In this episode of the McKinsey Podcast, Box CEO Aaron Levie speaks with McKinsey’s Simon London about how his team was able to start up and scale up the cloud-storage company, driving business growth while focusing on continuous innovation, and how artificial intelligence (AI) will factor into the company’s future.
Podcast transcript
Simon London: Hello, and welcome to this edition of the McKinsey Podcast, with me, Simon London. Our guest today is in some ways the archetypal Silicon Valley mogul. Aaron Levie cofounded Box while still an undergraduate and left his studies behind to focus on building the cloud-storage company. He talks fast, favors sneakers over highly polished oxfords, and now leads an organization valued at more than $3 billion. And yet, at college, Aaron was a student of business, not computer science. He’s an avid reader of business books, thinks deeply about strategy and organization, and, as we’ll hear, is interested not only in the future of technology but also in the future of management. So without further ado, fasten your seat belts for this conversation with Box CEO Aaron Levie. Aaron, thanks for doing this.
Aaron Levie: Thanks for having me. Really exciting.
Simon London: If you don’t mind, I’m going to ask, for people who might not know the company, just give us the story right from the very beginning. What’s the origin story? And take us through where you are today.
Aaron Levie: We started Box in 2005. We were sophomores in college, and we had this idea that it should be really, really easy to be able to share and access files from anywhere. It was a very basic idea. We launched this product. Eventually we got really lucky and got some early venture capital from Mark Cuban. That compelled us to drop out of college.
So we dropped out of college. We moved to Silicon Valley. As we thought about what we wanted to build as a company and as we wanted to scale the business, we decided to focus 100 percent on the enterprise market. Since 2007, we’ve been 100 percent focused on helping businesses securely manage, share, collaborate, and organize their critical information in the cloud.
We help companies manage everything: their financial documents, their media content, their marketing presentations, collaboration with ad agencies, collaboration with clients. We work with 85,000 customers globally, including about 70 percent of the Fortune 500.
Simon London: Employees now?
Aaron Levie: We have about 1,900 employees.
Simon London: And revenue this year is?
Aaron Levie: Last year, we did a little over $500 million in revenue. And this year, we’ve guided to about a little over $600 million in revenue.
Simon London: Thank you for that. The obvious follow-up question is: Along that journey, what have you learned about scaling a software company? Because that has special characteristics—but then maybe after that, just generalize a little advice to people who are on this journey.
You have to be nimble enough, you have to be agile enough, as new information is coming in, to constantly be evolving.
Aaron Levie: We’ve certainly learned a whole bunch of things not to do over the years. We’ve made our fair share of mistakes. The thing that has worked incredibly well is having a long-term vision for where you’re going.
There are so many things that have the potential to veer you off course. If you are not really, really sure of what you’re trying to do over the long run, like a ten-plus year vision, it is so easy to go off course. We were fortunate. Very early on, we had this vision, which was, we saw that everybody’s work style was going to be changing in the future. Every company was going to have to change the way they collaborated, the way they shared, the way they fundamentally ran their businesses. That was going to lead people to having to use the cloud to work and manage their data. And we were building an architecture and a strategy and a technology that could lead toward that vision of the future.
Simon London: And you were not afraid to pivot early on, right? So it’s not like you locked in too early.
Aaron Levie: No, we had a lot of flexibility early on. Probably too much. In the first year and a half of the company, I would say every 48 hours we’d change our business model. The way it would work is, we had four founders of the company, including myself, and what would happen was, if you went to bed too early one night, you were at risk of waking up and learning that the business model had changed.
There was a lot of early constant iterating, constant pivoting of the business. That became one of our core values as a company. We have seven core values that we really live by. And one of them is take risks, fail fast, and get shit done. The emphasis there being fail fast. The moment that we get information that we feel is going to be critical, we will make decisions rapidly on that. That has been built now into the culture of building an organization where you’re constantly iterating and you’re constantly testing ideas. You have to make sure that they line up against the long-term vision. But you have to be nimble enough, you have to be agile enough, as new information is coming in, to be able to constantly be evolving and veering the company in just slightly new directions.
That has stuck with us for a long time. There’s been a lot of near-death decisions and moments as a company where if we had gone a different direction, we’d be in a very different spot. One of those was the fundamental business model of focusing on the enterprise.
