The first step should be an Azure Content Understanding analyzer within Microsoft Foundry Tools.
This specific tool is designed to handle the complexities in this scenario--scanned text, complex tables, and multi-column layouts--by converting unstructured documents into a structured Markdown format. This preserves the document's original hierarchy and structural relationships, which is critical for accurate downstream reasoning by an agent.
Incorrect:
[Not A]
Azure Language is designed for text-based natural language processing (NLP) like sentiment analysis or PII redaction on raw text strings. It cannot parse scanned images, optical character recognition (OCR), tables, or complex multi-column visual layouts.
[Not C]
The first step in this specific Microsoft Foundry agent configuration should not be an Azure OpenAI Responses API call.While multimodal models can read images, they frequently hallucinate data, fail to capture complex multi-column reading orders, and degrade table structures when processing dense, mixed-format documents.
Reference:
https://learn.microsoft.com/en-us/azure/ai-services/content-understanding/document/elements