Introduction videos & labs

The videos on this page cover the fundamental concepts of Document AI. Each video has companion labs which provide step-by-step guides to get experience using the features presented in the video.

What is Document AI?

Welcome to The Future of Documents, a new series that helps you make the most of your unstructured data with Document AI. There's an enormous amount of data in the documents we interact with, such as PDFs and emails, that contain unstructured or "dark" data. Join us for this exciting new series where we show how you can use this data to automate processes, gather analytics, and much more!

Document AI Introductory Codelab

How to use Document AI

Want to learn how to use Document AI? This video shows how you can get started with Document AI using the Cloud Console or with the API using the Python Client Library. Watch to learn about Document AI processors and see how you can create your own!

Document AI Introductory Codelab

Document AI Processor Management Codelab

General Processors in Document AI

In this video, we take a deep dive into General Processors. These are pretrained and ready to use models designed to work on any document. Watch along and learn how to get detailed structured information from documents with General Processors in Document AI.

Document AI OCR Codelab

Document AI Form Parsing Codelab

Specialized Processors in Document AI

In the previous episode, we discussed General Processors in Document AI for most documents. However, did you know that there are more specialized processors for documents such as invoices, contracts, tax documents, and identification cards? In this episode, we discuss how to use Document AI for procurement, identity, lending, and contracts.

Document AI Specialized Processors Codelab

Document AI Workbench

Previously, we looked at Document AI Specialized processors and how they can provide intelligent insights into certain document types by providing consistent schemas and normalizing data to reduce the amount of pre/post-processing you need to perform. Document AI Workbench allows Uptraining of pre-trained specialized processors and creating Custom Processors from scratch, without needing any machine learning expertise.

Document AI Workbench: Uptraining Guide

Document AI Workbench: Custom Document Extractor Guide

Document AI Workbench: Custom Document Classifier Guide

Document AI Workbench: Custom Document Splitter Guide

Human-in-the-Loop

Previously, we've looked at lots of ways to process documents, including General Forms, IDs, Receipts, and other specialized document types. However, no Machine Learning model is perfect, and when you need the highest accuracy for your applications, it can be best to have people verify the results before you use them. Human-in-the-Loop (HITL) allows you to send documents to a dedicated Human Review team to verify the results and make corrections if needed.

Human-in-the-Loop Codelab

Document AI Warehouse

Document AI Warehouse is a platform to store, search, organize, and analyze documents and their structured metadata. Documents can include structured data such as forms and invoices as well as unstructured data such as contracts and research papers. The metadata for documents can be automatically extracted using processors in Document AI or manually input using fields and tags.

In this codelab, you will learn how to ingest, process, and search documents using the Document AI Warehouse user interface. Sample PDF documents are provided for this codelab, including a license agreement, loan form, and order invoice.

Document AI Warehouse: Codelab

Cloud OnAir Innovators Live: Getting Started with Document AI

Original air date: February 16, 2023

In this session, we'll cover the basics of processing documents, answer your questions about the technology, and show some demos for how to get the most out of Document AI.

Google Cloud Next '23: Generative AI

Presentation date: August 30, 2023

Generative AI is changing the way you process invoices, customer forms, and deliver document-based services. Using natural-language prompts with high accuracy and limited machine learning training, you can classify, extract, and glean deeper insights from document data. See how quickly others have learned to use our generative AI-based features.