Conference Transcript - BOX 2026-06-03

Operator

All right.

Ben Cuss, CTO

Well, welcome everybody and good morning.

Operator

Thanks so much for coming today. We are very fortunate today to have Dylan Smith, co-founder and CFO of Box, as well as Ben Cuss, the CTO of Box, to discuss the Box story and position in the agentic tech stack today. So maybe just to kick it off, there was a pretty interesting demo from OpenAI yesterday where Box was featured prominently. And it seems to me like Box is mentioned in virtually every major AI company press release when there's some major news. So maybe, Ben, if you could kick off the discussion by just giving us an overview of what was debuted yesterday.

Ben Cuss, CTO

Yeah, so for OpenAI, I did their event yesterday to talk about the future of work. And one of the things they did was to demonstrate how employees could utilize OpenAI doing kind of common work tasks in their demo they use Box to basically be the connected via their OpenAI agent to go pull data and this was because when you think about how somebody's gonna work in a company using AI they need access to the most critical data and of course OpenAI or Anthropic they don't store that data that needs to come from somewhere and so then since Box is a major repository of data they pulled it from Box this is similar to what Anthropica is showing. They did an event a couple weeks ago, and they did the same thing where they demonstrated that pulling the data out of Box, searching through it using our MCP server. And so this is a common thing that we see not only in these type of events, but also with our customers, is that they need to access their most critical data via these new agents.

Operator

That's interesting. And it dovetails into kind of the theme or the messaging from the company recently that Box is really positioned as a file system for AI. And so I wanted to ask you about how the value proposition of Box has evolved and will evolve as the primary users of software shift from humans to agents. Does this make Box essentially transform to, in some ways, an infrastructure layer for data? I would love to hear you speak more about that.

Ben Cuss, CTO

Sure. Sure. So in general for Box, we've always maintained the idea that we're an unstructured content platform. And so we are able to handle all aspects of dealing with files and dealing with your most voluminous data, which is your unstructured data, about 90% of data and organization. But typically or historically, only humans could really understand if you're going to watch that video or understand that Excel sheet or see that PowerPoint or for more specific tasks, like understanding the financial reports, understanding the proposals for the life sciences company, and just the very myriad of these very critical aspects of content in a company. So with AI, there's now a new thing that can understand this, which is the AI agents. They're kind of born on unstructured data. And so they very naturally use files. It's like almost the preferred mechanism of sort of the way that they read information, the way that they output information. So we're starting to see the idea that not only are people reading and creating data and collaborating with other people in our data, but agents are also. It's becoming very common now for agents to write out files. If you wanted to work on a project, you want an agent to help you, you basically collaborate with them in your files, and then they're able to then create information and share it back to you. And, of course, all the challenges around collaboration, making sure that they have access to it, making sure that they have everything secured, that becomes part of the interaction that these companies are doing with these agents. And so for Box, we maintain all of these capabilities for users, you know, on their web app, on their mobile app, but then now also for agents via MCP, via CLI, via all the new techniques that agents need to access data. So we kind of see it as an extension of a foundation we've built for a long time, but then very much focused on making sure that agents have the ability to work with people on some of the most critical data.

Dylan Smith, CFO

Yeah, and just a little bit, that seems louder, but to say that, you know, fortunate that really everything we've been building for the last 20 plus years and the way that humans interact with Box translates directly into the way that agents want to work with Box, minus they don't care about the user interface necessarily, but all of the same, you know, kind of file system, the way that they're most comfortable working with data. And because there's going to be an order of magnitude more agents and because those agents, unlike humans, don't necessarily have an awareness of, oh, I probably shouldn't be doing this or, you know, caring about security or kind of the, you know, EQ, all of the capabilities you build up around security, permissions, compliance become that much more critical because, you know, otherwise you could have those sort of, you know, leaks, inappropriate access to information. whatever it is and not even by you know ill intent there's no intent they're just trying to get the right information and so you know really well positioned for this new agent again fantastic and you know the messaging from the company

Operator

has been very consistent that improvements and AI models are actually a tailwind for the company and we've actually seen that manifest in growth which is pretty rare in the app sector today so maybe just help us think about as models get better and better, inference costs decline, where you think the most critical components of differentiation lie within the box platform? And what's most difficult to replicate by a competitor or a frontier model provider?