Simon London: It feels like you—very successfully as it turned out—decided to go for growth and to double down on growth. You wrote quite an interesting HBR [Harvard Business Review] piece about returns to scale. Do you just want to talk about that and the moments when you had to decide, “Do we double down on growth? Or do we start running the business for profitability and cash?”
Aaron Levie: Tell me if I’m wrong. I think you guys wrote a piece called, “Grow fast or die slow.”
Simon London: That’s it.
Aaron Levie: OK, good. Did I do a good plug just now?
Simon London: That was perfect.
Aaron Levie: Great. OK. It’s completely true in the software world where in some cases you have strong network effects. In other cases, the economies of scale you get by building software are obviously massive. There’s a premium—an orders-of-magnitude premium—on the companies that are number one or number two in their market relative to three, four, five, and so on.
So that means you really have to make sure you’re building for scale. The moment you believe that your product has product market fit and it’s working for customers, you’re pretty incentivized to growing as quickly as humanly possible [Exhibit 1]. We felt that moment in probably 2009, 2010 where, for the first time ever, large enterprises were saying, “OK, we’re fine with cloud. We think this is a more secure, safer way to do our work.”
The moment that we started seeing and experiencing that, and we saw that there were a bunch of tail winds that were driving our growth—things like mobile, cloud computing being adopted in the enterprise—we said, “OK, we’re going to now bet the whole business on growing as quickly as possible.” That was when we raised a few hundred million dollars over the subsequent few years because we had to make sure we were building up the technology team, the sales force, the go-to-market engine to reach every customer possible. We knew that this was a market that was going to be defined by the company that could have 90 percent of the large enterprises on the planet using their technology versus if you only had 10 or 15 percent of the enterprises.
Simon London: Presumably to pursue that strategy, you need to be very aligned with your investors.
Aaron Levie: Extremely.
Simon London: Because those are big bets. Did you need to reset with the original investors and say, “OK, maybe you thought you were going to start seeing some cash positive.”
Aaron Levie: I feel bad for our early investors because, I think from 2006 to 2012, I’m almost positive we pitched each subsequent investor on, “This is the round where we get to cash-flow positive.” We genuinely thought that was what was going to happen, and then what would happen is, our growth rate would exceed our expectations.
Or you just knew that with one more concentric circle of investment from where we were, we could grow even faster. That led everybody to conclude, including our early investors, that it was time to raise more capital and invest more in growth. Where start-ups get this wrong a lot is the sequence that is so fundamental. You have to make sure the product is working and that the customers want the product. Then you invest. A lot of start-up founders, VCs [venture capitalists], think that somehow you can buy your way to that, and you almost overfund your product development to the point where you might spend $10 million, $20 million, $50 million, $100 million just on building product well before the market actually even is interested in what you’re creating. And that’s where it’s really risky.
We were fortunate. We started the company with $15,000. We spent a total of $15,000 and proved that people wanted what we were building. That was in the consumer space. But then it only took a couple million more to prove that enterprises wanted what we had. From that point forward, just scaling. It’s really dangerous to start to prematurely scale using venture capital if you haven’t proved that your product both can scale, but then also the economics of the product actually make sense at scale.
Want to subscribe to The McKinsey Podcast?
Simon London: The interesting question is, to what extent is that specific to software, internet, and cloud businesses? It sounds quite dangerous, generalizable advice to a company making and selling shoes.
Aaron Levie: Oh, totally. The massive caveat should be, only take this advice if you have a recurring-revenue business where each customer you acquire is likely going to be with you for a long time and there are network effects. Certainly, if you’re selling a product on a one-off basis, no amount of scale is going to help if you have bad unit economics. You’re just now going to have bad unit economics at scale. It’s really hard to make a profit if each product that you’re selling is unprofitable, so it is pretty unique to software.
Simon London: The other thing that strikes me is, there’s a big debate about short term-ism and how providers of capital are too short term. But you didn’t show positive cash flow until how many years after the founding?
Aaron Levie: It’s too scary to talk about. But probably about 11 years.
Simon London: Right. And that’s really interesting.
Aaron Levie: Our most cash-flow-positive months were month one and month 137 or something.
