Very fortunate today to have Dylan Smith, Co-founder and CFO of Box, as well as Ben Kus, the CTO of Box, to discuss the Box story and position in the agentic tech stack today. Maybe just to kick it off, there was a pretty interesting demo from OpenAI yesterday where Box was featured prominently. It seems to me like Box is mentioned in virtually every major AI company press release when there's some major news. Maybe, Ben, if you could kick off the discussion by just giving us an overview of what was debuted yesterday. Yeah. For OpenAI 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 common work tasks. In their demo, they used Box to basically be the connected via their OpenAI agent to go pull data. This was because when you think about how somebody's going to work in a company using AI, they need access to their most critical data. Of course, OpenAI or Anthropic, they don't store that data. That needs to come from somewhere. Since Box is a major repository of data, they pulled it from Box. This is similar to what Anthropic is showing. They did an event a couple of 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. This is a common thing that we see not only in these type of events, but also with our customers, is with their need to access their most critical data via these new agents. That's interesting. It dovetails into the theme or the messaging from the company recently that Box is really positioned as a file system for AI. 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? Would love to hear you speak more about that. Sure. In general for Box, we've always maintained the idea that we're an unstructured content platform. 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 in an organization. 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 a more specific task, like understanding the financial reports, understanding the proposals for life sciences company, and just the very myriad of these very critical aspects of content in a company. 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. We're starting to see the idea that not only are people reading and creating data, 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 basically collaborate with them your files, and then they're able to then create information and share it back to you. Of course, all of 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. For Box, we maintain all of these capabilities for users, on their web app, on their mobile app, now also for agents via MCP, via CLI, via all the new techniques that agents need to access data. We kind of see it as an extension of a foundation we've built for a long time, very much focused on making sure that agents have the ability to work with people on some of the most critical data. Just build a bit. That seems louder. To say that, fortunate that really everything we've been building for the last 20+ 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 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 caring about security or kind of the EQ, all of the capabilities you build up around security, permissions, compliance become that much more critical because otherwise you could have those sort of leaks, inappropriate access to information, whatever it is, and not even by ill intent. There's no intent. They're just trying to get the right information. Really well-positioned for this new agentic era. Fantastic. The messaging from the company has been very consistent that improvements in AI models are actually a tailwind for the company. We've actually seen that manifest in growth, which is pretty rare in the app sector today. 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? I think when many customers who 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. 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. 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. 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. Having the AI, just like we talked about, OpenAI or Anthropic access it, is sort of where it all comes together in something you'd consider to be extra levels of enterprise productivity and things that they couldn't do before. I think many people are focused on the layer of the agents, which is really critical, and we have our own agents. The underlying aspect of the headless platform that powers it is where we see most of our traditional value and a lot of our work over the years. I want to get into specific product cycles because the Enterprise Advanced rollout, it's been out in the market for a little over a year now. It's been very well received. You've seen customers pretty regularly upgrade from Enterprise Plus to Enterprise Advanced. Dylan, I wanted to ask you about how you think about the cadence of Enterprise Advanced going forward. 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? Sure. 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. For context, last year we exited from starting from a standing start with Enterprise Advanced customers making up about 10% of our total revenue. We expect them this year with that number to about 20%. Three to five years out for that to be full 50% or more of our revenue. Still a multi-year kind of cycle generally that we expect to see with the adoption of Enterprise Advanced. On top of that, as we build more and more capabilities into the platform and we're innovating in large part because of all the capabilities of AI and what that unlocks, expect that to open up more and more use cases. Even Enterprise Advanced customers will be able to expand seats as more and more of those use cases resonate and are applicable to what they're doing in their organizations. Increasingly, we think there are some really compelling consumption-based use cases that we'd be able to monetize and are already starting to monetize through AI Units. We do think that at some stage, 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. 