Ben Cuss, CTO

I think when many customers we're talking to, they start to realize that they don't necessarily have an AI challenge. Specifically, they have a lot of really great sources of where to get AI from, right? Like Anthropic is great. OpenAI is great. Gemini is great. But they do have a challenge of making their data accessible to AI in a way that is secure, in a way that gives the agent the tools that it needs. Like imagine in an organization, let's say that it has a relatively sizable organization, hundreds of millions or billions of files, but different people only have access to different things. And you can't just have an agent go look at anything. If it's talking to you, it doesn't keep secrets. so it has to only access what you have access to. It has to find the data, and it has to keep that secure, and then it has to make sure that it obeys all of your GDPR and all of your compliance challenges for FINRA, HIPAA, and so on. So this foundation of storing and maintaining the permissions, the data, keeping it safe, keeping it collaborated on, and making sure the agents can use that effectively, that's where the vast majority of the hard-to-replicate challenges are. And then having the AI, just like we talked about, open AI or Anthropic, access it is sort of where it all comes together in something you consider to be like extra levels of enterprise productivity and things that they couldn't do before. And so I think many people are focused on the layer of the agents, which is really critical, and we have our own agents. But then the underlying aspect of the headless platform that powers it is where we see most of our traditional value in a lot of our work over the years.

Operator

And I want to get into specific product cycles because the enterprise advanced um rollout it's been out in the market for a little over a year now it's been very well received um you've seen customers pretty regularly upgrade from enterprise plus to enterprise advanced but dylan i wanted to ask you about um how you think about the cadence of enterprise advanced um going forward how do we how do we get comfortable with the notion that this adoption cycle is durable and we haven't seen some sort of pull forward in demand in the enterprise advanced upgrade cycle?

Dylan Smith, CFO

Sure. So at a high level, I'd say we're still very early days and very excited about the future opportunity and growth that enterprise Advanced can drive. So for context, last year, we exited from, you know, starting from a standing start with Enterprise Advanced making up about 10% of our total revenue. We expect them this year with that number to about 20%. And then, you know, three to five years out for that to be, you know, full 50% or more of our revenue. So still a multi-year, you know, kind of cycle generally that we expect to see with the adoption of enterprise advanced. And then on top of that, as we build more and more capabilities into the platform, and we're innovating, you know, in large part because of all the capabilities of AI and what that unlocks, expect that to open up more and more use cases. So even enterprise advanced customers will be able to expand seats as more and more of those use cases, you know, resonate and are applicable to what they're doing in their organizations. And increasingly, we think there are some really compelling, you know, consumption-based use cases that we'd be able to monetize and are already starting to monetize through AI units. And so, you know, we do think that at some stage, you know, The enterprise advanced upgrade directly will not be driving as much growth as they are today, but that's years and years in the future. And on top of that, you're thinking five years out, we may have another kind of whole suite on top of enterprise advanced. There are ways, especially as we're seeing a lot of the demand and a lot of our go-to-market efforts that we focused on are focused on verticals, we could verticalize some of these SKUs as well as a way to monetize. So we're really excited about the different ways that we can, you know, deliver more and more value to our customers and then certainly capture that value ourselves.

Operator

And then can you just help us think about how the different levers for ACV Uplift have evolved throughout the, you know, different stages of the enterprise advance upgrade cycle? obviously there's the pricing component there's the seed expansion component would love to hear from you dylan how that has evolved over the last year and how you expect it to to evolve going forward yeah so actually uh over the last year and what we've been seeing with enterprise advanced has actually been fairly consistent so we'll talk about the dynamics but um wouldn't say there's been a huge change in terms of what we're seeing with enterprise advanced upgrades you know nine

Dylan Smith, CFO

months ago versus today other than the types of use cases that customers are adopting enterprise advanced for just because we have more capabilities today than we did even a year ago but what we see in terms of the impact on contract value within an upgrade and we use as just one example from enterprise plus our previous kind of most premium suite into enterprise advanced because that's the most common upgrade in upsell motion. We tend to see an increase on a price per seat basis at the higher end of the 30 to 40 percent range. And then on top of that, in a little more than a third of enterprise advanced upgrades, we are also seeing seat expansion. And so that is because on day one, there's a very clear, okay, now I'm going to bring that into, you know, these net new use cases because these capabilities so i need more seats on box in addition to the enterprise advanced capabilities and we'd expect over time um you know once that upgrade has happened a customer is on enterprise advanced uh for more and more of those uh new use cases to emerge and for seat expansion to have a greater and greater impact um you know in those customers as well that makes a ton of sense.

