Box unveiled Architect at BoxWorks on November 5, proposing an AI layer that designs, tests and assembles workflows around content already stored in Box. The announcement combines a planned workflow builder with security controls that are available today.
Architect is intended to translate a business goal into stages, decisions and handoffs, then connect agents, forms, applications and automations. Teams can test outputs and routing before deployment and add human approvals at consequential steps. Box says existing content permissions and policies will govern the work.
Availability differs across the announcement. Architect is scheduled to begin a beta rollout to Enterprise Advanced customers in Q4. New threat protection, including detection and responses for suspicious downloads and account activity, is planned for early next year. Agent security controls are already generally available to Shield and Enterprise Advanced customers.
Those available controls include classification-based access policies, limits on agent tool actions and alerts when activity exceeds customer thresholds. The distinction matters: workflow design and future threat defenses should not be treated as already delivered. Box is describing a path toward broader process automation while adding controls for the access that agents have today.