Simon London: But that’s really interesting, isn’t it? It shows that there is patient capital out there, but you have to align the investors.
Aaron Levie: And it can be very scary at times. Because if you go off course for one or two quarters, then all of a sudden the whole model changes in some fundamental ways. When you’re burning cash, the model is very, very sensitive to what your assumptions are, and the moment those are wrong and either your growth is faster than you expected, so now you need more cash, or the growth is not there, but you spent the money. In both of those scenarios, it’s very dangerous. We have been blessed by a very strong finance and strategy-planning function. Without that, I think we’d be in a very different position.
Simon London: Right. Can we talk a little bit about technology trends?
Aaron Levie: I love those.
Simon London: You were born on the right side of history with respect to cloud, right?
Aaron Levie: Yeah.
Simon London: Then you got on the right side of history with mobile, as I understand it.
Aaron Levie: Yeah.
Simon London: That was, you saw that coming, and you moved on it. So what’s coming next, and how do you make sure you’re on the right side of history? What’s the big thing?
Aaron Levie: I think today’s version of those is, how are you or how is any company thinking about AI and machine learning? One of the really important things is, it’s not enough to just say, “OK, we’re doing AI” or “We’re doing machine learning.” The dependency of doing AI or machine learning really well is, do you have your data set in a form that you can make use of and make sense of?
And a lot of companies are not thinking through their long-term technology strategy to say, “Is my information being managed, stored, organized in a way that I’ll eventually get the kind of accretive benefits of AI on top of this data?” [Exhibit 2].
Simon London: I think it’s what people call data strategy, isn’t it? You’ve got to get your data strategy right before you even really start rolling out.
Aaron Levie: That’s exactly right. And you have too many organizations that have fragmented data where you don’t have the connection between different objects and between different data sets. For us, we are benefiting from the architecture decisions we made 13 years ago about being in the cloud. It means we have all the data in one place. I think a lot of customers have to think through, in three, or five, or ten years from now, are you going to be able to leverage best-in-class AI or machine-learning technology? Is your data in a format? Is it stored and managed in a manner that lets you take advantage of that? So that’s a big one.
When we think about our overall strategy, some of it is becoming one part technology as it relates to the deep technology architecture and one part more the science of management and how that is changing. The coupling of technology and business culture—and are we setting our product up and are we setting our business up to be at the center of where we see the future of work?
That obviously is super important to us because our whole product is 100 percent driven by, can we enable companies to work in a modern way? Which means we have to make sure we understand and we see what that modern way of working is all about. That’s where we spend the majority of our time.
Simon London: Let’s talk about the future of work, because it’s an interesting topic that has a number of different meanings. A lot of the stuff you see written about the future of work is just how much work there is going to be ten years down. But I’m guessing that’s not really what you’re talking about. You’re talking about how work will get done within enterprises.
Aaron Levie: To that point, there’s a future of jobs. So the jobs themselves: What are the job categories? Who is going to do what? Then to us, when we say future of work, we think about, OK, what does work look like? How does work get done inside of an organization? Whether that’s knowledge work or industrial, mechanical work, whatever that work tends to be.
We think we’re at this fundamental juncture where the 100-plus years of industrial-age management systems and practices are showing signs of not being able to scale for the digital age and not being able to translate well into the digital age. When you look at the hierarchies of organizations, when you look at the workflow patterns and processes of organizations, the waterfall methodology of product development and decision making, the asynchronous flow of data throughout an organization—all of these things we think are going to be rendered completely useless in the digital age. And not just useless, but in most cases slowing down companies and how they operate.
We think that, interestingly, a lot of the lessons to be learned are from software start-ups and the practices that smaller start-ups have had to learn in the software-development practice of being agile, being able to iterate quickly, having a tremendous amount of data to make decisions from, making sure that you have small teams that can move rapidly, being very, very close to the customer [Exhibit 3]. Those practices that were built out for software and for the internet translate now into every part of a business, whether that’s HR, finance, product development, and marketing. That’s the profound change in business, which is taking these agile, team-based practices of building things and now translating and having that manifest into every part of an organization.
Simon London: It’s moving from what I like to think of as agile with a capital “A,” or actual hardcore agile software-development methodology, to agile with a small “a,” or agility.
