Box adds Nvidia Nemotron to speed up AI search in documents

In short: Box is rolling out a new Box feature that uses Nvidia’s Nemotron to help AI agents find answers in documents faster and more accurately.

What happened

Box introduced a feature called Box Content Retrieval at its developer conference. Box is a company that stores and manages files for businesses, like a shared online filing cabinet.

The new feature builds on Nvidia’s Nemotron Embed system. “Embeddings” are a way to turn text into a kind of map of meaning (like adding smart index cards to every document, not just keywords). Box says this lets AI “agents” answer questions without scanning every file from start to finish. An AI agent is software that can take steps on your behalf, such as searching, summarizing, and pulling quotes.

Nvidia’s generative AI software VP, Kari Briski, told TechCrunch that the system should answer queries four times faster. She also said it should have 20% fewer errors and use 90% fewer “tokens,” which are the small chunks of text AI systems process (fewer tokens usually means lower cost).

Box CEO Aaron Levie said the system can return the most relevant parts of documents, instead of forcing an agent to read whole files. Box said it will be especially useful for work like legal research, where different wording can still refer to the same idea.

Box Content Retrieval is expected to roll out early next year for Box Enterprise Advance customers.

Why it matters

Many people are starting to use AI tools at work, but they are only helpful if they can quickly find the right information and respect who is allowed to see it. Box is betting that better “meaning-based” search, combined with Box’s existing permission controls, will make AI answers in company files faster, cheaper, and more reliable.

Source: TechCrunch AI