Box integrates nvidia's nemotron system for advanced agent content retrieval

Box has introduced a new capability, Box Content Retrieval, which integrates Nvidia's Nemotron embedding system directly into its platform. This feature enables AI agents to answer user queries with greater precision by utilizing the metadata generated by Nemotron. Box anticipates that this integration will result in queries that are faster, more accurate, and less costly for end-users. Nvidia's VP of generative AI software, Kari Briski, projected that the system could deliver answers four times faster, reduce errors by 20%, and decrease token usage by 90%. The functionality is slated for release to all Box Enterprise Advance customers early in 2027. Box CEO Aaron Levie noted that this significantly enhances the relevance of search technology for agents, allowing systems to return the most pertinent document segments rather than requiring agents to scan entire files. The implementation leverages Nvidia's Nemotron Embed project, which provides open-weight embedding models for improving automated retrieval. The system is particularly beneficial for applications like legal research, as Nemotron's embeddings map underlying concepts beyond simple keywords. While the models are free, the feature relies heavily on Box's existing metadata and permission controls, which Levie described as combining five different technologies into a single API.