Your team is working on a project to extract key information from a large volume of medical records. What are the key considerations you should keep in mind while implementing a custom document intelligence model for this use case?