Microsoft has expanded Azure Content Understanding with features meant to turn messy enterprise files into structured data that AI applications can use more directly.
The update adds broader support for the GPT-5 model series, new synchronous Read and Layout APIs, semantic chunking for retrieval workflows, and improved confidence scoring. Microsoft also says the refreshed CU 1.0 API is generally available for production workloads, while newer capabilities target more complex extraction from documents, forms, tables, images, audio, and video.
The practical change is that developers can rely less on custom parsing pipelines before feeding business content into agents or search systems. Semantic chunking helps split documents into pieces that preserve meaning for retrieval-augmented generation, while agentic document reasoning is intended for extraction tasks that require more than a single pass over a file.
The release does not remove the need to validate extracted data. Its strongest value is in reducing the amount of glue code required to prepare enterprise content for AI systems while keeping confidence scores and structured outputs visible to developers.