Box has shown off a new feature, Box Content Retrieval, that builds Nvidia’s Nemotron embedding system into its core product. Box says it will make queries faster and more accurate, while Nvidia’s Kari Briski expects fewer tokens to be used.
At its developer conference on Thursday, Box unveiled Box Content Retrieval, a feature that builds Nvidia’s Nemotron embedding system into its product. It is designed to help AI agents answer queries more effectively using embedded metadata.
Nvidia's vice president of generative AI software, Kari Briski, told TechCrunch that she expects queries to be answered four times faster, with 20% fewer errors and 90% fewer tokens used under the new system. Box Content Retrieval will be available to all of Box's Enterprise Advance customers, rolling out early next year.
Box CEO Aaron Levie said the integration dramatically boosts the relevance of search technology for agents. “Instead of an agent having to kind of read through each document, we can send back chunks from the documents that are most relevant to the user’s query or whatever the task is that the agent is doing,” he told TechCrunch.
The feature is an implementation of Nvidia’s Nemotron Embed project, which offers open-weight embedding models to improve automated retrieval systems. Nvidia’s most recent Nemotron release in July ranked first on the RTEB benchmark, which measures retrieval quality. Because Nemotron’s embeddings go beyond keywords to map underlying concepts, a search tool can surface related legal concepts even when different language is used.
Box gave the example of lawyers using the system to find relevant clauses and precedents more easily. While Nemotron models are freely available, the new feature relies heavily on Box’s pre-existing metadata and permission controls, making it difficult to replicate from scratch. “We’re probably packing in five different technologies into one simple API,” Levie said.
Analysis
Why This Matters
- Enterprise search becomes more efficient and context-aware, potentially reducing costs and improving accuracy for document-heavy industries such as legal, finance, and healthcare.
- Nvidia expands beyond hardware into enterprise AI software, embedding its models into widely used platforms.
- Box strengthens its AI capabilities, positioning itself against competitors like Google, Microsoft, and Dropbox that are also investing in AI-powered content retrieval.
Background
Embedding models convert text into numerical vectors that AI systems use to measure relevance and retrieve information. Nvidia's Nemotron is an open-weight model that has topped benchmarks for retrieval quality. Box, a cloud content management platform, has built its new feature on top of metadata and permission controls that are already in place for enterprise customers, making the system tailored for organisational data. The feature is intended to work with AI agents that act on the retrieved data, enabling more fluent operations.
Key Perspectives
Aaron Levie, Box CEO: The integration dramatically boosts the relevance of search technology for agents, allowing retrieval of document chunks most relevant to a user's query or task.
Kari Briski, Nvidia VP of generative AI software: Expects queries to be four times faster, with 20% fewer errors and 90% fewer tokens used, making the system more efficient for enterprise users.
What to Watch
- Rollout to Box's Enterprise Advance customers early next year and the pace of adoption in regulated industries like legal and healthcare.
- Competitive response from other content platforms that are also developing AI search and retrieval features.
- How developers and enterprises leverage the combination of Box's metadata and Nemotron's open-weight models for custom workflows.
Checked
12 claims in this story checked against 1 source: 9 supported.
Show each claim
- Partly supportedBox integrates Nvidia’s Nemotron embedding system into its product.AI News & Artificial Intelligence | TechCrunch: “Box showed off a new feature that will build Nvidia’s Nemotron embedding system into the core of its product.”
- SupportedBox Content Retrieval promises faster, cheaper and more accurate queries.AI News & Artificial Intelligence | TechCrunch: “Box says the upshot for users is faster, more accurate and less expensive queries.”
- SupportedThe feature is planned to roll out early next year to all Box Enterprise Advance customers.AI News & Artificial Intelligence | TechCrunch: “The system will be available to all of Box’s Enterprise Advance customers, rolling out early next year.”
- SupportedThe feature is designed to help agents answer queries more effectively using Nemotron-created embedded metadata.AI News & Artificial Intelligence | TechCrunch: “Called Box Content Retrieval, the new feature allows agents to answer queries far more effectively by relying on the embedded metadata created by Nemotron.”
- SupportedKari Briski expects queries to be four times faster, with 20% fewer errors and 90% fewer tokens used.AI News & Artificial Intelligence | TechCrunch: “Nvidia’s VP of generative AI software, Kari Briski, told TechCrunch that she expects queries to be answered four times faster, with 20% fewer errors and 90% fewer tokens used.”
- Partly supportedLevie said, “The integration dramatically boosts the relevance of search technology for agents.”AI News & Artificial Intelligence | TechCrunch: ““This dramatically boosts both the relevance of our search technology for agents,” Box CEO Aaron Levie said.”
- Partly supportedLevie said agents can return relevant document chunks instead of reading each document, ending the quote with “or whatever the task is that the agent is doing.”AI News & Artificial Intelligence | TechCrunch: ““Instead of an agent having to kind of read through each document, we can send back chunks from the documents that are most relevant to the user’s query or whatever the the task is that the agent is doing.””
- SupportedNemotron’s most recent release in July ranked first on the RTEB benchmark.AI News & Artificial Intelligence | TechCrunch: “Nvidia’s most recent Nemotron release in July ranked #1 on the RTEB benchmark, which measures retrieval quality.”
- SupportedNemotron embeddings can help a search tool surface related legal concepts even when different language is used.AI News & Artificial Intelligence | TechCrunch: “Because Nemotron’s embeddings go beyond simple keywords to map underlying concteps, a search tool can surface related legal concepts even when different language is used.”
- SupportedBox gave lawyers finding relevant clauses and precedents more easily as an example of the system’s use.AI News & Artificial Intelligence | TechCrunch: “Box gave the example of lawyers, who can find relevant clauses and precedents more easily under the new system.”
- SupportedNemotron models are freely available, but the feature relies heavily on Box’s existing metadata and permission controls, making it challenging to recreate from scratch.AI News & Artificial Intelligence | TechCrunch: “While the Nemotron models are freely available, the system also relies heavily on Box’s pre-existing metadata and permission controls, making it challenging to recreate from scratch.”
- SupportedLevie said, “We’re probably packing in five different technologies into one simple API.”AI News & Artificial Intelligence | TechCrunch: ““we’re probably packing in five different technologies into one simple API,” Levie told TechCrunch.”
