Hundreds of enterprise agents in CEO Aaron Levie's future vision as Box turns in solid Q2

AI-powered revenue growth helped Box to turn in a solid Q2, despite a drop in profits. For its Q2, the firm reported revenue of $294 million, up nine percent year-on-year, but net income took a dip from last year’s comparable $20.5 million to $13.4 million. The results were good enough for the company to up its full year revenue guidance to between $1.17 billion and $1.18 billion.

It was to customer adoption of Box’s AI offerings that CEO Aaron Levie pointed on the post-results analysts call:

A prominent US law firm became a new customer to Box, driven by Enterprise Advanced AI-powered metadata extraction capabilities and intelligent no-code apps to power its business processes. This is an enterprise-wide agreement replacing both an existing cloud-based platform vendor and an e-signature company.

In partnership with a systems integrator, a Fortune 500 hospitality chain upgraded from a non-suite plan into Enterprise Advanced as they move away from a manual process with multiple systems to manage global projects. The company is looking to use AI-powered metadata extraction, Box hubs, Doc Gen and Relay in design and planning workflows to scale projects and streamline execution.

And a global industrial automation company upgraded from Enterprise Plus to Enterprise Advanced and expanded seats as they look to centralize their contract management solutions, automate quote generations and enhance cross-entity document searchability. The company will use AI-powered metadata extraction to capture contract renewal dates and legal obligations to inform decision- making and ensure compliance. 

And there will be more to come with the rise of agentic tech, he predicted:

Enterprises know that AI agents are going to bring a new level of automation and deliver deeper business insights to their businesses. Software has historically been good for automating processes that deal with structured data, take payroll, CRM systems, accounting, HRIS or supply chain workflows. This is where data fits neatly into rows and columns in a database. But the vast majority of enterprise workflows revolve around unstructured data, which actually represents about 90% of our corporate information.

These are the workflows that drive client onboarding at a bank, M&A deals that get closed, contracts getting agreed on, clinical research advances, movies getting made and so much more. We've never been able to bring automation to these areas of work because they've been human-based manual processes dealing primarily with unstructured data.

Now for the first time ever, we can bring automation to this work with AI agents. With AI agents operating on unstructured data, enterprises can now accelerate product development processes, automate end-to-end hiring and training workflows, surface insights and automate clinical studies and speed up loan applications for better client engagement.

The vision thing

Levie is thinking big here:

We can imagine a future where there are over 100x more agents than people inside of an organization, where any task you want done in a company is only a matter of how much compute you want to throw at the problem. You'll have agents running in the background and in parallel for any workflow around content that you can imagine.

But this vision comes with some serious caveats:

Most companies can't tap into the full power of AI agents on their unstructured data because their enterprise content is fragmented or stuck in legacy repositories. And with this fragmentation, it means that AI agents have no core source of truth from which to answer questions about critical topics. It also means there's a risk that access controls are unmaintained, which can lead to AI agents leaking data to the wrong users asking a question. And finally, it becomes a massive nightmare integrating systems that don't play nice with one another in the AI era.

And there’s more:

We kind of imagine a future where you might have dozens, if not on the upper end of a large enterprise, hundreds of different AI systems that people are going to be working from. We obviously want to be the absolute best place to have you work with agents and unstructured data and content, but there's going to be just a tremendous number of other AI systems. You might have some users in ChatGPT. You might have some users in Claude. You might have some users in copilot. Some users might be in IBM, watsonx Orchestrate. And so there's a very real chance of, again, dozens or hundreds of these systems inside of organizations.

So then you, as an enterprise, have a decision. Do you replicate your data, your unstructured data across all of those systems, which is not only an incredibly costly problem, but it's also one that would lead to security risks and you have outdated information across those technologies. Or do you have a central repository that has your most important information and unstructured data that people can tap into from across all of those other environments?

Hence the need for Box’s AI tech, he said:

Importantly, Box AI agents work directly on top of the workflows that customers have already built on Box, and we're only accelerating what these combined capabilities can deliver going forward. We're already seeing the power of AI agents with customers building Box AI agents that can review and summarize documents, answer questions from a large data set and extract critical details from enterprise documents like contracts or invoices to orchestrate processes in legal, finance, healthcare and more.

In terms of future product enhancements, Levie has a to-do list:

We will introduce all new AI features within Box Notes, continued improvements for leveraging Box Hubs as an intelligent knowledge portal and all new core Box AI experiences to make it easy for customers to interact with AI agents and find information across their Box accounts no matter what they're looking for. And all of these AI agent capabilities will be available via our API so customers can take full advantage of summarizing, analyzing and extracting data from their content in any partner application, like Salesforce Agentforce, ServiceNow Agent Fabric, Google's agentspace, ChatGPT, Claude, Copilot, IBM's watsonx Orchestrate and more. And with our newly GA remote MCP server, customers can interact with the full Box API and AI agents as tools within their own AI-oriented applications

As for Box eating its own AI dogfood, there are multiple examples in action across the company, with Levie citing:

We're equipping every Box-er with the skills and tools to be productive with AI, encouraging experimentation, scaling best practices across the company and adding AI-first expectations in our hiring process. Across all of Box, we are using Box AI agents to augment our work in every area of the business from how we train and enable new sales or support reps to how we write product requirements or generate rapid account research with industry insights for each customer we sell to. AI agents are being used all across Box to help accelerate our workflows and drive increased productivity.

My take

More to come in a few weeks as the Box faithful gather for the annual collaboration at Boxworks in San Francisco. We’ll be keeping an eye on the outputs from that.