Operator

I want to talk about budgets because clearly, you know, the platform capabilities have expanded significantly over the last 12 to 18 months. But are you noticing in your discussions with other, you know, IT buyers that you're unlocking new budget or, you know, AI budget, so to speak, rather than the traditional content management budgets you've been selling into over the last several years?

Dylan Smith, CFO

Yeah, that's definitely been the trend. It would say that whether it is net new, AI-specific, you know, tapping into and much more directly, you know, kind of working more closely with CISOs, not just as a deal supporter, but an actual driver and funder of Vox deployments, that is definitely what we're seeing. And then in a lot of cases, I would say it is the exception rather than the norm that it's as clean as saying, okay, you know, I was spending X on, you know, open text, now I'm ripping it out or now I'm reducing that skimfully to Y, and that delta is how I fund box. You know, we do see that occasionally, and that may be the, you know, ultimate, you know, path for a customer, but much more often what we see is, you know, they're finding new budget pools, either net new or, in some cases, because of all the different things that we can do, you know, they are ripping out other IT systems, either directly because of Box or just orthogonally, and that's what's funding the purchase. So I'd say much more often it's net new or outside of the kind of traditional enterprise content management budget.

Operator

Right. And understanding it's still very early for AI unit monetization and API monetization, the platform component of growth, how do you expect customers to budget for the usage or the consumption component of box spend going forward? Is this something that they're going to allocate, you know, X dollars for and be comfortable with, you know, X plus Y spending? Is this something that people are willing to essentially increase spending on mid-contract? Or how do you think about budgeting when it comes to consumption?

Ben Cuss, CTO

Sure. I think what we're seeing is that many customers are getting, as AI improves, as the capabilities improves, they're starting to realize they can apply to more use cases. So we do see that sometimes they start out with a certain perception of what they can do, something maybe like document extraction, where they start to apply to some subset of data. They realize it works really well. They like the enterprise advanced features. And then they start to purchase more because something like enterprise advanced, or document extraction is a resource-based, AI unit-based. you'd see them sort of scale it linearly according to how much they need to do. So I think that a lot of customers are still sort of assessing the value and sort of how this works overall. But certainly there's many people are starting to notice that as we're seeing the value in it, that they can increase it pretty in a straightforward way where they purchase more of these units, and therefore that turns into sort of more value for them, oftentimes for these new use cases that before weren't really possible.

Dylan Smith, CFO

Yeah, and the only thing I'd add is the way that we sell Enterprise Advanced, which includes some of that consumption that gives customers the ability to test it out, has also really helped with that selling motion and with helping customers to get a better understanding of what they need in some of the use cases. So, for example, if you did have a large-scale, you know, in a contract management lifecycle type use case, you know, you're ingesting a bunch of content, extracting a bunch of metadata, and then, you know, kind of kicking off a workflow from there, they'd be able to test that and see, okay, which is the best model, you know, how many AI units is that using? okay great we're happy with these results it's accurate you know 98.5 percent of the time okay great now knowing that we you know process roughly x contracts you know on a monthly basis can get a pretty good sense of that and they've been able to validate that the cost you know of those ai units versus the value that they're providing or that they're getting out of it is a you know very very strong roi and they can just scale up from there um and so you know there's again as ben mentioned also you know a lot of uh you know additional use cases that are uncovered uh hypothesized whatever over time and then that might lead to um you know kind of mid-contract expansion or whatever but you know in in most cases um you know customers are then going to want to validate um you know those use cases as well um so they have a pretty good sense of what they need and that they feel good about that level of spending.

Operator

Understood. And I wanted to pivot back to product because certainly there's been, you know, quite a few interesting new product rollouts and announcements over the last six to 12 months. It's almost hard to dial into what exactly is going to be the most important in terms of needle mover for growth. So across, you know box extract, automate, box agents, apps, where should we as investors be centering our focus in terms of what's monetizable specifically over the next 12 to 18 months?

