Azure AI Services (formerly Cognitive Services) are pre-trained models that you can call in your .NET code via simple REST APIs.
- **Vision:** Face recognition, OCR (reading text from images), and content moderation.
- **Speech:** Real-time transcription (Speech-to-Text) and natural-sounding voices (Text-to-Speech).
- **Language:** Sentiment analysis, translation, and key-phrase extraction.
This is a 'must-know' for enterprise architects. It uses AI to extract structured data (tables, key-value pairs) from PDFs and scans. You can train it on your own forms (e.g., invoices, tax returns) with just 5-10 examples. It's much more reliable than traditional OCR.
Q: "Should I use pre-trained models or build my own?"
Architect Answer: "Always start with **Pre-trained Services**. They represent billions of dollars in R&D and work instantly. Only move to 'Custom Machine Learning' (Azure Machine Learning) if your problem is highly specialized and these services can't achieve your required accuracy."