Aaron Levie, Box CEO: A Fortt Knox Update

We got credit. Yep.

It's a good thing we're an AI company only and we don't sell any software.

Well, I think they're... first of all, I actually am sympathetic to 80% of the argument. I think we're going to have 100 times more software in the future. I think that we're going to further digitize more and more work. So that will clearly be the case. I think that the power of the system of record that is able to traffic cop the workflows of agents, what data they have access to, what they're able to do with that data, what permissions they're able to take on that information, I think the value of those systems will be even more valuable in the future because there'll be a hundred times more agents running around trying to do all this work.

Box is a system of record for your unstructured data or your documents. So imagine that you want to go and automate a workflow around your contracts or your marketing assets or your research information. To be able to do that, you need a place where you can securely store the files. You need to be able to make sure that you can give permissions and access to the agents to go work on those files. You need the workflow itself that you can go and define that those agents should participate in. So all of those things you need software for. And that software we think will be even more valuable in a world where those agents can take very consequential actions on your information where they could generate a report that might have the wrong data in the report where they can produce information that might move the stock price of your company or what you end up selling to your customers. So the value of that system of record where those agents are operating in we think will be even more important whether it's your unstructured data or a CRM system like Salesforce or an HR system like Workday. And I think that's the part that maybe some of the investor market doesn't quite perceive.

I think it's a perfect question. First of all, there's one X factor in there, which is how competitive is the pricing of AI model inference in the future? And in a world where you have three to four major players all competing for that inference capacity, I would argue that that will get squeezed more and more over time from a margin standpoint, which then obviously leads more value at the application layer for being able to create additional profit. But overall, my take would be that what will happen in the future is you'll pay some base fee for software. It could be correlated to the number of employees in the organization. It could be correlated to how much data you have, some kind of base fee. Today that's again seat-based, but that could evolve. And then you have some consumption model which is correlated to how much work are agents doing on your system. And then the big question is what is the margin a software player gets on that agent consumption? Do they get 30%, do they get 50%, do they get 70%. That will probably be determined by how much value have you added in what that agent is working on. What's the context engineering required? What's the domain expertise required? What is the sort of software to enable that agent to be effective required and that will then correlate with how much value you can effectively take in that mix. So I'm largely bullish on the general trend simply because customers are going to want to go to a vendor and say I just want you to automate how our sales process works. I just want you to automate our contract management process. I just want you to automate our marketing design process. And the software company that can provide that full stack solution will be in the command position to be able to generate the incremental profits on that. And then how much does OpenAI get or Anthropic or the model provider that remains to be seen due to just how much competitiveness there is in that space. But I don't lose sleep over how will an application layer provider generate profit in a world where agents are doing so much work within the software of those applications.

We are already losing sleep on that dimension. But software has added to the mix a little bit.

Well, I think that obviously investors need to continue to see more proof points of what I'm at least explaining and what other CEOs, Mark Benioff, Bill McDermott, Mike Cannon-Brooks, etc. are clearly both seeing in the business and trying to explain. We're seeing very strong durable markets from our customers and the demand for how do I combine agents with my system of record to be able to do useful work in this environment. In our Q3, the last public quarter, we're showing continued momentum on our enterprise advance plan which is the plan that brings together all of our advanced workflow automation capabilities and advanced AI agent capabilities. So we're seeing a very rich demand environment for that solution. And that best correlates with again the future of where our customers are going which is they want to go and automate a lot of their knowledge work and be able to augment what people are doing with their data and how they accelerate that work. So we feel very confident in our position, our path forward. Obviously, I think a number of peers feel the same way and it might just take whether it's quarters or new KPIs or better messages to hopefully give that clarity to the market. Before long, part of what happens is if prices fall enough, people at big companies have these M&A strategy boards where, things I'd like to buy that are too expensive. When prices drop far enough, people start going out making offers. And not just people, private equity too. How far are we away from that do you think?

I think it's been already a relatively robust environment on that front and certainly price pressure might increase that even further and so maybe that provides some kind of natural stop gap on the pricing dynamic. But I'll probably let other folks be the drivers of those predictions. But in general, I think you've already had a relatively robust environment for M&A, whether it's private equity or large enterprises.

Well, we've actually been very focused on a specific type of M&A strategy, which is we do tuck-in acquisitions of fantastic teams that have built out in an existing space. So we actually largely have not been very valuation sensitive in the past because we are looking for very specific point solutions that we can use to accelerate our roadmap. Interestingly actually AI coding aids us in this front because there's a lot of solutions where now we can go and quickly build the capabilities that we need and so we might be able to get a faster time to market in actually some cases of just building it ourselves now with AI than even doing an acquisition. So we'll still be inquisitive where relevant in our kind of traditional pattern. But our market valuation hasn't impacted that.