Considering implementing IDR, what are the issues?

Summary: We're thinking of implementing Intelligent Document Recognition and would like advice.

Content (required): 1. What is your experience with Machine Learning with high quality and low quality files?

2. Is it possible to allow IDR to process only high quality files?

3. What prep did you do before implementing the system? (e.g. clean supplier info to match invoice info.)

4. Some stakeholders want to use a digital\manual stamp to record vital data fields: PO Number, amount, invoice number. Thoughts?

5. Did you raise an SR to port the Adaptive Learning from TEST to PROD?

6. Is it possible to implement in phases, say department by department?