What changed
At its developer conference, Box unveiled a new feature called Box Content Retrieval, which integrates Nvidia’s Nemotron embedding system into its platform. This integration allows agents to answer user queries more effectively by utilizing the embedded metadata generated by Nemotron. Box claims the update will result in faster, more accurate, and less expensive queries.
What this means in practice
The new system will be available to all Box Enterprise Advance customers, with a rollout expected early next year. Nvidia’s VP of generative AI software, Kari Briski, stated that queries will be answered four times faster, with 20% fewer errors and 90% fewer tokens used. Box CEO Aaron Levie highlighted that the feature will enhance search technology for agents by returning relevant document chunks instead of requiring them to read through entire documents. The system is an implementation of Nvidia’s Nemotron Embed project, which provides open-weight embedding models to improve automated retrieval systems. It is particularly useful when paired with active agents, such as lawyers, who can more easily find relevant legal clauses and precedents.
Limitations and availability
The system relies on Box’s pre-existing metadata and permission controls, making it challenging to recreate from scratch. While Nemotron models are freely available, the full system requires Box’s infrastructure. Levie noted that the feature likely incorporates five different technologies into a single API. The system will be most effective with active agents and is expected to improve retrieval quality by mapping underlying concepts beyond simple keywords.
