Box Unveils Comprehensive AI Strategy at BoxWorks 2025, Revolutionizing Content Management

Box CEO Aaron Levie on AI's 'era of context' | TechCrunch

On Thursday, Box launched its developer conference Boxworks by announcing a new set of AI features, building agentic AI models into the backbone of the company's products. It's more product announcements than usual for the conference, reflecting the increasingly fast pace of AI development at the company: Box launched its AI studio last year, followed by a new set of data-extraction agents in February, and others for search and deep research in May. Now, the company is rolling out a new system called Box Automate that works as a kind of operating system for AI agents, breaking workflows into different segments that can be augmented with AI as necessary. I spoke with CEO Aaron Levie about the company's approach to AI, and the perilous work of competing with foundation model companies. Unsurprisingly, he was very bullish about the possibilities for AI agents in the modern workplace, but he was also clear-eyed about the limitations of current models and how to manage those limitations with existing technology. This interview has been edited for length and clarity. TechCrunch: You're announcing a bunch of AI products today, so I want to start by asking about the big-picture vision. Why build AI agents into a cloud content-management service? Aaron Levie: So the thing that we think about all day long - and what our focus is at Box - is how much work is changing due to AI. And the vast majority of the impact right now is on workflows involving unstructured data. We've already been able to automate anything that deals with structured data that goes into a database. If you think about CRM systems, ERP systems, HR systems, we've already had years of automation in that space. But where we've never had automation is anything that touches unstructured data. Think about any kind of legal review process, any kind of marketing asset management process, any kind of M&A deal review -- all of those workflows deal with lots of unstructured data. People have to review that data, make updates to it, make decisions and so on. We've never been able to bring much automation to those workflows. We've been able to sort of describe them in software, but computers just haven't been good enough at reading a document or looking at a marketing asset. So for us, AI agents mean that, for the first time ever, we can actually tap into all of this unstructured data. TC: What about the risks of deploying agents in a business context? Some of your customers must be nervous about deploying something like this on sensitive data. Levie: What we've been seeing from customers is they want to know that every single time they run that workflow, the agent is going to execute more or less the same way, at the same point in the workflow, and not have things kind of go off the rails. You don't want to have an agent make some compounding mistake where, after they do the first couple 100 submissions, they start to kind of run wild. It becomes really important to have the right demarcation points, where the agent starts and the other parts of the system end. For every workflow, there's this question of what needs to have deterministic guardrails, and what can be fully agentic and non-deterministic. What you can do with Box Automate is decide how much work you want each individual agent to do before it hands off to a different agent. So you might have a submission agent that's separate from the review agent, and so on. It's allowing you to basically deploy AI agents at scale in any kind of workflow or business process in the organization. TC: What kind of problems do you guard against by splitting up the workflow? Levie: We've already seen some of the limitations even in the most advanced fully agentic systems like Claude Code. At some point in the task, the model runs out of context-window room to continue making good decisions. There's no free lunch right now in AI. You can't just have a long-running agent with unlimited context window go after any task in your business. So you have to break up the workflow and use sub-agents. I think we're in the era of context within AI. What AI models and agents need is context, and the context that they need to work off is sitting inside your unstructured data. So our whole system is really designed to figure out what context you can give the AI agent to ensure that they perform as effectively as possible. TC: There is a bigger debate in the industry about the benefits of big, powerful frontier models compared to models that are smaller and more reliable. Does this put you on the side of the smaller models? Levie: I should probably clarify: Nothing about our system prevents the task from being arbitrarily long or complex. What we're trying to do is create the right guardrails so that you get to decide how agentic you want that task to be. We don't have a particular philosophy as to where people should be on that continuum. We're just trying to design a future-proof architecture. We've designed this in such a way where, as the models improve and as agentic capabilities improve, you will just get all of those benefits directly in our platform. TC: The other concern is data control. Because models are trained on so much data, there's a real fear that sensitive data will get regurgitated or misused. How does that factor in? Levie: It's where a lot of AI deployments go wrong. People think, "Hey, this is easy. I'll give an AI model access to all of my unstructured data, and it'll answer questions for people." And then it starts to give you answers on data that you don't have access to or you shouldn't have access to. You need a very powerful layer that handles access controls, data security, permissions, data governance, compliance, everything. So we're benefiting from the couple decades that we've spent building up a system that basically handles that exact problem: How do you ensure only the right person has access to each piece of data in the enterprise? So when an agent answers a question, you know deterministically that it can't draw on any data that that person shouldn't have access to. That is just something fundamentally built into our system. TC: Earlier this week, Anthropic released a new feature for directly uploading files to Claude.ai. It's a long way from the sort of file management that Box does, but you must be thinking about possible competition from the foundation model companies. How do you approach that strategically? Levie: So if you think about what enterprises need when they deploy AI at scale, they need security, permissions and control. They need the user interface, they need powerful APIs, they want their choice of AI models, because one day, one AI model powers some use case for them that is better than another, but then that might change, and they don't want to be locked into one particular platform. So what we've built is a system that lets you have effectively all of those capabilities. We're doing the storage, the security, the permissions, the vector embedding, and we connect to every leading AI model that's out there.

