Box taps Nvidia’s Nemotron for new agent-first content retrieval system

AI agents now retrieve content four times faster with 90% fewer tokens used.

Box announced a new feature, Box Content Retrieval, which integrates Nvidia's Nemotron embedding system into its core product. This integration aims to significantly enhance how AI agents interact with and retrieve information from Box's content repositories. The system leverages Nemotron's ability to create rich metadata, allowing agents to understand and process user queries more efficiently. According to Nvidia, this will result in queries being answered up to four times faster, with a 20% reduction in errors and a 90% decrease in token usage. Box CEO Aaron Levie highlighted that this enables agents to quickly access the most relevant document segments rather than processing entire documents. The feature is an implementation of Nvidia's Nemotron Embed project, which offers open-weight embedding models designed to improve automated retrieval systems. While Nemotron models are openly available, Box's integration relies on its existing metadata and permission controls, making it a proprietary solution. Box Content Retrieval will be available to all Enterprise Advance customers early next year.

The partnership between Box and Nvidia signifies a growing trend of enterprise software vendors embedding advanced AI, specifically generative AI embedding models, to improve core functionalities. Box's move to integrate Nvidia's Nemotron demonstrates a strategic effort to differentiate its content management platform by offering superior AI-driven search and retrieval capabilities, moving beyond traditional keyword-based searches to concept-level understanding. This also highlights Nvidia's expanding influence in the AI software layer, moving beyond hardware to provide foundational models that power new application features across various industries.

This development is particularly relevant for businesses relying on large volumes of unstructured data, such as legal or research-intensive organizations, where efficient information retrieval is critical. By making queries faster and more accurate, Box Content Retrieval aims to reduce operational costs and boost productivity for users interacting with AI agents within the Box ecosystem. The focus on agent-first retrieval suggests a future where AI agents are not just passive tools but active participants in information discovery, powered by sophisticated embedding technologies.