# Document AI Document AI is Google Cloud’s document-processing suite for extracting structured data, classifying documents, splitting composite files, and parsing text and layout. Developers can create processors in the Google Cloud Console or through an API, with options for pretrained and custom processing. - Findings24: - Website: - Pricing: Freemium - Categories: AI OCR, AI Document Extraction ## What is Document AI? Document AI is a Google Cloud document-processing suite for turning structured and unstructured documents into usable data. It provides processors for extracting fields and entities, classifying documents, splitting composite files into logical document types, and detecting text and layout with OCR. The platform supports both out-of-the-box processing and customization. Custom Extractor can work across a wide range of documents and can be fine-tuned with as few as 10 documents, while Custom Splitter and Custom Classifier can use customer-provided documents and classes. Developers select processors in the Google Cloud Console or use an API; workflows can automate document ingestion, processing, and storage. Document AI can connect processed document data with other Google Cloud services. BigQuery integration enables document metadata to be written to a BigQuery objects table and joined with other tables for analytics. Enterprise Document OCR can also prepare text from scanned documents, plots, reports, and presentations for storage or machine-learning workflows. ## What can Document AI do? ### Custom extraction Extract structured data from documents with a generative-AI-powered Custom Extractor. Users can start with the model and improve accuracy by fine-tuning it with a small document set or an API call. ### Document classification Custom Classifier identifies document classes using customer-provided documents and classes, with few-shot learning, fine-tuning, and iterative auto-labeling described on the product page. ### Composite-document splitting Custom Splitter separates a document containing multiple logical classes—such as a mortgage package—into individual documents. It can be used out of the box or trained with custom classes. ### Enterprise Document OCR OCR detects and extracts text and layout information from documents, including scanned archival material. Configurable features support different document-processing requirements. ### Workflow automation Workflows can automate document ingestion, processing, and storage after a processor is selected in the Google Cloud Console. ### BigQuery connection Parsed document metadata can be extracted into a BigQuery objects table and joined with other BigQuery data for analysis of structured and unstructured information. ## Use Cases ### Automated data entry Operations teams can use Custom Extractor and Document AI Workbench to extract structured fields from business documents for mail-room, procurement, shipping-yard, and mortgage-processing workflows. ### Mortgage package separation A mortgage package containing applications, income verification, and photo identification can be divided into its component document types with Custom Splitter. ### Document analytics Teams can send document metadata to BigQuery, combine it with existing tables, and analyze parsed document information alongside structured business data. ### Archive digitization Organizations can apply Enterprise Document OCR to scanned documents, plots, reports, and presentations before storing the resulting text in cloud storage or a data warehouse. ### Machine-learning data preparation Extracted OCR text from archival content can be used as input for training machine-learning models specific to an organization’s needs. ## Frequently Asked Questions ### How do developers use Document AI? Developers can select a suitable parser in the Google Cloud Console or call a processor through an API. Workflows can automate ingestion, processing, and storage. ### Can Document AI process documents without custom training? Yes. The product page describes several processors that can be used out of the box, including Custom Extractor and Custom Splitter. Custom Splitter can also be trained from the ground up with customer documents and classes. ### How can teams improve extraction results? Custom Extractor can be fine-tuned with as few as 10 documents. Custom Classifier supports few-shot learning, fine-tuning, and iterative auto-labeling with corrections. ### Where can processed document data go? Document AI can connect with BigQuery, where document metadata can be written to a BigQuery objects table and joined with other tables. OCR output can also be saved to cloud storage or a data warehouse. ### How is Document AI priced? The product page states that pricing is based on the number of processed pages per month, the relevant quota, and any purchased capacity reservation. Listed examples include Enterprise Document OCR at $1.50 per 1,000 pages, Custom Extractor and Form Parser at $30 per 1,000 pages, Layout Parser at $10 per 1,000 pages, and Custom Splitter and Custom Classifier at $5 per 1,000 pages; applicable ranges and current pricing should be checked before deployment. ## Quick Facts - **Category**: Document processing and AI/ML - **Platform**: Google Cloud - **Primary users**: Developers and teams automating document data extraction and processing - **Access methods**: Google Cloud Console and API - **Core outputs**: Extracted structured data, classifications, split documents, text, layout information, and document metadata - **Pricing basis**: Processed pages per month, applicable quota, and purchased capacity reservation