Box announces major AI platform enhancements, bringing powerful AI Agents to enterprise content

At Box, we’ve always believed that content is the heart of every business. It’s where strategies are crafted, deals are closed, and innovations are born. But much of this content remains underutilized—buried in folders, locked in PDFs, or scattered across tools and teams. The challenge has been unlocking it.

Today marks the next evolution of Box AI—one that leverages a new class of intelligent AI Agents designed to work with your content, understand your business, and accelerate real work. These task-driven AI Agents are natively built into Box AI, designed to search smarter, conduct deeper research, extract information faster, and operate with your organization’s specific context, all automatically governed by Box’s industry-leading security, compliance, and permissions model.

Box AI’s dynamic agentic reasoning framework 

Box AI now operates with a dynamic agentic reasoning framework that adapts based on the task at hand. This flexible approach allows Box AI to intelligently engage with your content, processing it in context to drive meaningful outcomes. 

Here’s how it works:

1. Plan: Box AI first identifies the goals or tasks at hand. It understands the objective, whether it’s finding specific information, gathering insights, or performing data analysis, and tailors its approach to achieve that goal efficiently.
2. Discover: Once the task is identified, Box AI begins the discovery phase. It processes and accesses your content, from documents and presentations to feedback and contracts, ensuring it pulls the most relevant data from the right sources.
3. Reason: At this stage, Box AI applies reasoning by analyzing the content it has discovered. This involves drawing conclusions, identifying patterns, and synthesizing insights to generate a solution or answer that is accurate and context-aware.
4. Act: Finally, Box AI takes action based on the insights it has gathered. Whether that’s delivering answers, recommending actions, or triggering processes, Box AI ensures that every step is aligned with your business needs and your content's value.

The Box AI agentic reasoning framework provides a structured and intelligent approach to automating complex tasks. This framework empowers Box AI Agents to first understand the objective (Plan), then securely locate and retrieve relevant enterprise content within Box (Discover). Next, they intelligently analyze and synthesize this information (Reason), culminating in a tangible output or action (Act), such as generating a document or providing a risk assessment — all while adhering to your organization's specific controls and leveraging your single source of truth.

The anatomy of a Box AI Agent

Box AI Agents are goal-driven, intelligent entities that operate directly over your enterprise content—securely, contextually, and intelligently. Designed for real work, Box AI Agents rely on structured logic, model choice, and real-time access control to reason through tasks and generate high-quality outputs. 

• Model: Box AI Agents leverage best-in-class models from Open AI, Anthropic, Google, and others to best match a model’s capabilities around reasoning, summarization, classification, and generation with the role that the agent is designed to accomplish
• Objective: Every agent is built around a specific goal, such as summarizing a quarterly business review, identifying risk in contracts, or generating marketing content.
• Instructions: These are the structured prompts, logic flows, or task plans that guide how the agent accomplishes its goal—much like an internal playbook the agent follows step-by-step.
• Capabilities & Tools: Box AI Agents bring specialized skills to the table, such as Q&A, summarization, classification, and the ability to call Box APIs to get the job done.
• Knowledge: The foundation of a Box AI Agent is the agent’s source material—the enterprise content it operates over, securely pulled from Box using permission-aware access, metadata, and real-time governance policies.

Box AI Agents leverage secure RAG and are governed by your enterprise's content controls. These agents securely access and intelligently process your specific Box files to automate complex tasks. This ensures actions, from content generation to risk analysis, are consistently grounded in your organization's single source of truth. Box AI Agents offer a powerful way to drive efficiency and leverage your enterprise content with precision and security

Box AI Agents will be able to do Search, Deep Research, and Enhanced Data Extraction

Box AI Agents are poised to revolutionize how enterprises interact with their vast (and rarely ideally organized) content. Today's reality of scattered files, ambiguous version control, and siloed data makes finding what you need, when you need it, a major impediment to productivity. Box AI Agents will cut through this complexity by offering intelligent ways to search, conduct deep research, and extract valuable data from your existing content repositories.

