Cloud storage company Box has unveiled a feature called Box Content Retrieval that uses Nvidia's open-weight Nemotron embedding models to help AI agents find information in stored documents, TechCrunch reports. It was announced at Box's developer conference on Thursday and will roll out to Enterprise Advanced customers early next year.
Rather than having an agent read whole files, the system returns the passages most relevant to a request, mapping concepts rather than just keywords. Nvidia executive Kari Briski told the outlet she expects answers to arrive about four times faster with fewer errors and far lower token use.
Box chief executive Aaron Levie said the product combines several technologies with Box's existing metadata and permission controls. The company gave legal research as an example, saying lawyers could locate relevant clauses and precedents more easily.
Source: TechCrunch