BoxWorks 2025: AI and Automation Take Center Stage

BoxWorks 2025 brought together Box customers and partners in San Francisco, September 11-12th with key announcements that underscore Box's commitment to AI and its power to transform unstructured data... i.e. content. Box's vision has been consistent for years: one source of truth with unified, secure content storage. Box is now layering an AI foundation into its core content platform. Box does not view AI as just an add-on to an existing content management system but envisions it as an integral part of a core offering, available to all clients. This foundation includes model flexibility, OCR, secure RAG, vector embeddings, markdown conversion, and support for multiple file types. Top Announcements: * Box Automate: Box Automate (expected beta in early 2026) is a new AI Agent-based workflow tool that will allow humans and agents to work together. Automate has an intuitive interface and provides a range of building blocks to build workflows, identify outcomes, use sophisticated conditional branching and orchestrate actions across both agentic and non-agentic workflows. It integrates with existing Box capabilities such as document generation and e-signature and can extend into third party applications via APIs. While Box has had its Box Relay routing/task management tool for several years, it has lacked a more robust workflow engine. Box Automate will help fill this feature gap. While Automate will co-exist with Relay for some time, expect that Box Automate will be the future path for workflow within Box. * Box Extract: Box Extract extends Box's intelligent document processing (IDP) capabilities allowing users to build and manage end to end data extraction processes. Extract is designed for the power users who would operate extraction workflows (ie for legal teams, finance, or operations), Extract brings advanced OCR (such as hand writing detection and table extraction), does math calculation, identifies and extracts metadata and taxonomy information, and provides confidence scoring. It also provides a document graph to help understand entities and parties in large complex documents. Box Extract is built by the team from AlphaMoon, the IDP provider acquired by Box in 2024. Extract Agents and APIs are available now, with the full Box Extract management console expected to beta in November 2025. * Enhancements to Box Apps: Box Apps (a no-code metadata and app design tool launched in early 2025) will be enhanced with natural language queries available on app metadata views and more data visualization. AI Agents will also now be available in Box Apps. The agents could be Box-provided agents (such as Q&A, Compose, Extract, Search and Research), or custom-built agents using Box Extract or Box AI Studio. Apps will also be embeddable in other applications, such as Salesforce. Expect these new capabilities to be generally available over the next few quarters. * Box Shield Pro: Box Shield Pro is a new add-on module for Box Shield customers that will bring enhanced threat analysis, ransomware detection, and AI classification. The threat analysis capabilities will provide security teams with summaries and analysis to provide more insights and help them focus on threats. The ransomware detection protects information beyond Box, helping to protect end users' endpoints by detecting file activities that could signal an attack. Anomaly detection from these end points can compel an admin to act and terminate a session, lock users out of a device, and know which files are affected. AI-based classifications can inspect content using context and content to define sensitivity, going well beyond rule-based approaches using keywords or text strings. This AI-based approach can look at nuances and take the overall meaning of a document into account, not relying on just rules and policies, It also understands context based on the author and who it is shared with. Classification labels are automatically applied, along with options for watermarking. This classification agent can also look at older managed documents, not just net new ones added to Box. Box Shield Pro is expected to be generally-available in 2025. What it means for Box customers: Access to approachable AI optimized for enterprise content. Box has always invested in intuitive user interfaces and makes usability and simplicity priorities. This extends to its AI evolution as well. Box customers have an opportunity to put a range of AI and AI Agent capabilities into the hands of their end users, and be confident that their information will remain secure, governed, and used appropriately. Box customers should look at their licensing tier to understand exactly what they'll get as these new capabilities roll out and determine if it makes sense to look at more comprehensive licensing to get the full swath of AI innovation, rather than trying to build it themselves.

Box debuts AI agents for almost every aspect of content management - SiliconANGLE