Smarter Search that gets the answer, not just the file

Box AI Agents will allow you to search smarter, and deliver precision, in two modes. For straightforward needs, the "quick lookup" feature will instantly surface contextual answers and specific data points from your documents, such as finding an expiration date in a contract. For more intricate queries, such as understanding all risks associated with a particular account, the "Deep Search" functionality will meticulously analyze hundreds of files, scanning them for relevant insights and delivering comprehensive results, all while allowing you to continue with other tasks. This means search adapts to the complexity of your question, providing precise information whether it's a sales VP needing the latest pricing deck or a legal analyst identifying specific contract clauses.

Deep Research that turns folders into executive insight

You may have used Deep Research in your favorite AI chatbot, but what if you could leverage that same experience across all of the content that lives in your organization. Introducing Deep Research from Box. Unlike general AI research tools that rely on public web searches, Box Deep Research securely analyzes your organization’s internal content, governed by permission-aware access and secure Retrieval-Augmented Generation (RAG). This ensures insights are accurate, context-specific, and compliant with your enterprise’s security policies. Box AI Agents will enable Deep Research, turning entire folders of unstructured data into actionable insights. When you need to analyze large volumes of content to uncover trends or synthesize critical information, Box AI Agents will take on the heavy lifting, including finding what files it should be looking through (compliant with Box’s permission-based, secure RAG, of course). For example, a customer success lead could ask about the key drivers of churn, and Box AI Agents will leverage Deep Research to analyze QBRs, support tickets, and feedback forms, surfacing crucial trends. Similarly, an HR executive could query shifts in employee sentiment, and the agents would comb through surveys and internal communications to provide a comprehensive analysis, empowering faster, data-driven decisions.

Data extraction that turns documents into structured insight

Box AI Agents also tackle the challenge of Enhanced Data Extraction, transforming messy, unstructured data like scanned PDFs, images, and handwritten forms into structured, usable insights. By employing Optical Character Recognition (OCR), advanced image and document understanding, and natural language processing, Box AI Agents will automatically identify and extract high-value data such as key dates, financial totals, contractual obligations, and clause types. This means a finance team could upload vendor invoices and have details like vendor names and due dates automatically extracted and flagged for inconsistencies, or a compliance officer could audit hundreds of contracts for specific risk clauses, turning previously unsearchable text into actionable, structured data without manual intervention.

Ground your Agent in real business knowledge from Box documents and Hub

Every company has its own way of working. Box AI Studio will soon let you attach knowledge to your custom Agents, whether it's a document or a Box Hub.

Your Agents will be able to :

• Flag risky language based on your real compliance policies
• Summarize business performance with QBR context in mind
• Draft content aligned to your brand’s tone, guidelines, and goals

Your custom Agents become a secure extension of your business knowledge, not just a chatbot, delivering responses based on governed, permissioned content from Box.

Quickly troubleshoot and optimize AI Agent performance with Debug Mode

Debug mode enables Admins to quickly restore previous settings, including the selected LLM and system prompts. It provides clear visibility into AI agent reasoning paths, helping Admins easily identify bottlenecks, troubleshoot issues, and refine responses for improved accuracy and performance.

Guide users effectively with custom suggested questions

Admins can configure custom suggested questions for each customized AI Agent, ensuring end-users immediately see relevant prompts when engaging with the agent. These tailored suggestions help guide user interactions, streamline workflows, and encourage adoption by clearly demonstrating the agent’s capabilities and intended use cases. By proactively surfacing helpful starting points, organizations empower their teams to quickly leverage AI Agents for faster, more productive outcomes.

This is how AI becomes useful:context-aware, content-secure, and business-specific.

