Box launches Architect, an AI-powered workflow builder

Box has introduced a new natural-language AI tool that helps users build automated workflows around document-related processes. Box Architect, announced at BoxWorks 2026 on Thursday, “effectively acts as an agentic operator of Box’s core content and workflow capabilities … to help customers design any content agentic workflow in Box,” said Aaron Levie, Box CEO, on X. To create a workflow, users describe in plain language a goal or process that they want to automate. Architect then defines the necessary stages and decisions, applying existing permissions, classifications, and retention policies related to content in Box. Users can test the workflow before it’s published, with options to insert requirements for human review and approvals. Once deployed, Architect can monitor a workflow’s outputs and recommend improvements as it’s continuously used. Architect interacts with multiple existing Box tools. A bank might use Architect to design and build a client onboarding workflow that collects applications via Box Forms, for example, then verify the documents with Box Extract, sign agreements with Box Sign, and begin account setup with Box Automate. Jason Andersen, vice president and principal analyst at Moor Insights & Strategy, said Architect will make it easier to build agents and workflows based on the collective knowledge of a team or business, rather than focused only on their own productivity. As with any AI system that involves business data, customers will still need to evaluate how those workflows perform, said Andersen. “I’d still recommend human oversight during the creation process with any agent development project and using observability tools to make sure that the agents are running in a fashion that meets the risk and cost management needs of the business,” he said. Architect is due to launch in beta in Q4 2026 for Box Enterprise Advanced customers. Pricing has not been disclosed. Box doesn’t publish a price for Enterprise Advanced, its highest tier. Enterprise Advanced includes an allocation of 20,000 AI Units that are used to measure consumption of certain Box AI features. Box also announced a new Context Retrieval engine that Levie said will “deliver better results for agents” that connect to documents held in Box. Built on Nvidia’s Nemotron Retriever open model, Context Retrieval enables faster indexing and lower latency is “designed to find and deliver more relevant enterprise context to AI agents,” said Box CTO Ben Kus. It uses multimodal indexing to scan documents, images, audio, and video, and combines keyword and semantic search to find information when a user’s query doesn’t contain the exact terms used in a document. The Content Retrieval engine applies existing permissions in Box so that agents can only retrieve what a user is authorized to access. Internal testing showed that tool led to a 20% improvement to search accuracy, said Kus, with “fourfold faster question-answering.” It also resulted in “up to 90% fewer tokens sent to downstream models,” he said. Andersen from Moor Insights sees a wider increase in use of context retrieval systems more generally. These types of tools are “significantly improving output accuracy and agentic performance, both of which lead to higher end user satisfaction and lower costs,” he said. Box’s use of Nemotron’s open-weight model further improves the value proposition, according to Andersen. “However, to get the most benefit from context management, a good set of design principles and governance are also needed, so customers need to be sure to account for that effort to get the most benefits,” he said.  The Context Retrieval engine will be generally available to Enterprise Advanced customers early next year, Box said.