Box's Aaron Levie: Why SaaS still matters | Pioneers of AI

For us, everything's at stake. The key point is probably that other than that very early founding period, maybe the first year or so, it's never felt as much like a startup as it does today. We feel like we're very much back in that early formalization period, mostly because of the size of the opportunity.

Every single day there's a new model release. There's some new breakthrough in how you design and build agents. It really is critical that we are moving again at this very fast-paced way. Everybody working on AI feels that type of pace and intensity right now.

Thanks for having me on.

I got enamored with technology fairly early on. In the early years of high school, I had a couple friends who were building software and internet websites. That piqued my interest quite a bit, so I started building lots of websites. Through that, I created a bunch of projects with various friends, sometimes on my own, sometimes with what would eventually become one of Box's co-founders. I was just trying anything that seemed interesting and a good use of the internet. None of those ideas really panned out. I went to college and tried to explore another passion area, which at the time was film—it still remains an area of interest. I thought maybe I could combine digital technology and media and do something with that. So I got some internships in the entertainment industry. The one you referenced was at Paramount Pictures in 2004. At the exact same time, I had to do a project in college to identify a market opportunity. I chose the idea of online storage through a random set of events: working from a corporation and seeing how sluggish legacy software was, and in school, emailing yourself files to move between computers. A few stars aligned, and I eventually decided to study this space. There had been companies that would let you store your data on the internet, but they all died off in the late '90s. They didn't really innovate; they were zombie products and companies. It felt like this was the moment where you could finally deliver on the promise of putting your data in the cloud—it wasn't called the cloud at that point, but you could put your data online, store it securely, access it from anywhere. Browsers were getting faster, the internet was getting faster, storage was getting cheaper. That was the general thesis we had. We started building that idea out and launched it in early 2005, and it just started growing from there.

Yes, very much so. He gave us that profound advice very early on. We were debating a couple of different business model options and even technical options. We were split between a couple of different approaches. He basically said, as a startup, just don't hedge your bets. It's either going to work or it's not going to work, but it won't work if you try to do both things. From a game theory standpoint, that's the best advice. If you have a startup with very limited focus, attention, and bandwidth, and you try to do two things in opposing directions, you're guaranteed to lose. If you do one of those things, maybe you only have a 50/50 chance depending on which one you chose, but that's your only path to getting a return. Your energy will be split in too many directions. We live by it today. Every time I see a situation where it feels like we're hedging—we can't quite make a decision, so we want to try both things—we channel Mark Cuban. We just fully live by it. That can be different from keeping your options open, so it's important not to confuse the two. We often keep our options open with a flexible technology framework that can move in different directions depending on market trends. But if there's something where you really need to make a fundamental decision, we do not parallel track it. We make a bet and go big on that bet. So far, our hit rate is pretty good, and it's avoided a lot of disasters that would have happened if we had tried to go both directions.

Once we pivoted into the enterprise market, which we did relatively early on, a couple of years into the business, it's a great example of not hedging. We were all in on enterprise. We had a lot of customers asking if we could deploy an on-premises version of the software where they could run it in their own data centers. This was a time when the cloud wasn't 100% obvious. Many enterprises, especially banks, pharma companies, and governments, didn't want a cloud version; they needed to manage and control it themselves. A lot of companies at the time were giving enterprises on-premises systems while building out cloud products because they wanted to win deals without convincing the customer to go to the cloud. We made the bet that architecturally, it was way more important to be fully in the cloud because as new innovations happened—mobile devices or eventually AI—you need all your customers to have access to that technology on day one. When you have customers on on-prem environments managing their own systems, you're at the mercy of their upgrade cycle. And what's funny is that end users end up blaming the vendor, not their internal IT. They think, 'Oh, that Box product isn't very easy to use because it doesn't work on my phone.' We realized we didn't want to be in a situation with fragmented deployments. We needed to be in control of all the new innovation our customers get. So we said we're not going to hedge our bets; we're betting fully on the cloud. It meant we would lose deals along the way—lots of banks, pharma companies, and government agencies. But years later, as our cloud customers grew, they got all the benefits of the most modern features. And many of the companies that said they had to stay on-prem eventually realized they had to go to the cloud anyway. We had a better architecture for them to move into, and the players doing on-prem software were the ones they were moving from. So yes, you had to be much more patient, but if you're right architecturally, you end up in a way better spot assuming you stick to your principles. The general advice for any startup is certainly don't hedge your bets, but absolutely don't change your technical principles. If you're right, customers will eventually move to your side. If you're wrong, the business wouldn't have worked out anyway.

I can occasionally produce a magic trick.

Well, first of all, you're not supposed to know that. I don't know who told you.

That's our little secret. You don't have to tell people this.

Interesting segue. Actually, we do use magic principles in the actual product. There's a lot of stuff you try to do where you need to buy yourself some time in the product experience while the system is working. It's not classic misdirection, but it's ways of making the system thinking process not boring and giving people a real experience. But on the overall space, I'm such a deep enterprise person that my general belief is enterprise AI is where most of the power of AI will be delivered. AI, at the end of the day, for the most part, minus a couple of consumer-oriented categories, is an enterprise industrial technology. It's much more akin to classic industrial automation systems, but for knowledge work. If you saw a gigantic forklift for the first time, you wouldn't immediately think it's a massive consumer market; you'd think it's going to be in warehouses. I think of AI like that. There are fantastic consumer tools—search, finding information, maybe therapy—but the big market, the big TAM, is on the enterprise side. I'm probably a pragmatist in the sense that model progress will continue at an incredible rate, but model progress will always have a capability overhang until you have the workflow systems that take the models and put them into the business process. That part will take longer than some of the most optimistic AI people believe. Change management and enterprise workflows take longer than most realize, and that will cause cognitive dissonance. In Silicon Valley and tech, we'll see incredible breakthroughs and use them quickly, but for the rest of the world to adopt and transform a business process, it will take two, three, five years, maybe a decade. We saw this in the cloud. Within two years of 2007, 2008, 2009, we all knew cloud computing was the future, but it took two decades to actually play out across all businesses. I was with a customer four weeks ago who are the biggest believers in cloud, and they still have massive on-prem systems they're only now starting to think about migrating. That's the kind of change management that happens in large enterprises. So this is going to take a little longer than people think. The good news is, if you're on the entrepreneurial side, this means you have time to build great solutions. If everything could instantly be solved by the model, there wouldn't be much opportunity for entrepreneurs; the labs would have all the value. So it's mostly a good news story; we just have to be a bit more patient with our timelines.