Coming soon: Box AI Agent for Microsoft 365 Copilot

Box is committed to ensuring its AI Agents seamlessly integrate across the applications where your team gets work done. We've already launched Box AI Agents in platforms like IBM watsonX Orchestrate, Salesforce’s Agentforce, Google’s Agentspace, and ServiceNow’s AI Agent Fabric. Today, we're proud to announce a new Box AI Agent for Microsoft 365 Copilot. This allows users to leverage Box AI within Microsoft Teams, Copilot Chat, Word, and PowerPoint, bridging the gap for those whose daily work is in Microsoft 365 but whose critical content resides securely in Box. This integration eliminates disruptive app-switching and empowers users, like a lawyer reviewing contracts, to query renewal dates, analyze documents for specific risks, and cross-reference findings against internal company guidebooks stored in Box—all from within Copilot. This transforms lengthy manual processes into minutes, making your secure Box content smarter and more accessible right where you work.

Now available: Box AI remote MCP server

Beyond direct application integrations, Box is also deeply invested in the rapidly evolving landscape of data and developer frameworks to ensure compatibility within our Agent ecosystem. We facilitate connections with numerous established frameworks such as the OpenAI Agents SDK, Microsoft Copilot Studio, and Google Cloud's Vertex AI Agent builder, as well as popular developer tools like Pinecone, LangChain, and LlamaIndex, underscoring our dedication to market neutrality and enabling your organization to build in the way that best suits your needs. A significant recent development in this area is our integration with the Model Context Protocol (MCP). This allows third-party AI agents from platforms including Copilot Studio, Cursor, Anthropic's Claude, and OpenAI to access Box content seamlessly. Following the March 2025 release of a locally hosted MCP server, Box will soon offer a company-hosted server for remote access, further streamlining and standardizing how various agents can connect to and utilize your enterprise knowledge.

The future of enterprise content is intelligent

With powerful new AI Agents embedded directly in Box, your content becomes more than just stored, it becomes active, contextual, and actionable.

From Extraction to Action: Box AI Extract Agents and Box Apps

Unlocking the valuable data within unstructured data, like contracts, is a common challenge, often involving immense manual effort. Box AI Extract Agents are designed to automate this, pulling key details from documents and images—even scanned paper files—and saving them as structured metadata in Box. This metadata makes content significantly easier to find, search, and manage. For instance, when integrating thousands of inherited customer contracts after an acquisition, instead of a manual review, AI Extract Agents can automatically populate metadata fields. This extracted information can then be leveraged at scale using Box APIs, turning vast amounts of previously inaccessible data into actionable insights for critical business decisions, like contract renewals or renegotiations.

Streamlined Compliance and Content Lifecycle Management with Box AI

Box is also enhancing its platform to streamline compliance and provide true end-to-end content lifecycle management, all built on a secure and scalable cloud infrastructure. New capabilities like Box Archive offer long-term content preservation, allowing retention policies and legal holds, including activity-based preservation. The modernized Content Manager provides admins with faster, more granular control to find and manage content across the enterprise, supported by significant scalability enhancements such as uploads of 500 files per second. Using Box Shuttle for high-speed content ingestion from external repositories ensures that permissions and metadata are maintained. Box AI agents and Box Shield can then automatically extract data, classify content based on sensitivity, and apply sophisticated access controls and governance—saving organizations significant time while ensuring compliance throughout the content lifecycle, from creation and migration to archival and eventual disposition.

Transform your enterprise content by unlocking smarter decisions and accelerate your business with Box AI Agents

With powerful new AI Agents embedded directly in Box AI, Box’s Intelligent Content Management suite transforms your content into an active, contextual, and actionable asset for your entire organization. From smarter search and deeper research to precise data extraction, streamlined compliance, and fully customized AI workflows, Box AI is engineered to help your teams move faster and make better decisions. This isn't just a glimpse into the future of work; it's happening right now— welcoming you to a new era where your content truly works for you.

Join us for the Content + AI Virtual Summit to see it all in action.