Legal documents aren't just made up of text. They're full of charts, diagrams, signature blocks, floor plans, org charts, and tables that lose critical meaning when reduced to Optical Character Recognition (OCR) output in AI systems. When a lawyer asks an AI tool, "what does the revenue chart on page 47 show?", they need an answer that actually reads the chart and understands the colors of each bar and how they map to the chart key, rather than simply extracting the numbers and text as chunks.