Ben Cuss, CTO

So I think especially the ones you mentioned around automate apps extract is they often work together very well and so you start to see that like customers who have different sets of challenges one of those solve and start to adopt the other ones. So something specific like many companies that have a lot of content that need to know more about it, they'd use Xtract to basically structure their information. And so then they can go through and sort of take all sorts of various types of documents, oftentimes millions or more of these kind of things, structure them, understand what they're all about. And then as they do that, then they say, well, now that I have this information, I want to better sort, filter, understand it, create dashboards around I'd be able to search it and go through it and then that's something that box apps would do and as they do that They say okay. Well now I need to we'd have normal processes associated with these They're manual sometimes So how about I use automate to help me go through and automatically process these oftentimes based on that extract information? And so those kind of things are where advanced all works together So you see that like something like the fundamental ability powered by AI to extract a document is valuable by itself But this question that comes to all the customers ask, which is like, now what can I do is then answered by those other capabilities. And that's what we're seeing a lot of the enterprise advanced adoption is that people are saying, this data has become so much more valuable than it was before that we can now then process it at scale, much like you do with structured data, and then be able to then use it at scale. And then you need new tools, which is what Automate and Apps provides.

Operator

Maybe talk about any of the gating factors preventing, you know, greater uptake of the AI capabilities today. Is there, you know, we talk or we hear a lot about data readiness as an issue for enterprise adoption of things like agents. Is that influencing Box today, or how do we think about kind of the gating factors?

Ben Cuss, CTO

I think in general, for companies, data readiness is one of the key things. Now, for Box specifically, like, our data is sort of inherently secure and ready to use on agents, so we start to see people who are utilizing that. But I think right now the trend in general that you see is that people are still getting comfortable with what agents can do. Even if you tried it three months ago, six months ago, and you tried to do something and maybe it did or didn't work. Like nowadays, because of newer models, because of newer harnesses, because the way that they can access data, it's able to do more. And then I still think that not only at a per-individual level, but also at an enterprise level, they still are sort of continuing to try and experiment and to trust these new things. And I think these days, you start to see that the people who are using the more advanced use cases around it, they're starting to get actual, like, serious value of it. For instance, something like using document extraction, powering a workflow, using an agent to then process and create outputs very quickly to power, let's say, like a financial review or a customer review, that kind of thing. And so I think you see a parallel in the world of engineering with everybody who's using the latest engineering tools, you know, cloud code codex all the different new engineering agents and even sophisticated users of of engineering are still getting used to what's possible what's not possible and I think we're seeing the same trend across knowledge work a little bit behind where even though ages can do something you haven't seen everybody fully adopt it yet because they're still sort of understanding the capabilities that continue to change so I think this is one of the big trends that we're seeing is that people are saying I can't believe it can do that I'm going to use it more both for process in addition to just individually to help, you know, so your daily work.

Operator

And Dylan, I want to talk about efficiency and margin because clearly Box is already benefiting from, certainly from an engineering standpoint, but also on a go-to-market front from AI efficiencies. But how do you think long-term about, because you're getting all these positive signals from your new AI features, how do you balance the, you know, accelerating investment in growth to continue to yield those benefits versus driving margin expansion?

Dylan Smith, CFO

Yeah, so I would say philosophically, you know, there hasn't been a whole lot that's, you know, changed in terms of how we think about, you know, the growth versus profitability dynamic, where the biggest input into what we think about as the right level of investments tend to be on this sort of return that we're getting from our sales and marketing investments, right? You know, Salesforce productivity and how the newer cohorts of AEs are ramping, you know, what's the return we're getting on our marketing programs, our partnership investments, things like that. And, you know, I think AI changes the equation in terms of what we're seeing in different parts of the organization and that impact, but, you know, hasn't really changed the overall philosophy, and we remain very confident in and committed to increasing both our growth rate and our operating margin over the next several years. So, you know, really what we are seeing is, I would also say that different parts of the business and where we're driving efficiencies, what we do with that can be a little bit different, right? Because, for example, we can say, hey, with, you know, the way that we've been able to automate some of these different processes, you know, with a lot of the capabilities Ben was just talking about, you know, our, you know, commercial legal team can now, you know, handle twice as many sales contracts. We don't need to double or grow, you know, our, that team. But if we're saying, okay, because of all the ways that we're able to automate, you know lead flow and move customers through that buyer's journey much more efficiently our sales force productivity is up 50 percent you know our bias is going to be let's hire more sales people not like great now we can grow at the same rate with you know a third of the sales force and so there's a lot of moving pieces in there but you know ultimately you know really comes down to what is the ROI we're seeing on particularly our go-to-market investments and we're able to fuel that as we've been doing this year and as we've talked about because of a lot of the efficiency we've been able to drive across the business in other areas.

Operator

And then maybe just frame the the long-term gross margin outlook for Box because there's there's obviously a lot of moving pieces on both product mix, business model with you know platform fees becoming more important, and AI units, and so on. So how do you frame the long-term gross margin outlook for Box?

