7.4. Reflection Checkpoint

Key Takeaways

Before proceeding, ensure you can:

• Design Azure AI Search indexes with appropriate field attributes (searchable, filterable, sortable, facetable)
• Configure indexers and skillsets for AI-enriched search pipelines
• Choose the right knowledge store projection type (Tables for structured, Objects for JSON, Files for binary)
• Select the appropriate Document Intelligence model (pre-built for standard docs, custom for unique forms)
• Distinguish Document Intelligence (structured field extraction) from Content Understanding (general comprehension)
• Handle hybrid search combining full-text, semantic, and vector queries

Connecting Forward

Phase 8 brings everything together for exam readiness. You'll apply all the frameworks, service selection decisions, and implementation patterns you've learned in practice questions that mirror the actual exam format.

Self-Check Questions

1. An Azure AI Search query returns documents but the expected results aren't appearing. The index has a "department" field that's used in a $filter expression. What's the most likely cause, and what field attribute was probably missing?

2. A company has invoices from 20 different vendors, each with unique layouts. They want to extract vendor name, invoice total, and line items from all of them. Should they use pre-built models, custom models, or composed models? Why?