Mastering Box Extract for Metadata Extraction at Scale

Learn how you configure your Extract Agents and write per-field AI instructions to metadata extraction at scale. This roundtable brings together Box product experts and early adopters to share practical prompting techniques, agent selection strategies, and iterative refinement workflows that enable you to make your content into AI ready and drive your processes. We will draw lessons from having deployed Box Extract to millions of documents at scale.

Topics we'll cover:

• How to write effective per-field AI instructions that dramatically improve extraction accuracy
• When to use Standard vs. Enhanced Extract Agents (and how document complexity should drive that decision)
• How to use confidence levels (new Jan 2026) to identify which fields need better prompts
• Iterative prompt refinement workflows: start simple, measure, improve -- build an AI system for your documents
• Common prompting anti-patterns that hurt accuracy
• How extracted info is being used to power AI workflows and applications