Azure AI Content Understanding turns messy PDFs, images, audio, and video into structured, searchable data using schemas—with confidence scores and grounding so you can verify results. In this talk we’ll unpack analyzers and show how schemas drive field behaviors (Extract, Classify, Generate) that power automation and RAG. First Demo is a playful Pig Personality Test: attendees draw a pig, snap a photo, and watch an image analyzer extract traits and generate a summary, illustrating image support and confidence-aware outputs. 2nd Demo is an Audience Survey with a custom analyzer that reads selection marks, numbers, and free text to produce an instant session report—wired through the simple Analyze → Get Result REST pattern that returns JSON.

Takeaways

Know when to use Extract vs. Generate, set confidence thresholds, and verify with grounding to enable safe automation.

Leave with two reusable demo patterns (image + document) and the REST recipe to integrate CU outputs into apps and agents