Aaron Levie: Yes.
Simon London: And how do you take some of these practices and that whole ethos and spread it out across a company?
Aaron Levie: Now, the question is, are large enterprises prepared both culturally and technically to be able to get there? We spend a lot of time with large-enterprise customers that are going through that journey, and they’re kind of saying, “OK, it’s not enough to just have modern technology. I can buy every cool new Silicon Valley software start-up product, but if my culture doesn’t change, I actually can’t get that much use from this technology.”
Conversely, you can have the world’s best HR leader and the world’s best CEO driving transformation, but if your technology stacks mean that people can’t share data, they can’t work in real time, they can’t collaborate across the traditional divisions in their organization, then no amount of “cultural change” is going to manifest in real productivity.
The thing that we’re seeing, which is a pretty interesting trend, is the complementary transformation that’s being driven from the IT technology organization as well as HR and operations. These things are feeding off of each other in a profound way and that we think are going to collude to then ultimately transform how companies will operate, which is that you have to get to smaller teams that move much more rapidly and that have permission to fail but iterate constantly. They have to have data to get their jobs done. They have to be able to connect up to the rest of the business, so people know what’s going on. That creates an environment of a lot more transparency, a lot more openness, and a lot more inherent accountability, because you can see what’s going on.
There are a lot of cultures that are not prepared for that. That’s going to be one of the biggest tests of which companies in the Fortune 500 make it to the digital age, or make it through the digital age, and those that don’t.
Simon London: How do you think about that in terms of Box’s own operations? Because, increasingly, you are delivering at scale, it has to be an incredibly reliable, secure service. So in many ways, you’ve got that foundation that you need to operate. It means we are totally reliable and secure. But on the other hand, as you say, you’re trying to move fast. You’re trying to iterate. You’re trying to learn quickly in agile-type practices. How do you marry those two things together? Which is the challenge that a lot of big companies face.
Aaron Levie: This is absolutely the paradox of the digital age, which is that you have to move insanely quickly to stay competitive. At the same time, there are a lot of business processes that were not built for that kind of speed of iteration: if you’re building a car, if you’re building a jet engine, if you’re delivering a new medical treatment.
Simon London: Yeah. Is it even desirable?
Aaron Levie: Right.
Simon London: I’d rather have my plane made slowly. I don’t want it to fail fast.
Aaron Levie: This is definitely the challenge that a lot of organizations face. I think the key is to be very, very clear on which business processes you can afford to have this kind of level of iteration, agility, and fail-fast mentality. We don’t have a take-risks approach in our accounting process. However, we still use a lot of agile practices when we are closing the books.
We effectively have scrum team meetings where everybody’s getting together, standing up, talking about what are the latest activities that we have to do. The failing fast is, if something’s off track, we’re going to iterate, and we’re going to know within a day or two of that happening, as opposed to the very typical process, which is, I only find out three weeks later that somebody upstream from this process ended up doing something in a way that had the numbers wrong or created some problem.
The idea of agility, and the idea of agile with the capital “A,” is, you get people much closer to the work and much closer to the ultimate customer. They are sharing information on a more frequent basis. They are much more accountable and fully own the problem. That transcends whether you’re building the most trivial software or doing jet-engine design or missile design.
Would you like to learn more about our High Tech Practice?
Simon London: So just get a little more specific, if you don’t mind, about AI, machine learning. There’s a lot of hand waving about it. I know you’ve started to build this into the product and platform already. But what’s it going to allow the product and the platform to do that it doesn’t do today?
Aaron Levie: Box stores tens of billions of files, hundreds of petabytes of data. Most of that information is relatively unstructured. So you take a Word document, or an image, or a video, and you store it in Box. You share it with somebody. But it’s a relatively closed, contained object that we and you don’t know much about at an abstract level. You can look at your file, but you don’t know the essence of the data.
Simon London: No, you don’t know what’s in there, apart from the file name.
Aaron Levie: Exactly.
Simon London: Usually. It could be anything.
Aaron Levie: Exactly. And so you have to go and read it manually to know what’s going on inside that piece of content. The power of AI is that you can extract information and data from the content itself. If you take an image that a CPG [consumer packaged goods] company has of a new product, you would be able to extract what the objects are in that image. So what are all the product names? What SKU number is it? And be able to take that information.
