Box

Most teams are still manually sorting files that could route themselves. A question from our community said it best: "Can we automatically tag metadata and sort files into the right client folder without a third-party tool?" The answer is yes, using a simple Extract → Branch → Act pattern that most modern workflow platforms already support. What's a process your team is still doing by hand that you suspect could be automated?

An exciting example of how we're using Box AI to cut alert triage from minutes to seconds and reduce the fatigue that comes with reviewing hundreds of alerts a day. The time saved is going back into more automation and proactive threat hunting, while analysts stay in the loop for the calls that need human judgment.

OpenAI launched GPT-6.1 Sol yesterday, and we ran it against a real lending scenario in Box. A borrower submits a $1.25M funding package marked complete. But in this Box Agent demo, GPT-6.1 Sol reviews all 11 documents and finds what a checklist missed: two unmet conditions, apparent discrepancies in the delivery address and seller identity, and a source-cited report showing what the lending team actually needs before funding can proceed. Explore custom agents for financial workflows with Box AI Studio.

Box is now available in Muse, from Meta! Your personal AI agent can now find information buried in documents, manage contracts, and work with clients and partners, all with governed enterprise content from Box, with permissions and governance intact throughout. Small businesses don't need more AI tools. They need one secure place where people, content, AI, and workflows work together. That's what Box brings to Muse. Whether you are reviewing a contract, tracking down a client file, or pulling together materials for a customer meeting, Box gives Muse the business context it needs to get the job done. Connect Box in Muse today.

We evaluated Claude Sonnet 5.5 on Box's Complex Work Eval. Real document tasks, end to end, graded against a detailed rubric with confidently-wrong answers penalized. Sonnet 5.5 scored 65% overall against Sonnet 5's 61%, reached a finished deliverable 2.4x faster, and used 12% fewer total tokens doing it. Accuracy and speed usually trade off against each other in agentic work; improving both is a real step forward. What stood out across industries: → Financial Services (63% to 81%): in a due diligence review of a trading-tool acquisition, Sonnet 5.5 recalculated misreported interest totals and option prices in the deal book, identifying errors that Sonnet 5 repeated as fact. → Life Sciences (54% to 62%): calculated edge counts for protein interaction networks that Sonnet 5 missed, giving researchers a stronger starting point for examining potential targets. → Legal (60% to 67%): more consistently surfaced missing protections in a commercial lease review and proposed concrete language for counsel to assess during negotiation. Sonnet 5.5 is available today in Box AI Studio and through the Box API.

BoxWorks is almost here! And it's filled with hands-on Master Classes designed to help you go from experimenting with AI to actually leading it across your organization. The Box AI Strategist Certification is a half-day, hands-on experience designed to give you the practical skills and frameworks to turn AI experimentation into real business impact. Built for admins, power users, workflow owners, and team leaders who are ready to move beyond using AI and start leading AI transformation. What you’ll learn: 🗺️ Build an AI strategy. Assess your organization's AI readiness, identify high-value opportunities, and create an adoption roadmap. 🔧 Design AI solutions for your teams. Go beyond prompting to build custom agents and MCP-connected workflows that can scale. 🚀 Turn pilots into programs. Develop a practical approach for rollout, change management, user enablement, and success metrics. 🔒 Lead responsible AI adoption. Apply governance frameworks and guardrails while continuously improving your AI program. 🏆 Become an AI leader. Build the skills, language, and confidence to champion your AI strategy within your organization. BoxWorks 2026. November 5-6 in San Francisco. Save your spot today! Comment below for a discount code. 👇

Muse Spark 1.3 and Claude Opus 5.5 landed in Box this week. And Dreamforce changed the conversation around the headless enterprise. This week's newsletter goes deeper on what it all means for enterprise AI. What's inside: 👉 What Muse Spark 1.3 gets right about enterprise document work 👉 Why Claude Opus 5.5's efficiency story matters 👉 Why connecting process and content in the same place changes how governed content travels through a workflow 👉 What Dreamforce revealed about the headless enterprise Check out the full newsletter.👇

Instead of asking "How do we hire more people to keep up with this?" the Box Policies and Standards team asked a different question: "How do we make our existing content answer these questions itself?" Every quarter, the team had to manually open nearly 100 policy PDFs, scroll to signature pages, and extract key details like ownership, last review date, and approver names. Preparing a single quarter's batch took about 12 hours. Across four quarters and dozens of ad hoc requests, the team was spending roughly 100 hours a year chasing down information buried in PDFs. That changed when the team built a solution using Box AI, Box Apps, and a metadata template. Annual review prep dropped by 96%, from 12 hours to 2, saving roughly $14,000 in year one. Here's how: • A metadata template that automatically extracts policy ID, owner, approvers, last signature date, and review history directly from each PDF • A custom Box Apps dashboard that organizes all 100 policies by quarter and org, updated in real time with no manual sorting required • A self-service Box Hub where any employee can search, read, and ask AI-powered questions about policies, logging more than 6,500 views since launch This is a pattern any compliance or operations team can replicate. And with a Box Automate workflow now in pilot, even the last manual step, triggering reviews and assigning ownership, will soon take care of itself.

Extracting reliable, usable data from large volumes of unstructured content shouldn’t require endless manual review. The new Box Extract agent builder makes it easier to test and refine extraction for more accurate results at scale. See how it works for a clinical site monitoring workflow. 👇

Agents don't fail because they lack capability. They fail because they don't understand the business context. At Dreamforce last week, Box and Salesforce showed what changes when they do. Salesforce brings the structured business data. Box brings the governed content behind it, the contracts, proposals, case files, and documents that give agents the full picture. And because both are headless, that context goes wherever your teams and agents work. What headless looks like across the Salesforce ecosystem: ➡️ Box and Claudeforce. Claude can now reason across Salesforce records and governed Box content simultaneously, answering questions and completing tasks without files ever leaving Box. ➡️ Box in Slack. A question asked in Slack can reach across Box content, combine it with Salesforce data, and return an answer grounded in the full picture of the business. ➡️ Box for Agentforce. Salesforce agents can review documents, summarize account materials, and route content through business processes, all grounded in governed Box content. Files stay in Box, and records stay in Salesforce. Your people and agents can work anywhere.