Box Unveils Architect: A Smart Workflow Builder for Document Management

Box has announced the launch of Architect, a new natural-language AI tool designed to assist users in creating automated workflows for document management. This innovative tool was unveiled during the BoxWorks 2026 conference.

Intuitive Workflow Creation

Architect serves as an intelligent assistant that allows users to outline their automation goals in simple language. Once a user specifies the desired workflow, Architect automatically shapes the required stages and decisions, taking into account existing permissions, classifications, and retention policies that relate to the content managed in Box.

Before finalizing the workflow, users have the opportunity to test it. This testing phase includes the option to incorporate specific criteria for human reviews and approvals. Once the workflow is active, Architect continuously monitors its performance, offering suggestions for optimization based on its usage.

Integrating Seamlessly with Box Tools

Architect integrates smoothly with a range of established Box tools. For instance, in a banking setting, it could facilitate the creation of a client onboarding process, gathering applications through Box Forms, verifying submissions with Box Extract, signing documents with Box Sign, and initiating account setup with Box Automate.

Jason Andersen, vice president and principal analyst at Moor Insights & Strategy, highlighted that Architect empowers teams to develop workflows drawing from their collective expertise rather than focusing solely on individual productivity. However, he emphasizes the necessity for businesses to continually assess the performance of these workflows.

“Human oversight should still play a vital role in any project involving agent development to ensure compliance with risk and cost management,” Andersen recommended.

Beta Launch and Pricing Insights

Architect is anticipated to enter beta testing in Q4 2026, specifically for Box Enterprise Advanced customers. While specific pricing details are not yet available, it's known that this tier includes an allocation of 20,000 AI Units, which are a measure of usage for certain AI features within Box.

Enhancing Search Capabilities with Context Retrieval

In addition to Architect, Box also revealed a new Context Retrieval engine aimed at improving the accuracy of results produced by AI agents linked to documents. According to Box CEO Aaron Levie, this tool is set to enhance the relevance of the information delivered to users.

Built on Nvidia’s cutting-edge Nemotron Retriever model, the Context Retrieval engine allows for expedited indexing while maintaining low latency. This feature is designed to deliver precise enterprise context to AI agents quickly and effectively. It leverages multimodal indexing, scanning not just text but also images, audio, and video, optimizing the search process through both keyword and semantic queries.

The internal testing of this engine revealed a remarkable 20% increase in search accuracy, along with a fourfold enhancement in question-answering speed. Additionally, it managed to cut down the number of tokens sent to downstream models by up to 90%.

Andersen observes that such context retrieval systems are gaining traction and are pivotal in enhancing output accuracy, agent effectiveness, and overall user satisfaction, while also reducing operational costs.

Looking Ahead

Box is planning for the Context Retrieval engine to be generally accessible to Enterprise Advanced customers early next year. By combining these advanced tools, Box continues to strive toward creating a more efficient and user-friendly environment for document management.