Box unveils AI-agent search built on Nvidia Nemotron
Box Content Retrieval will reach all Box Enterprise Advanced customers early next year, with Nvidia estimating 4x faster answers.
Box announced Box Content Retrieval, an AI-agent search feature built on Nvidia's Nemotron, at its developer conference on November 5, 2026, according to digitaltoday.co.kr. The feature uses Nemotron metadata to help AI agents answer questions more effectively.
Box says users get faster, more accurate answers at lower cost. Nvidia VP of generative AI software Kari Briski estimated that answer speed will be 4 times faster, errors cut by 20 percent and token usage reduced by 90 percent.
Box CEO Aaron Levie said agents can be sent only the document parts most relevant to a user question or agent task, without reading documents one by one.
The feature will be available to all Box Enterprise Advanced customers starting early next year.
Nvidia's retrieval work
Nvidia is developing Nemotron Embed, an open-weight embedding model aimed at improving retrieval quality for AI agents and RAG systems. Embedding models convert data such as text, images and audio into vector arrays computers can process. TechCrunch reported that major AI research companies had been reluctant to invest in such embedding models while Nvidia focused on the area.
Quick answers
When will Box Content Retrieval be available?
Box says the feature will be available to all Box Enterprise Advanced customers starting early next year.
What improvements did Nvidia estimate?
Nvidia VP of generative AI software Kari Briski estimated answers will be 4 times faster, with errors cut by 20 percent and token usage reduced by 90 percent.