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. We're really excited about the different ways that we can deliver more and more value to our customers and then certainly capture that value ourselves. Can you just help us think about how the different levers for ACB uplift have evolved throughout the different stages of the Enterprise Advanced upgrade cycle? Obviously, there's the pricing component, there's the seat expansion component. Would love to hear from you, Dylan, how that has evolved over the last year and how you expect it to evolve going forward. Actually, over the last year and what we've been seeing with Enterprise Advanced has actually been fairly consistent. We'll talk about the dynamics, but wouldn't say there's been a huge change in terms of what we're seeing with Enterprise Advanced upgrades nine 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. What we see in terms of the impact on contract value with 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 and upsell motion. We tend to see an increase on a price per seat basis at the higher end of the 30%-40% range. Then on top of that, in a little more than a third of Enterprise Advanced upgrades, we are also seeing seat expansion. So that is because on day one, there's a very clear, okay, now I'm going to bring that into these net new use cases because these capabilities, so I need more seats on Box in addition to the Enterprise Advanced capabilities. We expect over time, once that upgrade has happened, a customer is on Enterprise Advanced for more and more of those new use cases to emerge and for seat expansion to have a greater and greater impact in those customers as well. That makes a ton of sense. I want to talk about budgets. Clearly, the platform capabilities have expanded significantly over the last 12 to 18 months. Are you noticing in your discussions with other IT buyers that you're unlocking new budget or AI budget, so to speak, rather than the traditional content management budgets you've been selling into over the last several years? Yeah, that's definitely been the trend. I would say that whether it is net new AI specific, tapping into and much more directly, working more closely with CISOs, not just as a deal supporter, but an actual driver and funder of Box deployments, that is definitely what we're seeing. In a lot of cases, I would say it is the exception rather than the norm that it's as clean as saying, okay, I was spending X on OpenText, now I'm ripping it out, or now I'm reducing that significantly to Y, and that delta is how I fund Box. We do see that occasionally, and that may be the ultimate path for a customer. Much more often, what we see is they're finding new budget pools, either net new or in some cases, because of all the different things that we can do, they are ripping out other IT systems either directly because of Box or just orthogonally. That's what's funding the purchase. I'd say much more often it's net new, or outside of the traditional enterprise content management budget. Right. 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 X dollars for and be comfortable with X plus Y spending? Excuse me. 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? You start? 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. 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 it to some subset of data. They realize it works really well. They like the Enterprise Advanced features, then they start to purchase more. Something like Enterprise Advanced and document extraction is resource-based, AI Unit-based. You'd see them sort of scale it linearly according to how much they need to do. I think that a lot of customers are still sort of assessing the value and sort of how this works overall. Certainly many people are starting to notice that as they're seeing the value in it, that they can increase it in a straightforward way, where they purchase more of these units, and therefore, that turns into more value for them, oftentimes, for these new use cases that before weren't really possible. The only thing to 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. For example, if you did have a large scale, in a contract management life cycle type use case, you're ingesting a bunch of content, extracting a bunch of metadata, and then you're kind of kicking off a workflow from there. They'd be able to test that and see, okay, which is the best model? How many AI Units is that using? Okay, great. We're happy with these results. It's accurate 98.5% of the time. Okay, great. Knowing that we process roughly X contracts on a monthly basis, can get a pretty good sense of that. They've been able to validate that the cost of those AI Units versus the value that they're providing, or that they're getting out of it is a very strong ROI, and they can just scale up from there. There's, again, as Ben mentioned, also a lot of additional use cases that are uncovered, hypothesized, whatever, over time, and then that might lead to mid-contract expansion or whatever. In most cases, customers are then going to want to validate those use cases well, so they have a pretty good sense of what they need and then they feel good about that level of spend. Understood. I wanted to pivot back to product because certainly there's been quite a few interesting new product roll-outs and announcements over the last 6- 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. Across 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- 18 months? 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 customers who have different sets of challenges, one of those solved start to adopt the other ones. Something specific like many companies that have a lot of content but need to know more about it, they'd use Extract to basically structure their information. 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. As they do that, they say, well, now that I have this information, I want to better sort, filter, understand it, create dashboards around it, be able to search it, and go through it. That's something that Box Apps would do. As they do that, they say, okay, well, now 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 extracted information. Those kind of things are where eAdvanced all works together. You see that something like the fundamental ability powered by AI to extract a document is valuable by itself. This question that comes, all the customers ask, which is, "Now what can I do?" Is answered by those other capabilities. That's where 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. Maybe talk about any of the gating factors preventing greater uptake of the AI capabilities today. We talk or we hear a lot about data readiness as an issue for enterprise adoption of things like agents. Yeah. Is that influencing Box today? Or how do we think about the gating factors? I think in general, for companies, data readiness is one of the key things. For Box specifically, our data is sort of inherently secure and ready to use on agents, so we start to see people who are utilizing that. 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, nowadays because of newer models, because of newer harnesses, because the way that they can access data, it's able to do more. 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. 