When humans process unstructured data the old-fashioned way, we employ a cadre of sophisticated capabilities our brain has learned and refined since birth. Eventually, a human’s ability to identify objects, faces, and text from images or words and sounds from audio is automatic, effortless, and implicit. Machines attempt to mimic these capabilities with deep learning neural network models iteratively trained on very large datasets. This may seem obvious, but it’s worth reiterating: AI that synthesizes human (or above-human) capabilities as digital solutions can instantly scale to meet the needs of a business, where the challenges of growing and managing a human workforce are no longer the limiting factor. In this chapter, I’ll demonstrate the capabilities of Amazon Bedrock Data Automation for multimodal data extraction. I’ll discuss and compare the capabilities of other AWS services utilizing both probabilistic and generative AI approaches for each of the modalities: images, video, audio, and text. I’ll dive deep on Intelligent Document Processing (IDP)Intelligent Document Processing (IDP), a very common use case spanning many industries. IDP consists of processing documents—as scanned images or PDFs—to extract data from form elements like key-value pairs, tables, or even visual elements like charts and graphs.