Box Unveils Box Context Retrieval Powered By Nvidia Nemotron Models

Introducing Box Context Retrieval, a new engine for enterprise AI

Agents are increasingly embedded across everyday workflows, but their value depends on access to the unique context that defines how a business operates. This context comes from the files, such as contracts, policies, customer history, and institutional knowledge, that live within enterprise content. To equip agents with the accurate information they need to work effectively, Box is introducing Box Context Retrieval, an advanced engine powered by NVIDIA Nemotron Retriever, a collection of open models that transforms enterprise data into a structured knowledge layer that agents can reason over.

With NVIDIA Nemotron Retriever embedding models, the Box Context Retrieval engine will enable agents to more accurately and efficiently retrieve relevant, permission-aware information across millions of files, including content that has historically been difficult for AI to interpret, such as scanned documents, images, audio, and video. Customers will have the potential to improve search accuracy by 20 percent (Recall@5), reduce token usage by 90 percent, and accelerate question-answering speeds fourfold, helping them access trusted enterprise knowledge and increase productivity.

With these new capabilities, customers will:

Receive answers they can trust through permission-aware retrieval, which provides AI agents with relevant information from only the content users are permitted to access, enabling precise responses grounded in their unique enterprise knowledge;
Access relevant information more efficiently through significantly faster indexing and lower latency, making new and updated Box content available to AI agents in real time;
Find the right information across more of their enterprise content using multimodal indexing to make scanned documents, images, audio and video searchable, and hybrid search to combine precise keyword matching with an understanding of meaning and context – even without knowing the exact words to search.
Bringing NVIDIA’s open weight models to Box

The collaboration will also bring NVIDIA’s frontier-scale, open weight Nemotron models, including NVIDIA Nemotron 3 Ultra, to Box. Box’s multi-model approach to enterprise AI gives customers the flexibility to select the right model for each task based on performance, cost, and control. With access to NVIDIA’s open models, Box customers can now access frontier-scale intelligence with more cost-efficiency for specialized, high-volume workloads – all while maintaining transparency and keeping their content securely managed in Box.

By leveraging the capabilities of NVIDIA’s open weight and embedding models in Box:

Financial analysts can ask questions about loan applications and receive answers grounded only in the content they are permitted to access, helping protect sensitive customer information;
Legal professionals can find relevant clauses and precedents across contracts and matter files, even when different terminology is used, by combining precise keyword search with an understanding of broader legal concepts;
Clinical researchers can access newly added or updated policies, care protocols and research in real time, helping them work with the most current information available in Box;
Insurance auditors can analyze complete claim packages across scanned forms, property images, recorded statements and video evidence, making all unstructured information available to AI agents;
University administrators can use NVIDIA’s open-weight models within Box to ask questions, extract information, and complete multi-step research workflows across university policies, research materials, student services documents, and other institutional content.
Securing agent execution

As members of the Open Secure AI Alliance, Box and NVIDIA are collaborating on AI safety to help enterprises deploy agents with greater control over how they access content, use tools, and execute tasks. The collaboration brings together Box’s enterprise content security and governance capabilities through Box Mount, and NVIDIA OpenShell’s secure runtime, policy enforcement, and supervision to advance safer agent workflows in the Box platform.

Availability

The Box Context Retrieval engine will be generally available to Enterprise Advanced customers early next year. NVIDIA's Nemotron models, beginning with 3 Ultra, are now available in Box AI to Enterprise Advanced customers. For more information, visit the Box blog.