Box Unveils Box Automate: Ushering in the Era of Context-Driven AI Agents
Box Unveils Box Automate: Ushering in the Era of Context-Driven AI Agents
Box has taken another major step in embedding artificial intelligence into enterprise workflows, unveiling a suite of new AI-powered features at its annual Boxworks developer conference. The highlight of the announcements is Box Automate, a new system designed to serve as an operating system for agentic AI, allowing businesses to augment and automate complex workflows involving unstructured data.
AI Agents for the Unstructured World
According to Box CEO Aaron Levie, the company's focus is on the rapidly evolving world of work, where AI's greatest impact is now on processes reliant on unstructured data—think legal reviews, marketing asset management, or M&A due diligence. While structured data automation is mature, unstructured data has remained a challenge. Levie explains that AI agents now enable organizations to finally automate these complex, document-heavy workflows:
- Review and approval processes for contracts and legal documents
- Asset management for marketing teams
- Due diligence in mergers and acquisitions
"For us, AI agents mean that, for the first time ever, we can actually tap into all of this unstructured data."
Introducing Box Automate: Modular AI-Driven Workflows
Box Automate acts as a backbone for deploying and managing AI agents within enterprise processes. The system breaks down workflows into modular segments, each handled by a specialized agent—such as submission, review, or data extraction. This modularity brings a new level of flexibility and control:
- Decide how much work each agent should perform before handing off tasks
- Reduce risk by limiting the scope of each agent’s actions
- Enable organizations to scale automation safely across departments
Box Automate workflow, with AI agents assigned to specific tasks. Image Credits: Box
Addressing AI Risks: Guardrails and Security
Levie notes that successful AI deployment depends on strong guardrails and clear demarcation points within workflows. By segmenting tasks and assigning them to different agents, Box reduces the risk of compounding errors and prevents agents from "running wild" after repeated actions. The system also lets administrators define which steps must be deterministic and which can benefit from more flexible, agentic AI behavior.
Data security and compliance remain central. Box leverages its long-standing expertise in access controls and permissions, ensuring that AI agents only access and act upon data users are authorized to see. This robust governance layer is built into every AI-driven workflow.
Future-Proofing AI for the Enterprise
Box’s approach is model-agnostic: enterprises can choose their preferred AI models and aren’t locked into a single provider. As AI models and agentic capabilities improve, Box’s architecture is designed to seamlessly incorporate these advances—future-proofing enterprise automation as the field evolves.
Competing in a Rapidly Evolving Landscape
With foundation model companies like Anthropic introducing new features for file management, Box distinguishes itself by providing:
- Enterprise-grade security, permissions, and compliance
- Deep integration with storage and content management
- Flexible connectivity to leading AI models and APIs
Box Automate represents a significant leap for organizations seeking to safely and effectively harness the power of AI for the most challenging, unstructured workflows—marking what Levie describes as the true beginning of the AI "era of context."
