In the context of Document Understanding, which activities do NOT consume AI units?

In Document Understanding, activities are designed to process documents and extract relevant information. The activities that do not consume AI units are generally those that either involve basic processing of documents or human intervention that does not rely on AI capabilities.

The pair of activities that include Digitization and Human in the Loop are the correct choice because digitization refers to converting documents into machine-readable formats (often performed using Optical Character Recognition, or OCR), which typically does not require AI units. Human in the Loop refers to the involvement of a human to validate or correct data but isn't a computationally intensive AI process itself; rather, it serves as a quality assurance step.

In contrast, activities such as Classification and Extraction are fundamentally reliant on AI capabilities. Classification involves understanding and categorizing documents based on learned patterns, while Extraction focuses on pulling specific information from documents. Both tasks require AI models to effectively analyze and interpret data, which consumes AI units.

Thus, the correct answer highlights those activities that are foundational and do not leverage advanced AI resources, allowing organizations to perform necessary document processing without incurring additional AI unit costs.