Box has introduced Box Content Retrieval, integrating Nvidia's Nemotron Embed to optimize how AI assistants interact with document sets.
By leveraging models that perform conceptual rather than keyword-based searches, the tool aims to improve retrieval relevance for complex tasks.
Developers and enterprise users can expect the feature to roll out in early 2027, with performance targets suggesting significantly reduced token overhead and improved accuracy for large-scale data queries.
Nemotron Embed prepares data in advance to aid in search. Instead of just searching for the same words, it can find related topics even when they are described in different words or in a different language. For example, according to Box, lawyers will be able to more easily find relevant clauses and legal precedents.
Kari Briski, Nvidia’s vice president of productive AI software, estimates that queries will be answered four times faster, with 20% fewer errors and 90% fewer units of text processed by the AI. These numbers are Briski’s expectations.
Box CEO Aaron Levie said the service can return snippets that best match the AI assistant’s query or task. The service also leverages the information Box already has about documents and the controls that determine who can see them. According to Levie, this reliance on Box’s existing capabilities makes it difficult to rebuild the service from scratch. Nemotron models are available for free.
↳
READ ALSO
HACKS.GR / RELATED
HACKS.GR
AI Unlocked
Britain creates COBRA team to detect forest fires faster
↗
Box says the feature will be rolled out gradually in early 2027 to all customers with the Enterprise Advance package. The Nemotron version released in July ranked first in RTEB, a test that measures how well relevant information is identified.
>Stay connected with Hacks.gr
Set us as a preferred source on Google and find us on social media.
Preferred source
↗
end of article
Top comments (0)