Workbench Custom Extractor is now GA. Try it now below.
Use artificial intelligence to process documents—extract, classify, and split. With Document AI Workbench, you can now use foundation models to quickly create parsers without the need to label, annotate, or train new models.
Custom Extractor is now powered with generative AI to parse documents with speed and accuracy
Fine-tuning is now available for foundation models
Create a custom classification or splitting model quickly
Creating a processor is free of charge
New Google Cloud customers get $300 in free credits
Benefits
Higher accuracy
Experience fewer errors and faster, more-accurate results using generative AI. Fine-tune foundation models to achieve even higher extraction accuracies.
Lower cost
Spend less time labeling and annotating datasets with generative AI, fine tuning and auto-labeling tools
Faster time to value
From unstructured documents to fixed layout forms, you can now use generative AI to process your business documents quickly and out of the box.
Demo
Extract data from your documents using generative AI. For full product capabilities head to Document AI in the Google Cloud console.
Key features
Uses generative AI to accurately parse and extract data from structured and unstructured documents in a production environment. The Custom Extractor provides a simple way to create such processors with a UI or API.
To achieve higher accuracies, Custom Extractor enables you to fine tune a foundation model in cases where you have little or no training data available or when the layout variation of a document is high. These documents can be contracts, terms of service, invoices, bank statements, bills of lading, payslips. The fine tuning feature is simple and easy to use.
Create a processor that uses artificial intelligence to identify a document type from a user-defined set of document classes, helping save time and resources. Use this feature to validate document intake, automate downstream processes, or redirect a document for further processing.
Custom Splitter helps developers create a processor that identifies classes of documents from a user-defined set of classes. The Custom Splitter analysis will provide the class (or document type) and page number for a file containing multiple documents. For example, the Splitter will provide the page numbers of a driver’s license, paystub, tax form, and bank statement within a single file.
What's new
Sign up for Google Cloud newsletters to receive product updates, event information, special offers, and more.
Documentation
You can use generative AI to create a document parser in minutes. Fine tune the model for higher accuracies or auto-label a new dataset as needed.
Create custom classifiers that identify documents from a user-defined set of classes.
Custom splitters can split and classify multiple documents within a single file. Your custom splitter is trained and evaluated with your data.
Learn how to apply labels from your model schema to imported documents in your dataset.
Train template based processors to extract data from forms with fixed layouts or train a custom model to extract data from documents with moderate layout variation.
An evaluation is automatically run whenever you train or uptrain a model. See how to run a manual evaluation to get updated metrics after modifying the test set.
Use cases
Use Document AI Workbench to automate data entry by extracting structured data from your documents. Enable downstream analytics or search features. Use this data to make more efficient and effective business decisions.
Try out the Custom Extractor.
Technology customers and partners can improve and expand their document processing solutions quickly with generative AI. With a simple API prediction endpoint and JSON document response format, you can integrate powerful Document AI processors into your platform or solution.
Quickly create accurate document processors to classify and extract data from supply chain and procurement use cases. These can include purchase orders, invoices, bills of lading, vendor intake, receipts, and more.
Create processors to help automate banking, lending, and insurance workflows. These include financial statements, insurance forms, account documents, identification, and KYC documents.
Pricing
With Document AI Workbench, you pay only for hosting, auto-labeling, and prediction (processing). Importing and training are free of charge.
New pricing for 2024 is on the link below.
Partners
Get help with implementing Document AI Workbench with these partners.
Start building on Google Cloud with $300 in free credits and 20+ always free products.