1. Extract data from *any* document types with minimal setup.
2. Customize the JSON format you receive as a response.
3. Save loads of time on tedious pre-training tasks.