Box AI Built an Agent for Corporate Content on LangChain's Deep Agents

Box, a corporate content management platform with more than 100,000 customers worldwide, has transitioned its AI system to the Deep Agents architecture from LangChain. The new Box Agent is capable of independently searching, analyzing, and synthesizing information from corporate repositories — while fully respecting access rights already configured in the system.

From Chatbot to Agent

The difference between a regular AI chat and an agent is fundamental. A regular chat answers a question based on one document or a small fragment. An agent — plans a task, calls the necessary tools, iteratively refines the result, and is capable of working with thousands of files simultaneously.

Box Agent is built on the Deep Agents framework from LangChain. This means the agent can:

• Decompose complex questions into subtasks
• Call search and analysis tools in the right order
• Work with long contexts from multiple documents
• Iteratively check and refine its conclusions
• Generate structured reports with links to sources

For a corporate user, this means: you can ask a question like "What is our travel expense policy and how has it changed over the past two years?" — and receive a coherent answer assembled by the agent from dozens of documents without manual search.

Security Without Compromises

The main challenge for corporate AI is not the accuracy of answers, but respecting access rights. In a large organization, there are hundreds of user groups with different access levels to documents. If an AI agent does not account for these restrictions, it becomes a serious security hole and compliance violation.

Box solved this problem by integrating Box Agent directly into the existing permissions system. The agent only sees files that a specific user has access to at any given time. If a document is marked as confidential or accessible only to a specific department — the agent will not disclose its content to those who do not have the right to see it.

This is critically important for industries with strict security requirements: financial companies, law firms, medical organizations. Box is most actively represented in these sectors, and it is here that access control requirements are most stringent.

Independence from Provider

Another key feature of Box Agent is complete independence from any specific AI provider. The Deep Agents architecture allows switching between different language models without reworking agent logic. In practice, this means: if one model is optimal today, and a more powerful or cost-effective one emerges tomorrow — Box can update the backend without changes to the user-facing product.

In an environment where new AI models are released every few weeks, such architectural flexibility becomes a significant competitive advantage.

LangChain acts here as an orchestration layer that abstracts the business logic of the agent from specific LLM providers. Box can optimize model selection for each task: one for search, another for synthesis and generation of the final report.

What This Means

Box Agent is one of the first examples of a mature corporate platform transitioning to a deep agents architecture in production. This is not a pilot or a demo: Box serves the world's largest companies, and betting on Deep Agents means that agent AI is ready for enterprise load with real security and scalability requirements.

For the rest of the market, this is a signal: the pattern "agent plus corporate access rights" works and is becoming a working template for the industry.