If you were a bank and you had a customer record or a contract, you’d be able to pull out all of the important data about that contract from that customer. So the power of AI is that you can begin to create structure and context from all of your information. And then the question is, what could you do with that? Can I automate business processes based on different rules? As certain images come into the system, I want to be able to route those images to the right person to approve them or review them.
I want to be able to automate an insurance-claims process by taking in all of the video or an image from an insurance adjuster or from the consumer themselves and automate that entire process. Or, going as far into some of the more mission-critical aspects of it, I want to be able to understand how my data’s being shared, because there might be security events that are happening where somebody’s sharing a previously unknown-to-be-confidential document, but with AI, I now know that there’s confidential information inside that document. That person’s sharing it with somebody that they’re not supposed to be. I can now alert the security team or block that off.
Simon London: So it’s things like optical character recognition, image recognition, basically creating this sort of metadata layer that’s autogenerated. And then, as you say, what can you do with that?
Aaron Levie: Natural-language processing.
Simon London: Natural-language processing. And then can you start to trigger automated business processes as the information moves through the system?
Aaron Levie: That’s exactly the vision. So take a whole bunch of work flows that were either being done in a highly manual way or never being done at all, because they would be cost prohibitive to do. And begin to take the power of AI, structure your information, and then begin to automate those processes. Again, everything in a healthcare-delivery process to an insurance-claims process to a digital-asset-management-review process for a CPG company, that’s how we are approaching AI.
But I think the exciting thing is to start to think through all of the tasks that we spend time on inside of a workplace that you just know a computer could do better. A computer can connect the dots between our collective calendars way faster than we can. A computer can connect the dots between different information systems much faster than we can as we’re searching for data.
Why does it takes us hours just to be able to pull up the latest business results when we’re looking for our Q2 [second quarter] business performance? Why can’t I just ask a question to Siri and say, “What was our revenue in Q2?” and it spits out the answer? We know that computers can do these things, there just has never been elegant software to be able to connect it all together and deliver that from a user-experience standpoint.
These are the things that we see as being incredibly exciting about where AI is going in the workplace. Then we want to begin to make sure that we can spend our time on the areas that humans are, frankly, way better at than computers and are the more fulfilling parts of our job: the creative tasks, the tasks where we’re collaborating with other people, where we’re creating new products and ideas and not just searching for information or doing things that computers are going to be way better at.
Simon London: Let’s go back to the question I asked around big technology trends and moments being on the right side of history. Building all of these things into the product: These are your bets to be on the right side of history when it comes to AI, right?
Aaron Levie: Correct. We feel like the way that we are going to manage, share, and collaborate our information in three, or five, or ten years from now is going to look unbelievably different than how it has in the past. And AI is going to be at the center of that.
Simon London: The other interesting thing, looking at your AI strategy, is, you’re working with Google. You’re working with IBM. You’re working with Amazon. You’re working with just about everybody, right?
Aaron Levie: Yeah.
Simon London: That raises a question about ecosystems, which again is something that you’ve written about. Just talk to me about ecosystem strategy [Exhibit 4].
Aaron Levie: This is one of the areas that, especially for a lot of incumbent companies, whether they’re in software or whether they’re not digital, companies are going through this evolution. In a world of platforms, you need to interoperate with other platforms. Furthermore, in a world where we are still relatively early on this journey of digitizing the globe, and all of our businesses, how everything around the world works, we’re in a world of non-zero-sum gains and growth in so many different spaces, which means that you have a lot less time to compete. You have to start to collaborate to go after these markets. And so a lot of incumbents are realizing that the competitors that they once had are now going to be their partners in this new era. Because there’s simply no way for any one company to deliver an innovation to solve all of the world’s problems, because of just how much growth and how much opportunity there is.
So our approach to this is no different, which is to say, we are going to be, hopefully, the world’s best product and company at managing, structuring, organizing, and securing information. But it turns out that Google has more engineers than us. And it turns out that they’re going to be spending billions of dollars every year on AI. And IBM is going to be spending billions of dollars every year on AI.