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 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, a financial review or a customer review, that kind of thing. I think you see a parallel in the world of engineering with everybody who's using the latest engineering tools. Claude Code, Codex, all the different and new engineering agents. Even sophisticated users of engineering are still getting used to what's possible and what's not possible. I think we're seeing the same trend across knowledge work a little bit behind, where even though agents can do something, you haven't seen everybody fully adopt it yet because they're still sort of understanding the capabilities. They continue to change. 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 a process, in addition to just individually to help your daily work. Dylan, I want to talk about efficiency and margin, because clearly Box is already benefiting certainly from an engineering standpoint, also on a go-to-market front from AI efficiencies. 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 accelerating investment in growth to continue to yield those benefits versus driving margin expansion? I would say philosophically, there hasn't been a whole lot that's changed in terms of how we think about the growth versus profitability dynamic, where the biggest input into what we think about is the right level of investment tend to be on the sort of return that we're getting from our sales and marketing investments, right? Sales force productivity and how the newer cohorts of AEs are ramping. What's the return we're getting on our marketing programs, our partnership investments, things like that. I think AI changes the equation in terms of what we're seeing in different parts of the organization and that impact. It hasn't really changed the overall philosophy, and we remain very confident and committed to increasing both our growth rate and our operating margin over the next several years. Really what we are seeing is, and also the different parts of the business where we're driving efficiencies, what we do with that can be a little bit different, right? For example, we can say, hey, with the way that we've been able to automate some of these different processes with a lot of the capabilities Ben was just talking about, our commercial legal team can now handle twice as many sales contracts. We don't need to double or grow that team. If we're saying, okay, because of all the ways we've been able to automate lead flow and move customers through that buyer's journey much more efficiently, our sales force productivity is up 50%, our bias is going to be let's hire more salespeople, not like, great, now we can grow at the same rate with 1/3 of the sales force. There's a lot of moving pieces in there, but ultimately, 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 efficiencies we've been able to drive across the business in other areas. Maybe just frame the long-term gross margin outlook for Box, because there's obviously a lot of moving pieces on both product mix, business model with platform fees becoming more important, and AI Units, and so on. How do you frame the long-term gross margin outlook for Box? Can you start? Sure. Some of the key thing for We operate infrastructure at scale, and one of the jobs of doing that is to do it very efficiently. We operate at a scale where efficiency matters to us in terms of, we can do things to like in aggregate that maybe is hard for, let's say, a smaller company to do in terms of driving the right margins. Whenever we're delivering something, let's say something like Box Extract or the way our Box Agents work, not only are we focusing on 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, also we're able to really focus on the efficiency of it at scale. That is usually what drives the margins that we expect overall. Yeah. I would say that with AI and AI units and all these things, Ben and I work very closely on this. We've done a good job positioning and kind of abstracting 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. We don't expect the gross margin outlook to be materially different from what we've talked about, what we're delivering right now. 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% 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. I think that mix shift is going to be one of the core drivers of just ultimately where that gross margin profile kind of ends up. Absolutely. 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. 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. Yeah. There's just a huge amount of data in organizations that have the form of PDFs, these different kind of client files, these different research projects, and it's just some of the most valuable data in these organizations. They just live in these folders historically, different places. The way that it works in Box is you just say, if you want to pull out this structured data, think of a row in an Excel file or in a database and just say, apply it to these 1,000, 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. That is the highest volume, the most interesting for customers. 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. Because there's so much variation in the data, it's sort of huge amount of data was never structurable. 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. A lot of people are really focused on that. Maybe just to follow up on that, on that use case specifically, how does Box differentiate on the extraction side, given that agentic capabilities are advancing very rapidly, and you've even got vendors like Databricks, Snowflake, who are showing pretty impressive data extraction capabilities. How do you layer that into the Box platform and differentiate? I think the structured data companies are great in that they're able to take unstructured data inside of their structured systems to be able to gain insights from those. 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, audits, just this tremendous volume of normal stuff across every company. 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. Now you can't just say to an agent, do it. There's a ton of work around being able to convert it to the right formats, like some sort of markdown style, being able to look through the different pages. 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. All of the technology that goes into this, it's really a multi-step part. Only the last bit is what you consider a traditional AI agent. This is where a lot of customers are saying, hey, if I try myself, something works, but it doesn't work as well as yours. Most people don't like to spend all their time building these things. We have whole teams to do it. 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 the ton of value out of it. Well, it looks like we're out of time here. Thank you so much, Ben, thank you so much, Dylan, for showing up today and educating us on Box. Thanks so much. Thanks for having us.