Box debuts AI agents for almost every aspect of content management Box Inc. is beefing up its already powerful artificial intelligence capabilities in an effort to automate even more enterprise work with AI agents. The cloud content management platform announced a host of new "agentic" AI features at its annual user conference BoxWorks 2025 today, including an AI agent-powered data extraction tool called Box Extract and a new workflow automation system called Box Automate. There's no official timeline, but the new tools will be made available to Enterprise Advanced customers in the coming months, alongside several enhancements to the company's no-code application builder Box Apps, and an all-new Box Shield Pro offering for securing sensitive data in Box. Box debuted its first AI agents in February, introducing a series of robotic workers powered by generative AI that can complete tasks such as querying documents and extracting data from files with minimal supervision. With today's offerings, it's building on those initial agentic capabilities to enable more comprehensive AI automation. For instance, Box Extract is all about unleashing AI agents to dig up critical insights that human workers might miss by carefully scanning everything from contracts and invoices to spreadsheets and images. As Box explains, this kind of data is unstructured, which makes it especially difficult to access. Until the rise of AI agents, the only way to unearth insights from these kinds of documents was to sift through them manually, but of course, that's completely impractical for an enterprise that has amassed thousands of them. With AI agents, it's no longer a problem, because enterprises are getting what is essentially a team of robotic workers on steroids. As the name suggests, Box Extract is all about data extraction, relying on the "reasoning" skills of its enhanced extract agents to not only understand the documents they're looking through, but to pick out the most vital bits of information within them. These agents, Box says, have particularly good context understanding skills, which make them uniquely able to interpret documents with tables, charts, handwriting, barcodes and other complicated inputs. They can understand the semantic relationships between different fields to dig up "nested and interrelated data points" and then filter this information to different applications and business processes using the Box Extract application programming interface. As has long been the case with Box's AI offerings, customers can choose virtually any large language model they desire to power these capabilities. Its catalog of LLMs include the latest and most powerful models from OpenAI, Google LLC, Anthropic PBC, Meta Platforms Inc., Amazon Web Services Inc., xAI Corp. and IBM Corp. Box Automate, meanwhile, is a new agentic workflow automation tool that aims to better orchestrate work across AI agents and human workers. The idea is that companies can hand off as much work as possible to AI agents, such as manual data entry tasks and accounting processes, so that humans can focus on the more complex and nuanced tasks that AI can't yet be relied on to do effectively. The company said Box Automate makes it simple for companies to design and manage automated AI workflows using a streamlined no-code or low-code interface. As part of this process, users can create and customize foundational AI agents for each specific task. So, for instance, the Box Extract agent can be fine-tuned on the specific documents a healthcare organization has to deal with each day, while a research agent can be trained to create reports based on a specific format. As users are creating each workflow, they can assign each of the different tasks involved to the most suitable entity, be it an AI agent, a human or a specific system, based on business logic and real-time context. Box co-founder and Chief Executive Aaron Levie said Box Automate can be used to create an incredibly wide variety of workflows that span client onboarding, sales automation, invoice processing, field operations and more. With it, users can specify exactly which tasks in those processes should be automated and which should be performed by humans. "As AI agents rapidly integrate into every vertical and domain, we're witnessing a fundamental shift in how work gets done," Levie explained. "This new generation of intelligent agents... [is] enabling organizations to automate previously impossible tasks and unlock massive new value." The Box Apps platform has proved to be one of Box's most popular AI tools, making it relatively easy for workers to create intelligent applications that can automate or streamline tasks such as contract management, digital asset libraries, employee onboarding experiences and more. Users will now be able to make even more capable AI-generated apps that can integrate with AI agents. For instance, Box talks about its new agent-assisted analysis capabilities that can help applications to identify new trends and anomalies and recommend actions to users based on what it finds. Other new features include enhanced natural language search tools to help users quickly explore the content stored within their AI applications, and new "dynamic data visualizations" including charts and graphs that can evolve in real time as more data is fed into them. Box has also worked hard on the integration side, so customers can now embed their Box Apps within third-party software platforms such as Salesforce and Workday. Finally, Box said it's expanding its AI-native content protection tool Box Shield, with a new "Pro" version that uses AI agents to enhance risk classification, accelerate threat response and proactively strengthen security. When Box launched the original Box Shield more than five years ago, the company didn't mention AI once. With this new update, virtually all of its main capabilities can now be automated with AI agents, so security is not only stronger but also simpler to implement. Box Shield Pro's AI security agents have been trained specifically to safeguard every sensitive document, and they can do this without human input. Every time a new file is uploaded to Box, an AI classification agent will quickly check it and then decide if it's sensitive or not, based on the file's context, the type of data within and the organization's existing security policies. If it's deemed to be sensitive, it will automatically be secured and governed with the appropriate access controls. As an example, Box said a healthcare organization can use its security classification agents to automatically detect documents that contain a patient's personal information and health records and restrict access to it. Meanwhile, the AI threat analysis agent works around the clock, under the hood, continuously checking which documents are being accessed, when and by whom. Should it discover any unauthorized access, it will create an alert together with a concise, easily understandable summary that explains the nature of the alert, so security teams can respond faster. There's also a third type of agent that's focused on ransomware activity detection, which oversees Box Drive and the rest of the Box content management platform. Its primary job is to detect mass encryption of content, which is a sure sign of ransomware, and then take rapid action to remediate such attacks. Moor Insights & Strategy analyst Melody Brue said Box Shield Pro can be useful, because the rapid adoption of AI agents creates new risks for organizations. "Traditional security protocols were not designed to manage these agents at a machine's processing speed," she said. "Box Shield Pro's approach of integrating security controls directly into AI agent operations provides a method for maintaining cybersecurity when human oversight isn't practical."