OCR Output Should Match the Job

Here's a simple way I think about it:
Need editing?
→ Word
Need calculations?
→ Excel
Need searchable archive?
→ PDF
Need structured records?
→ JSON/CSV/XML
Need documentation?
→ Markdown
Need mathematical notation?
→ LaTeX
Need diagrams?
→ Mermaid/Graphviz
The OCR step is the same idea, but the output changes the entire workflow.