🎧🍌 Aaron Levie | The $1 Trillion AI Startup Opportunity
🎧🍌 Aaron Levie | The $1 Trillion AI Startup Opportunity
Plus stories from the early days of Box, pivoting from consumer to B2B, navigating an IPO, fending off an activist investor, and advice for large companies approaching new technology cycles
The episode is brought to you by Attio
Attio is the next generation of CRM.
It’s powerful, flexible, and easily configurable.
Join Replicate, ElevenLabs, Modal, and more and try it today.
To inquire about sponsoring future episodes, click here.
Aaron Levie is the co-founder and the CEO of Box. He dropped out of college to start Box in 2005, and has since scaled the company to $1B+ in revenue, building through multiple technology waves and platforms shifts.
Our conversation covers all the opportunities in AI, advice for large companies approaching new technology cycles, and the $1 Trillion market he sees for startups. We also talk through his founder journey, including the early days of Box, pivoting from consumer to B2B, turning down an acquisition before IPO, and fending off an activist investor during COVID.
If you don’t have time to listen or read the transcript below, my biggest takeaways:
It’s easy to underestimate the size of new markets. Bringing down the cost and increasing the ubiquity of new technology often dramatically increases the size of new markets. We’re probably under estimating the impact AI will have on the economy and our daily lives over the long-term.
AI will benefit both startups AND incumbents. This is a very rare occurrence, as new technology usually benefits one at the detriment of another.
If your product benefits from multiple users, make it as easy as possible for people to share it.
Always put your company in a position of riding whatever tailwind is in the market. You want to be doing things as much as possible that are not running into just general headwinds. For example, at the dawn of cloud, mobile was a headwind for all on-prem software. Their software wouldn't work on mobile, and they had to re-platform their technologies. Whereas Box just rode this mobile wave.
Timestamps to jump in:
03:11 Why ChatGPT was an iPhone moment
04:37 Advice for large companies incorporating AI
11:16 Why AI will add jobs, not steal them
16:13 How AI is supercharging Box’s products
19:03 AI agents: the $1 trillion opportunity
25:27 Estimating size of new markets
29:58 Starting Box with high school friends
33:18 Living out of their first office
34:52 Why early investors passed on Box
37:24 Pivoting from consumer to B2B
39:53 How Box got its first customers
41:57 Should founders talk to Associates at VC firms?
43:26 How Mamoon at Kleiner saved Box at its Series B
46:25 Turning down an acquisition before IPO
50:51 Why Box’s IPO was so hard
54:11 Fending off an activist investor during COVID
Find Aaron on Twitter and LinkedIn
Transcript
Find transcripts of all prior episodes here.
Turner Novak:
Aaron, how's it going? Welcome to the show.
Aaron Levie:
Good, good. Thanks for having me. Wild times right now, but yeah, glad to be here.
Turner Novak:
Yeah, definitely wild times. You had a comment recently that I thought was super interesting. You've indicated, hinted, you think AI is like an iPhone moment. It could be even bigger than the iPhone. What did you mean by that and why are you excited about AI?
Aaron Levie:
Yeah. So I think, tell me if I'm wrong, I think it was probably around the start of the ChatGPT early craze. It felt like one of these big platform shift moments. It's always imprecise how you compare these things to prior technology events, whether this is the web browser moment, if it's the iPhone moment of a particular technology category. But I think empirically, the moment ChatGPT came on the scene, it really opened up everybody's eyes to what you can now do with AI at a much greater scale.
And we've been thinking deeply about AI for quite some time. But I'd say the classic issue has been, before these very large language models, you basically could solve one narrow AI task at a time. You could do image labeling. You could do autocorrect. You could do type ahead. But you really couldn't solve general intelligence type use cases across a wide variety of domains, spanning legal, and math, and business, and science. And so you really need one of these generalizable horizontal platform shifts for really there to be one of these big breakthrough moments, and I think that's what ChatGPT represented.
And now obviously, it's very obvious that I think AI is creating these new platform opportunities for software companies. And I think you've got this classic issue of which areas do incumbents have a strong advantage in where they can very quickly adapt to this new environment, they can incorporate AI into their products and services without really much of an innovator's dilemma in the process, and then where do you have areas where actually it is a bit of a dilemma? You do have a technology that's going after a core profit stream, serving customers maybe in a better, faster way and where that is relatively disruptive.
