About this Video

• Video Title: Agents in action: Document processing 2.0 | OD814
• Channel: Microsoft Developer
• Speakers: Gwenael Bego, Antonio Rodrigues
• Duration: 00:36:19

Introduction

This Microsoft Build video demonstrates new Copilot tools and agents for automating document processing workflows. The presenters showcase a new out-of-the-box document processor agent, its customization options, and the creation of custom agents using custom tools. The video highlights the ease of setup, human-in-the-loop capabilities, and integration with other Microsoft tools like Power Automate, Power Apps, and Teams.

Key Takeaways

• Easy and Fast Agent Creation: The new document processor agent is readily available in Copilot Studio and features a setup wizard for quick configuration and customization, reducing setup time from days to minutes.
• Human-in-the-Loop Control: The workflow incorporates human validation through a monitoring and validation station app and Teams approvals, ensuring accuracy and control. Conversational interaction with the agent allows for real-time status checks and clarification.
• Multimodal Capabilities and Extensibility: The agent handles various data types (documents, images, emails), and integrates with external data sources (Salesforce, Dataverse, etc.) through prompts. Power Fx integration enables precise calculations and deterministic outputs. The agent is reusable and can be invoked by other agents.
• Enhanced Customizability: New Copilot tools provide granular control over agent tasks. Users can define custom instructions in natural language and create validation rules to manage data extraction and formatting.
• Improved Custom Document Processing Models: AI Builder's custom document processing model now leverages Azure Document Intelligence 4.0, offering improved OCR capabilities and pre-built models for various document types (invoices, contracts, healthcare forms, tax forms). The ability to output in Word format is under development.