Why would we want to be in a world where we are trying to compete with those players and their R&D budgets, in a space where they’re going to do unbelievably well, instead of flipping the model on its head and taking all of the strength of their technologies and letting our customers leverage those capabilities? That’s the architecture that we’ve built.
The idea is to take advantage of all the unbelievable innovation happening in AI and to start to incorporate and include that in Box, so that customers can turn on any of those services for all of the data that they have inside their platform. It’s not natural for a lot of organizations to be so open and so dependent on other companies and on an ecosystem for their strategy. For us, we were, as you mentioned earlier, born into this new way.
We were born on the right side of this era of ecosystems, and interoperability, and integrations. The kinds of companies, again, whether they’re digital or not digital, that think that they’re going to somehow be able to build all of the technology themselves, they’re going to control all of the architecture themselves, that is a fool’s errand in the digital age. The amount of innovation that is happening from so many companies—if you can’t take advantage of all of that and allow all of that to become a tail wind for your strategy, then you’re just not going to be able to compete with the companies that do take advantage of that. You see some of companies these days that say, “Hey, I really want to move to the cloud, but the best cloud-computing provider might be one of my competitors.” That doesn’t mean that you shouldn’t leverage their technology. They’re going to compete with you whether you like it or not.
One of the most profound examples of this is Netflix, where they run primarily on Amazon web servers. Amazon is one of their most significant competitors, but that didn’t change the fact that architecturally, Amazon is the better cloud-computing provider. That has no bearing on how they’re going to compete with Amazon, which is on a completely different part of their business.
That ability to say, “OK, we’re going to compete in one area; we’re going to complement each other and partner in another area” is what we see as the way organizations are going to have to operate in the digital age, because of how interdependent all of our technologies and all of our products are.
Simon London: Before we run out of time, I want to talk a little bit about business. I also read somewhere that every year you reread Clayton M. Christensen’s book Innovator’s Dilemma: When New Technologies Cause Great Firms to Fail (Harvard Business Review Press, 1997). Is that true?
Aaron Levie: I don’t know if I’ve kept up with it every year. For the sake of accuracy, I would say it is not true. However, more or less, every couple of years, I reread it as a reminder about, again, how important these moments are when you’re thinking through, is that a competitor that’s coming up behind you, is that an actual disruptive threat, or is more of a sustaining innovation that you’re going to be able to adapt to?
Is there something that you’re not responding to in the market because your organization finds it unattractive to respond to that thing? Or are we just not moving fast enough? These are fundamental decisions that teach you when you are at risk of being disrupted. But then, conversely, are there areas where you can be more disruptive to competition? And how do you take advantage of that? That’s why I think it’s always important to be grounded in the core of Innovator’s Dilemma and why it matters.
Simon London: I think it’s a massively wise book. Over the years, Clay and the book have taken a lot of stick because people throw around “disruption” as a buzzword. But actually, if you really read it, particularly the tensions inside companies, smart people in companies often fail at this.
Aaron Levie: I would say 98 percent of the time.
Simon London: And it’s not because they’re stupid or they didn’t see it coming. It’s all to do with incentives, and processes, and how you navigate it.
Aaron Levie: That’s exactly right. It’s funny, because being a digitally born company, you’d think you would know how to avoid this type of issue from happening in your company. But I see it every day, where you can just start to subtly see how one organization’s, or an entire business’s, incentive model is built around something that is now going to leave us in a position to be flanked in one particular area.
How do you mitigate that preemptively, before maybe even the competition arrives in terms of attacking you in that area? But it’s all people. It’s all organization. It’s all how people are incentivized. What are people’s goals? The best position to be in is where you’re using Innovator’s Dilemma as an offensive technique to be able to be disruptive in other markets. But at a minimum, understanding it so deeply [helps you understand] why people are not responding to disruptive threats in their organization. And how you make sure that you can avoid that, from a defensive measure, is also equally important.
Simon London: Well, thank you. This was a wonderfully nerdy, pleasurable conversation.
Aaron Levie: Thank you. Hopefully we don’t lose too many people in your audience after this one. I don’t want the podcast-subscriber count to drop after this.
Simon London: No, this is in our sweet spot. This is great stuff.
Aaron Levie: OK, cool.
Simon London: So Aaron Levie, thank you very much.
Aaron Levie: Thanks a lot. Really good to be on.