We, I don't think, as an industry have totally figured out exactly how this is going to shake out, which is actually what makes it so incredibly exciting, is that in any given day, one day you're thinking Google is going to be disrupted. The next day, you're thinking, wow, Sundar is back and he's actually executing on this. It's not going to be disrupted. And I think that rollercoaster is what makes obviously this moment incredibly exciting for all of us.
Turner Novak:
So then, how do you think if you were, let's pretend you're the CEO of a publicly traded company, let's just pretend, what would you do?
Aaron Levie:
It all depends obviously on what kind of company you're in. But I think those of us in enterprise software, I think AI looks, feels, smells like a technology that makes all of our categories of software just better for customers. And I think there's a little bit less of the stress on the core innovator's dilemma piece because AI just thrives on more data, more business workflows, more user interaction. And so if you have an existing technology where AI can make your software better, that's actually just a great thing for you and for customers. And so the dilemma is less severe.
And so for us, our scenario is the best case scenario, which is AI is incredibly good at reading documents, understanding unstructured information, so we see this as a breakthrough in how we can actually tap into all of the data and insights that an enterprise has inside of their files. And this has been this age-old problem for us for nearly two decades. We've had this challenge which is you create a file, you share the file, you share the file, and you probably don't really go back and look at it ever again for a significant portion of that data.
Now all of a sudden, AI is able to let you go and understand and synthesize that information at scale in a way that we just never could have done before. So it makes all of your data an order of magnitude more valuable for your organization. So that makes us incredibly excited because we can do so much more with information.
I think every enterprise software company is on some continuum of either AI is an extremely massive breakthrough or at least usually incrementally positive. I think there's not as many players that have this as a direct attack on their core business. I think again, there is continuum so maybe if you're on the pure customer support side, maybe this looks like maybe a seat reduction in some of your core software. But other than that, I think enterprise software companies look at this as a huge opportunity.
On the consumer side, it could be a little bit different. There's obviously an element to hey, does this change how we search for information, how we find things. You can just see Zuck for instance is like, "Hey, I'm going to put a search box in every one of my products in the form of an AI chat experience." So he's clearly using this as an opportunity to get much more into that part of the business. So what's great is that nothing is a given at the moment in this landscape that we can fully rely on because things are changing so quickly.
Turner Novak:
In terms of maybe what you guys are doing at Box, how do you shift a company that's new technology? You're doing over $1 billion in revenue. I would say you're a mature business. Maybe you still like to act like a startup as much as you can, but you're a very large company.
What do you have to do internally to shift everyone to just take advantage of these different technology shifts that come about while you already have a mature business?
Aaron Levie:
Yeah. Again, it's a little bit, I think it all depends on if you feel like this new technology is something you can fully take advantage of and it's a tailwind for you versus oh, man, this could be a headwind with pivot. And we've been in both modes to be clear.
AI is one that it feels a bit more of you're more riding a tailwind as opposed to having to dramatically cause the company to understand how severe the disruption element could be, which is frankly 10 times more fun because then you're not spending as much time convincing people of why something is important as much as showing what's now possible, and that becomes a propelling force on its own.
So for us, I think like any company, the moment we saw that we had these generalizable AI models, we basically got into a room and we said, "Okay. How is this going to impact us? Let's put together a tiger team." That tiger team starts relatively small, you start to work on architecture, then it actually becomes a much more concretely funded team in the next planning cycle and then you expand its resources, and then a year later, it's like a core part of your building blocks as a platform. And so that's how it's evolved over the past year.
We fortunately, I think this is pretty important, we fortunately have a number of people in the company that have been at startups. They've seen how quickly these spaces move. They understand you have to hop on these moments very quickly. And so I think we have a strong set of folks that knew, hey, this is one of those moments where you basically just have to bet that this is going to accelerate way faster than you realize.
So let's make sure that we have a dedicated effort on this and that we're fully resourcing it right out of the gate. So this was a little bit less difficult for us because of just the sheer upside that we saw with AI.
Turner Novak:
Because you can basically just go to your customers and say, "Oh, by the way, here's all these cool new things you can do and you don't have to change anything. We're just giving you more features."
Aaron Levie:
Yeah. It's like 90% that pitch. They have other things they have to deal with and there's a cost implication or whatnot.
But again, if you have data, if you have workflow and if you have users, AI in a lot of categories, not all categories, but in many categories, will become just an incrementally positive capability for your product.
