Document AI has been extremely useful for us in OCR implementation. It was pretty straightforward to train, get up and running, and to integrate into our existing flow. The reader has been very accurate, even with a wide variety of documents that we process, without much additional training. Review collected by and hosted on G2.com.

Some parts of the documentation were not 100% up to date, or were a bit ambiguous to implement. I would've wished for more code examples for different use cases than they currently have. It was a bit unclear how to train (e.g. what parts of the desired text to include in training data exactly?), so more clear instructions especially for non-technical implementers would be greatly appreciated! Review collected by and hosted on G2.com.