7.2. Implement Document Intelligence

💡 First Principle: Document Intelligence transforms unstructured documents (PDFs, images of forms) into structured data you can process programmatically. The key decision is whether your document type matches a pre-built model (use it!) or requires custom training (more work, but handles any document type).

What breaks without proper model selection:

• Using custom models for invoices → Wasted training effort; pre-built works better
• Using pre-built for proprietary forms → Poor extraction; need custom model

Model selection framework:

Pre-built vs. custom trade-offs: Pre-built models work immediately with high accuracy on standard documents. Custom models require labeled training data but handle any document structure.