Turner Novak:
Yeah. And when you think about incrementally positive, do you ever see ... Maybe how would you advise people to not get tripped up over AI will steal people's jobs? It seems like you're very pro. I've heard you say and read a lot of your stuff. You're like, "It actually is going to add more jobs."
How would you talk through that and just convince them on, hey, AI is actually very good for the labor force?
Aaron Levie:
Well, my general thought experiment is if you look at most areas of a business and you said if you made that part of your business more productive by, name a percentage, 10%, 20%, 50%, 100%, it doesn't matter the number, what would you do as a result of that productivity gain? Just think about it from purely a business owner standpoint.
So the example that I threw out there a few weeks ago was if you made an engineer 50% more productive, which is the extreme end of what anybody is seeing from an AI standpoint, so it's purposely-
Turner Novak:
Like worst case or best case scenario?
Aaron Levie:
The worst case in terms of, oh my god, that could reduce jobs. And best case in terms of wow, that's impactful for now and maybe it'll be 300% in five years from now.
But let's just say best case/worst case, 50% increase in productivity. I don't know that many companies where if they got a 50% gain in engineering productivity wouldn't immediately want to then go build more features faster. Very few companies would accept I can now build 50% more features faster and so thus, what I really like is my current pace of innovation, so what I really want to do is now take those savings. I don't want to deliver any more features. I just want to reduce the spend in engineering because now my engineers are 50% more productive. I don't know of many companies.
I'm sure there are some that are in very static industries that I'm sure that will exist, but the vast majority of the world is under-resourced in a lot of these highly creative, highly innovative functions of their business. Engineering being an important one in that sense.
Where if you've got a 50% increase in engineering productivity, well, first of all, ideally, what that's meaning is that you're building features your customers want. That should turn into revenue or else, you're doing something wrong. And with that productivity gain, most organizations would say, I'd like to actually build even more features now faster. So that's one I think way to think about it.
The other way to think about it is let's say you were the one company that said, "You know what? I'd like to bank these savings and my engineers now are more productive. I'd like to bank the profitability." That implies that you're also in somewhat of a static competitive market where somebody else isn't going to take that productivity and turn that into a competitive force for them, which then again causes you to have to go build more features or solve more problems.
So if you imagine a complete zero sum market that was totally static as a space, it could be reductive from the job standpoint, but basically there's very few categories that actually look like that. We're all in dynamic industries. We're all constantly competing. So the idea of me getting a productivity gain is going to usually lead to me wanting to go build more functionality and thus ultimately need to hire more engineers.
Take sales as an example. Sales is even better I think one to make a point. Any organization that can make their sales team 10% or 20% or 50% more productive, which basically the simple math is a dollar goes into the sales rep's salary or expense, how many dollars do you get back?
So if I can make that ratio more productive or the amount of total revenue that they generate 10% or 20% higher on the same cost basis, what we're going to do as a company is not then define with the total number of sales reps we have and just basically bank that 10, 20% productivity. We're going to now hire as many sales reps as we can that will continue to deliver that same level of productivity before we start to see diminishing returns again. And there's basically no company on the planet that wouldn't make that exact same decision.
So the more productivity we get in many of these parts of our organization, the more hiring we're going to do in those areas. Now there are some examples that I think you could at least debate around the edges like okay, what do you do in customer support? If we could deflect 30 or 50% of our inbound customer support inquiries, what do you do? I'm actually still going to take the case that a significant portion of those dollars of savings will be reinvested in things like, well hey, what if we actually go and do more outbound customer success with our customers? What if we get much more tailored and personalized? What if we actually get closer to our customers?
But the idea that we want to be spending so much time doing the more basic level of customer interaction versus using those dollars to actually do much more high-touch customer success, it's crazy. We'd love to be able to actually go spend more time with our customers, not spend time on password reset requests.
So I think there's a lot of areas where again, if we can get AI-driven productivity gains, that's going to mean reinvestment back into the company and back into growth areas of the business, which then ultimately then creates more jobs.
Turner Novak:
Yeah. So in terms of Box, what you're doing, what are you comfortable talking about? Some of the internal stuff. Maybe initiatives inside the company, products, anything specifically you're most excited about right now?
Aaron Levie:
I'll throw out just the general thing that we're thinking about and then I think this even expands to probably just startups in general.
So for us, again, it's a really simple idea. There's a lot of data inside of Box. Obviously, that's driven by enterprises storing their contracts, new marketing assets and financial records and HR files. And inside of all that information are important insights.
