• I've read a few ideas here
• Identifying whether an uploaded document is a template might be a useful feature we can accomplish.
  • It's a basic classification task. We can also find ways to extract named entities from the documents and remove them from the document to get a template. Then ask the user to set these entities and get a full document.