Slow search is a big problem.
We’re all making more content than ever before. At Box alone, customers add roughly 30PB of data every month, and all of that content is useful only if people can find it whenever they need it.
AI agents are even more dependent on fast, accurate search. As AI agents plan, research, summarize, and more, they often start with one or more searches to find relevant documents and information across large volumes of unstructured enterprise data to provide the context needed to return accurate, useful responses to prompts.
Enterprise search has lagged behind search engines like Google for decades, and the reason has always been structural: not enough signals, too much access restriction, too much reliance on keyword matching. AI doesn't just solve these problems; it fundamentally reimagines what search can be.
We’re excited to share that Search 3.0 is now generally available. Withthe latest AI-powered search experience in Box, it’s faster and easier than ever to find the right information.
Key takeaways:
- The new Box Search is a major, scalable infrastructure upgrade designed to deliver near-real-time indexing and lower search latency across all search surfaces
- Box Search is significantly faster, with search results that perform 4- 10x better, and file availability for search from 15 minutes down to seconds
- The new foundation enables external and native AI agents to perform deep-document searches via Box APIs, paving the way for upcoming features like multimodal content search, contextual retrieval, and hybrid search
More speed, relevance, and intelligence with Box Search
Our new Box Search is a foundational infrastructure upgrade to our search layer, a scalable search engine built for speed, relevance, and intelligence at scale. The result is dramatically faster, more relevant content discovery for both end users and AI agents, unlocking capabilities that will directly improve the user experience and position Box for a future where search goes far beyond keywords
Core updates include:
- Faster search with new infrastructure built for better scale and performance, delivering 4–10x lower latency
- Better relevance that improves result quality and gets users to the right content faster
- Stronger non-English support for our global customers
- Faster indexing with real-time improvements and significantly lower indexing delay
- Simple, powerful public APIs with the same latency and relevance improvements as end-user search
- A stronger foundation for AI Agents, Hubs Q&A, and future advanced search experiences
Notably, Box APIs expose context retrieval capabilities to external agents, and our own native AI agents call upon these same APIs to do the kind of intelligent, deep-document search necessary for the enterprise. Whether you’re a developer building on top of Box or an end user asking questions directly, the infrastructure to find real answers is already there.
Performance metrics for the new Box Search
The new Box Search performs 4-10x better compared to the previous search infrastructure. Where there was a 15-minute delay for files to be available to search after creation, now, it’s just seconds. This is true for both Quick Search and Full Search, and here’s a more detailed explanation.
Quick Search is a fast way to find something with a simple search. It usually looks at the most obvious matches first, like file names or a small set of likely results. It’s the right tool to use when you want a quick answer or already roughly know what you are looking for.
Here’s what Quick Search looked like before:
And here's what it looks like now:
What is Full Search?
Full Search is a deeper search. It looks through more places and more content, so it can find things that Quick Search might miss. It may take a little longer.
And here’s what it looks like now:
The number one change you’ll experience with the new Box Search is improved speed, as well as improved relevance.
Indexing time has dropped from 15 minutes to seconds
In some situations, it's helpful to be able to search for a document quickly. We have reduced the time from when a file is uploaded to when it can be found in search from 15 minutes to seconds.
This near-instantaneous indexing is especially critical for external and native AI agents. Because AI agents rely on deep-document searches to retrieve real-time context, having files available for search immediately after creation ensures that agent responses are always accurate, up-to-date, and fully informed by your latest enterprise data.
What’s next for Box Search
Enterprise search is finally starting to get really efficient and useful. And for organizations that run on information (which is all of them) that's not a minor upgrade. It's a transformation.
Up next? Look for:
- Hybrid search for meaning-based and natural-language retrieval
- Multi-modality search across images, video, audio, and scanned PDFs
- Richer search workflows powered by snippets, proximity search, and more
Box Search is not just about finding files faster. It is about unlocking the intelligence in our content for users, teams, and AI agents across Box.
Read more in this Roundtable summary and recording, or try out the new Box Search in your Box Drive. If you’re a developer interested in building on Box, don’t forget to sign-up for a free Box Developer account and join the Box Dev community.