And also, that's data that largely you have not been able to automate the workflows around in the past because it's unstructured data which to a computer looks like a blob of text. So it's hard to automate things that are just these blobs of text.
But with AI, we can actually understand what's in that text. We can understand what's in that image. We can understand what's in that video, and we can automate business processes around it.
So we see AI really as an accelerant to automate workflows. Pull out insights from information, better secure our data, and that's what we're super excited about.
I think the medium to long-term approach that this then takes as a characteristic is what are the kinds of things that people do with content but where there's simply not enough people in the world to do those things for most organizations?
So again, you go to most companies and you say, "Hey, how are you managing your contracts?" It's a bunch of data inside of a bunch of folders and email attachments and they don't have the ability to say, "Hey, what are exactly the five customers that have this particular type of clause in their contract where they're up for renewal in the next week that I need to make sure I get ahead of?" Very few companies actually have those types of insights, and it's because they have not had the human capacity to go read through all their documents, go label all the information in those documents and then build a system for actually managing that information.
And you can apply that to literally every category of business. It doesn't have to be contracts. It could be anything. So imagine if we have AI that basically can do that for you. Now, first of all, back to the prior point, there's very few jobs in the world that were ever spent on those things because only basically, the largest companies in the planet ever actually belabored at that problem. Most companies just don't have that information.
So the first thing is that AI is now just purely accretive. It's not taking any job from what we were doing previously, but it is doing the task that a human in a perfect world of abundant resources would have been doing.
So the thing that we're really excited about is what happens when you can create effectively AI agents that can go and do things for your data and your information that were just never possible to gamble for? Help me analyze this information, help me structure this data, automate this workflow. That becomes this huge supercharger activity for most organizations and for most business processes as an enterprise.
Turner Novak:
So if I'm hearing this word AI agent for the very first time, can you just give an example, a pretty down the fairway example of what that might be just in the context of Box?
Aaron Levie:
In our context, there's what we think of right now in most of AI is AI is this assistant where I ask a question, it gives me back an answer. And so that's ChatGPT, that's Siri, that's Alexa, that's the more assistant functionality.
Where we obviously, I think as an industry collectively, are now talking about AI going is not just when it serves back an answer or information, but actually, when AI goes and does something for you. And that can be an extremely big task like, hey, go book a flight and plan the whole vacation or whatnot. I think we're still years from probably wanting or even trust AI to go do that, but that's a farther out example to something just as simple as, hey, do one decision or take one action for me as an AI agent.
So it's really moving AI from being this read experience where I'm asking a question and basically, I get a read-only view of data, do something that's a little bit closer to a read-write operation, which is okay, it actually can take an action, it can take a step, it can move something forward.
And so in our world, I can either ask a document a question like, hey, document, can you summarize this or what's the answer to this question? That's basically a read-only experience, to something that would be more agent driven, which is, hey, I'd like you to actually take out the metadata from this document, structure this contract and then kick off a workflow based on certain data that's inside the contract.
So one is I'm asking one question, getting an answer. The other is I'm saying, "Can you go this thing," and an actual business operation happens. The latter is probably a 10x or 100x even chains and the kinds of things now you can do with your data because again, you wouldn't have been able to throw human labor at that problem before, but now as tokens come down in cost, we can just scale up intellectual labor and knowledge work or imagine more than we could have with people.
Turner Novak:
So it's like your example of searching for contract renewals. You can basically say, "Okay, agent. Email all these customers. Set up meetings. Send them contract renewal. Give them updates on new features." You can go down the list of what would an actual human probably do and just make it easier for the sales team to do all that stuff and just kick it off and make it be even more efficient.
Aaron Levie:
Yeah. I had to parse through the individual examples. Some of those we probably wouldn't do. We would hand that off to somebody else and then some, we would more directly do. We're so focused on just the content side that we want to focus on the things that you do just to the content. What can you actually have an agent now do? As you get to actually now reaching out to the customer or whatnot, that would probably be the role of the sales force agent or some other part of the ecosystem. But I think in general directionally, that's where we see AI going.
The thing that I find incredibly exciting though is if you think about let's say a year ago, a year and a half ago, the general pattern of an AI startup was a bit more in this Copilot direction, which is this is an AI service that it's going to sit alongside me and give me insights and information as I do my work. Very powerful. Unquestionably a productivity gain and improvement. I think a lot of the value were true to the incumbents that already have the workflow, already have that user in their interface, but still very powerful, and there will be startups that emerge that build incredible products and use cases there.
