About Google Document AI
Where your scanned PDFs get turned into rows.
Google Document AI is Google Cloud's managed document-parsing service, sitting between a document file and a machine-learning model that reads it. The output is a Document object: full text from the OCR pass, a page-level layout, and a list of entities with a confidence score, a mention text, a normalised value and a bounding box on the page. The same shape is returned by every processor, which is what makes it usable as a warehouse table instead of one custom parser per document type.
The product ships pre-trained processors for the documents most businesses already deal with: Enterprise Document OCR for digitising text in over two hundred languages, Form Parser for key-value pairs and tables, Layout Parser for RAG-style chunking, and a set of specialised parsers for invoices, expense receipts, bank statements, pay slips, W2s, utility bills, US passports and US driver licences. Where none of those fit, Document AI Workbench lets a team train a Custom Extractor or Custom Classifier on their own labelled documents, and Custom Splitter cuts multi-document PDFs into the right pieces. Documents are processed one at a time through the process endpoint or in bulk through the batch process endpoint, and a Human-in-the-Loop review step can be wired in for the fields where confidence is low.
