Box Extract is an AI-powered data extraction service within the Box content platform that uses agentic AI—large language models (e.g., Google Gemini, Anthropic Claude, OpenAI GPT), advanced OCR, chain-of-thought prompting and RAG—to automatically pull structured data from unstructured content (documents, images, spreadsheets, video) and save it as searchable metadata; users can pick standard or enhanced extract agents, build custom agents and call Extract APIs to automate high-volume, complex document processing, improve search and dashboards, trigger workflows, and maintain enterprise-grade security with human-in-the-loop confidence controls.