tasks (Foundry Tools) and generative AI models (Foundry Models).