Box is bringing Nvidia’s Nemotron technology into a new agent-first content retrieval system, aiming to make it easier for companies to search, understand, and act on their internal content with AI.
According to Box, the system can answer queries four times faster, while producing 20% fewer errors and using 90% fewer tokens. If those gains hold up in real-world enterprise settings, they could translate into faster workflows, lower AI costs, and more reliable answers for teams working across large document repositories.
Why it matters
Enterprise content platforms hold massive amounts of business knowledge, but finding the right information can still be slow and fragmented. Agentic retrieval systems promise a more proactive experience: instead of simply returning files, AI can help interpret requests, locate relevant content, and generate useful responses.
- Faster answers: Reduced latency can make AI assistants more practical for everyday work.
- Better reliability: Fewer errors are critical when teams rely on AI for business information.
- Lower costs: Major token reductions can improve the economics of deploying AI at scale.
This is a strong example of AI moving from experimental demos into productivity infrastructure, where efficiency improvements can have an immediate impact on enterprise users.