Box launches Automate, AI agents to handle enterprise grunt work

Box is pushing deeper into enterprise AI with Automate, a new service Aaron Levie said would take over tedious work that still drains time inside large companies. The product is aimed at tasks such as invoice management and document extraction, with AI agents plugging into business workflows to pull data from corporate files before a human makes the final decision.

The pitch is less about replacing core systems than about making them usable. Box says Automate runs on a workflow engine built for conditional branching, parallel processing, iterative loops and advanced routing. Customers can create and deploy custom agents without coding, using Box AI Foundation Agents such as Q&A, Compose, Search and Research. In practical terms, that means the system is designed to read, route and organize documents at scale, not just answer prompts.

The limits matter as much as the promise. Box support says only Enterprise Advanced customers get the full suite of agentic workflow automation, including AI agents, Box Extract integration, Box Forms integration, Box Doc Gen integration, Box Apps integration and Hubs support. Every agent run inside an Automate workflow consumes Box AI Units, giving customers a built-in meter for how much AI work they are actually buying.

That structure suggests the first real test will be operational, not theatrical. If Automate is doing useful work, customers should be able to measure faster invoice cycle times, fewer manual extractions from lease agreements, finance documents and other unstructured files, and less staff time spent on repetitive routing. Box’s own platform is built around unstructured data storage, AI-powered extraction, search, governance, retention and security, and it says its AI can tap models from OpenAI, Anthropic and Google. The question is whether those capabilities cut labor or simply add another layer of software between employees and the files they already manage.

Levie has argued that companies are unlikely to rebuild core ERP or CRM systems with general-purpose chatbots because those systems are too important and too risky to recreate from scratch. That position fits Box’s broader strategy: instead of selling a chatbot fantasy, it is trying to turn its content platform into infrastructure for safe automation on top of existing workflows.

The launch comes as Box still faces skepticism from investors about how far enterprise AI can go. The company reported fiscal 2025 revenue of $1.09 billion and fourth-quarter revenue of $280 million, and said it had more than 100,000 paying organizations as of January 31, 2025. Enterprise Advanced launched in January 2025, and Box has spent much of the past year moving toward agentic AI workflows, including Box Extract for pulling actionable data from unstructured content. Automate is the latest test of whether that shift produces real productivity gains or just a more sophisticated interface for the same old office grind.