Box Inc launches AI agents that automate, extract, and analyse enterprise data
A Silicon Valley cloud content management provider introduces agentic AI tools to streamline workflows, uncover insights, and strengthen enterprise security
Box, Inc. is transforming its cloud content platform into an intelligent workspace built for the age of artificial intelligence.
The company’s next-generation products—Box Extract, Box Automate, and enhanced Box Apps—will soon enable enterprises to deploy AI agents that can automatically extract key data from documents, analyze content in real-time, and orchestrate complex workflows across systems.
These AI agents are designed to simplify how businesses interact with their information, turning everyday content into actionable intelligence while maintaining Box’s hallmark focus on security and compliance.
“These tools are designed to solve the problems enterprises face daily—finding, analysing, and creating content securely and efficiently,” Yashodha Bhavnani, Head of AI at Box, told TechJournal.uk in an interview. “We’re re‑engineering our platform so that content, security, and AI coexist at the core. The future of Box is no more content‑first; it’s content security with AI first.”
Bhavnani said Box is embracing AI as “the larger shift in how people work—with their knowledge, with their content,” adding: “The future of Box is no more content first; it’s content security with AI first.”
The strategy marks a decisive shift for the California-based company, as it brings AI-driven intelligence to where data already resides, eliminating the risks and inefficiencies associated with transferring sensitive information between tools. The new AI agents help users locate files, summarise reports, and generate new content using natural language prompts—all within the trusted boundaries of the Box ecosystem.
From storage to intelligence
Box has long been recognised for enabling secure file sharing and collaboration across large organisations. But the latest AI-first vision aims to move beyond storage and transform Box into an intelligent content management hub capable of understanding and acting on data at scale.
At its BoxWorks 2025 conference in San Francisco, the company unveiled a suite of agentic tools that redefine the boundaries of automation:
Box Extract: Uses agentic reasoning to interpret and process large volumes of unstructured content—from PDFs and spreadsheets to scanned images and handwritten notes. It identifies relationships between fields, extracts structured information, and validates data automatically to deliver insights in seconds.
Box Automate: Provides a no-code builder for orchestrating workflows across AI agents, employees, and external systems. It enables end-to-end automation for processes such as HR onboarding, contract review, and invoice reconciliation.
Enhanced Box Apps: Expands Box’s no-code dashboard capabilities, allowing users to visualise trends, create dynamic charts, and embed AI-driven insights directly into business applications such as Salesforce.
Together, these tools create a unified environment where AI works directly with enterprise content, accelerating productivity and reducing operational complexity.
Bhavnani shared a practical example of this in action: a U.S. real estate client that once relied on manual audits of thousands of property leases across different states. Using Box Extract, the process was automated—documents were scanned, key fields identified, and compliance reports generated instantly.
“This is where the rubber hits the road of AI,” she said. “You can have the coolest technology, but if you can’t apply it to real‑world problems, it’s just fun.”
Standing out in a crowded AI field
The market for AI-driven cloud and SaaS platforms has become increasingly competitive, with companies such as Google, Microsoft, and Salesforce all launching their own AI assistants and workflow agents. Bhavnani acknowledged that Box operates in a crowded space but argued that the firm’s differentiation lies in its simplicity and trustworthiness.
“Silicon Valley has a new trend almost every year,” she said. “We take advantage of the right trends to help our customers with this focus on content.”
She noted that many organisations already use a patchwork of tools for CRM, HR, and project management, but Box’s approach is to provide an AI layer that connects them.
“The whole point is the connectedness,” she said. “Now you can do it in one trusted platform where your content lives.”
Asked why companies should choose Box over other enterprise AI offerings, Bhavnani highlighted three key themes: connectedness, control, and simplicity. She described customer feedback this way: “It’s so simple.”
She also stressed Box’s emphasis on “security and control,” alongside the value of having an “integrated platform” rather than multiple vendors.
AI agents built on choice and control
At the heart of Box’s innovation lies a strong commitment to user control. Through Box AI Studio, customers can build or customise AI agents using models from OpenAI, Google, Anthropic, IBM, Amazon, and others—and, in Bhavnani’s words, Box is “one of the first, if not the first, platforms where you can choose over 100 models.”
“While choice is important, it can be overwhelming. We provide a default that includes the best models in the market for the best use cases,” she said.
She added: “We may use one model for question answering on text, a different model for question answering on video, and a different model for data extraction.”
She further underscored Box’s security stance: “You can use AI on your most critical content. We’ll make sure that your data is never leaked back into the model.”
This philosophy of control extends to Box’s integration capabilities. Using APIs and developer tools, businesses can link Box with Salesforce, Microsoft 365, or ServiceNow to create a unified AI ecosystem. Box also supports emerging frameworks such as Google’s Agent-to-Agent (A2A) and Anthropic’s Model Context Protocol (MCP), which allow multiple AI systems to communicate seamlessly.
To ensure its technology remains at the cutting edge, Box maintains close collaborations with the world’s largest AI model providers. These partnerships enable Box customers to typically access new AI models on the same day they are launched. The company also collaborates with service integrators, such as Slalom, to help enterprises tailor Box’s capabilities to their specific operational needs.
In September 2025, Box said in a press release that IBM’s data division now utilizes Box Extract with Watsonx to convert vast troves of unstructured data into trusted, actionable intelligence. At the same time, Sage Hospitality Group has leveraged technology to process operational records more efficiently and with greater accuracy.
Enterprise roots and global reach
Founded in 2005 by Aaron Levie and Dylan Smith, Box started as a consumer file-sharing platform before pivoting to serve enterprise users around 2009.
Today, it is a publicly traded company with about 2,800 employees across the United States, the United Kingdom, EMEA, and Japan. From its headquarters in Redwood City, California, Box serves clients that include many Fortune 500 companies, which have entrusted their most sensitive intellectual property and confidential information to the platform.
Despite its scale, Bhavnani described Box’s expansion model as intentionally measured.
“I don’t think we think of expansion as more people. I think we think of expansion as serving our customers better,” she said. “Every region has unique compliance and data requirements, and our job is to meet those needs.”
The company’s long-term vision is to create an intelligent, interconnected ecosystem where AI agents collaborate across departments and systems. Using the new Box Automate and Extract frameworks, these agents can act as digital colleagues—interpreting data, launching workflows, and delivering insights instantly.
Bhavnani said the near-term focus is on interoperability across systems—citing launch partnerships for Google’s A2A framework and support for the MCP—so that agents can work with content without needing to move data out of Box.
As Bhavnani framed it, the near-term test is pragmatic: agents that stay close to the content, respect enterprise controls, and quietly make everyday work simpler without shuttling data between tools.