Azure AI Document Intelligence

Azure AI Document Intelligence (formerly Form Recognizer, and now part of Microsoft's Foundry Tools / Content Understanding capabilities) is a machine-learning-based OCR and intelligent document processing service. It extracts printed and handwritten text, tables, key-value pairs, selection marks, and document structure from forms and documents, returning structured, business-ready output through REST APIs and SDKs for .NET, Python, Java, and JavaScript.

The service is organized into three model families. Document analysis models (Read for text/OCR, Layout for text, tables, and structure) provide general extraction. Prebuilt models handle common business documents out of the box -- including invoices, receipts, bank statements, checks, pay stubs, contracts, identity documents, health insurance cards, and US tax forms (W-2, 1098, 1099, 1040) and US mortgage forms -- without any training. Custom models (custom neural, custom template, custom composed, and custom classification) are trained on a customer's own labeled documents to extract fields specific to their use cases or to route mixed document batches.

For ERP and finance teams, the invoice and receipt models are the most relevant entry points: they extract vendor and customer details, line items, totals, tax, and dates that can feed accounts-payable and expense workflows. Because the service is a Microsoft cloud product, it connects natively to Power Automate, Power Apps, and Dynamics 365 through a first-party connector, and it can also be deployed as a Docker container -- including in fully disconnected/on-premises environments under a Microsoft licensing arrangement -- for data-residency-sensitive scenarios.