Ben Cuss, CTO

Sure. So some of the sort of the key thing, I mean, we operate infrastructure at scale, and one of the jobs of doing that is to do it very efficiently. And we operate at a scale where efficiency matters to us in terms of, and we can do things to like an aggregate that maybe is hard for, let's say, a smaller company to do in terms of driving sort of the right margins. So whenever we're delivering something, let's say something like Extract or the way our agents work, not only are we focusing on our value so that customers will be happy to pay the resource-based usage because they're getting that value and it's hard for them to replicate themselves, but then also we're able to really focus on the efficiency of it at scale, and that is usually what drives sort of the margins that we expect overall.

Dylan Smith, CFO

Yeah, and I would say that, you know, with AI and AI units and all these things, I think we've done, and Ben and I work very closely on this, we've done a good job positioning and kind of extracting some of the underlying costs and with tokens through AI units, which gives a lot of flexibility as models are constantly evolving in addition to the way that we're able to kind of intelligently swap out models to get the best outcome for customers, but also to make sure that we're doing that as efficiently as possible. As most people are aware, there's an order of magnitude difference in the cost depending on which models you're running. And so we don't expect the gross margin outlook to be materially different from what we've talked about, what we're delivering right now. But certainly, over time, we could see that if AI units become a much bigger part of our overall revenue, we wouldn't necessarily expect those kind of pure AI use cases to have 80 plus percent gross margins. We would expect them to be very strong and to be accretive, both to the profit that we're generating as well as to our overall growth. But I think that mixed shift is going to be, you know, one of the core drivers of just ultimately where that gross margin profile, you know, kind of ends up.

Operator

Absolutely. And we have about three minutes here left in the fireside. So I wanted to take this opportunity to open it up to the audience. If we had any investor questions, feel free to raise your hand and we'll get you a microphone.

Ben Cuss, CTO

Is there a consumption use case that you can go to customers with right now that's just super easy to sell that you're seeing kind of viral adoption of? document extraction like there's um just a huge amount of data in organizations that have the form of pdfs uh these different kind of client files these different research projects and and this is some of the most valuable data in these organizations but they just live in these folders historically different places and so you you say the way it works in box is you just say if you want to pull out this structured data think of a row in a in a in an excel file or in a database and just say, apply it to these thousand, these 10,000, these million, these billion files, and then you'll have that data, and then you can sort, filter, do whatever you want with it, investigate it, use it for agents, use it in apps. Like, that is the highest volume, the most interesting for customers. And historically, they'd have been able to do something like that if they had a team of 20 machine learning scientists who only dealt with that one type of data. But, like, because there's so much variation in the data, it's sort of, huge amount of data was never structurable but now for box for box extract it's just using latest AI capabilities you say structure any info and you get structured info and unstructured info together and that's what many companies are saying that's amazing the accuracy levels continue to go up it continues to be operationalized on how the agent can use it how the humans can use it that's that's a lot of people are really focused on that and maybe just a follow-up on that on that use case specifically how does box differentiate on the extraction side given that you know agenda capabilities are advancing you know very rapidly and you've even got vendors like databricks

Operator

snowflake who are who are you know showing pretty impressive data extraction capabilities how do you layer that um into the box platform and differentiate i think those companies are the structured companies are great in that they're able to take unstructured data inside of their structured systems to be able to gain insights in those.

Ben Cuss, CTO

Our world is really about the files and about how people have so much of their data in the form of these files in different places, client records, research proposals, like audits, like just this tremendous volume of normal stuff across every company. So our specific value and the thing that we sort of really excel at is this idea of being able to take those in whatever arbitrary form and then be able to extract the structured info and now it's um it's not just you know you can't just say to an agent like do it it has there's a ton of work around uh being able to convert it to write formats like some sort of markdown style being able to look through the different pages like the some of the formats are notoriously hard to do for agents so you have to spend a lot of time converting it this idea of ocr continues like old concept continues to evolve rapidly so all the technology that goes into this it's really a multi-step part only the last bit is the what you consider a traditional ai agent and so this is where a lot of customers are saying like hey if i try myself something works but it doesn't work as well as yours and so and most people don't like to spend all their time building these things we spend have whole teams to do it so this is where they're saying okay i'm happy in a good price point they'll be able to do that and they get a ton of value out of it it looks like we're out of time here but thank you so much ben thank you so much dylan for for for for showing up today and and educating us on box thanks so much thanks for having us