Box has unveiled a new feature called Box Content Retrieval, which integrates Nvidia’s Nemotron embedding system directly into its core platform. The tool is designed to enhance how AI agents process and retrieve information from stored documents.
According to Nvidia, the implementation of Nemotron is expected to deliver significant performance improvements. Specifically, the company anticipates that queries will be answered four times faster, with a 20% reduction in errors and 90% fewer tokens consumed.
The new system leverages Nvidia’s open-weight Nemotron Embed project, which focuses on improving retrieval quality for automated agents. By mapping underlying concepts rather than relying on simple keywords, the tool allows agents to identify relevant information even when different terminology is used.
Box CEO Aaron Levie noted that the system combines five distinct technologies into a single API. It also integrates with the company’s existing metadata and permission controls, which Box claims makes the solution difficult to replicate independently.
The feature is scheduled to roll out to all Enterprise Advance customers early next year. It is intended to assist professionals, such as lawyers, in locating specific clauses and precedents within large volumes of data more efficiently.