In the past couple of quarters, I've seen a pretty healthy increase in more of these agent-oriented startups. And the reason why that's so exciting is because I think we're starting to see a little bit of, for lack of a better word, probably just a template for what that startup looks like. And it seems to look like a startup realizing that a person or a company has to hire some function in their business, I need an outbound sales rep, I need a security engineer, I need a DevOps engineer, I need a legal support person. And what they're basically doing is building software that is a vertically integrated version of what that work used to be on the human side.
And that's a breakthrough because if I can now as a startup hire effectively software to do work for me, I can hire it to generate leads. I can hire software to review legal contracts. I can hire software to keep my infrastructure up.
All of a sudden, this is game changing for basically businesses of all sizes, but especially smaller businesses because you've always been in this massive talent asymmetry war with big incumbents, which is the big incumbents have all the resources. You start with basically nothing. You're doing all of the jobs and all of the tasks. And now all of a sudden, AI basically becomes this massive point of leverage where I can now actually generate app on leads, review the legal documents, actually protect my enterprise, and in the form of now this AI laborer that's able to go do that work.
So I think this is going to be incredible because there's not a lot of software incumbents in these markets, so you have a lot of greenfield opportunities, and you are starting to see a pattern emerge of how do you actually build these types of technologies? And that's I think going to be ... that would be ... If I had to close my eyes and bet on the AI market that produces the next trillion dollars of in aggregate startups, it would be an agent for all of these domain specific vertical parts of our business.
Turner Novak:
Fascinating. Okay. Yeah. It's interesting when you think about AI has a ton of benefits for incumbents like you were just talking about. If you're a big company, lots of data, lots of workflows, it's great. But then if you're a small business, you also benefit from using them.
It reminds me of cloud where the big companies definitely benefited, but Uber probably wouldn't have existed without the cloud. Could Facebook's mobile app, which now benefits every small business in the world with advertising? That might have not existed if the cloud wasn't a thing.
All these different technologies, there's different value that accrues at different places. I feel like AI is super interesting where it's almost like both, like everyone, all businesses can benefit if you leverage it right.
Aaron Levie:
Yeah. In the early days of cloud, let's say between early 2000s to even late 2000s, we probably all selectively dramatically underestimated the second derivative effects of having infrastructure and software that was 10 times cheaper and a hundred times more ubiquitous and democratized. Because initially, you usually measure a market off of the existing market and you say, "Okay." It's hard to imagine anything more than any market doubling or something. And so you're like, "Okay, well maybe the market could be two times larger."
But then there's this other factor which is like, well, if it's actually cheaper in the cloud, maybe actually the market is smaller. If you say, "What's the global IT spend in 2005?" And then you have cloud computing, but then you might be like, wow, well, cloud computing.
What you're doing with cloud computing is you're creating shared resources for a large number of customers, which means all of that excess wasted capacity doesn't exist anymore. That is in everybody's data centers.
So you could easily squint and be like, "Maybe actually the market for cloud is half the size of the on-prem market." And so that's like your initial thinking. You're analyzing the market size or whatever, but what you forget is actually, well, when you actually make something 10 times cheaper or 10 times more available, what happens is enough generations later of startup formation and ubiquitous technology, all of a sudden, you have so many more customers that never were in the market for buying the massive data center footprint.
You have the two-person startup, the Instagram that gets created and all of a sudden, they start scaling and needing infrastructure that they would never have been in the market to buy data center equipment previously. And so you multiply that up by just the scale of the economy and you actually in many cases have something like a 10X increase in the amount of then the scale of these markets.
So think about somebody trying, Tobi talked about this lot, but think about trying to size the e-commerce software market before Shopify existed. You'd be like, "Okay, maybe there's 100,000 companies that need to sell things online. What's the upper limit of this kind of business?" It just turns out, no, actually, you give people really powerful, simple e-commerce technology, you'll have tens of millions of businesses that get created as a result of you lowering the cost of this technology. And so that has been true. AWS dramatically increased the total scale of how we're using infrastructure. Shopify did that for e-commerce. Stripe did that for developers. Twilio did that for communications technology.
And so that's what we basically saw in the early days of SaaS. And now, I think you can almost create an analogy to what AI would do, which is you would basically be scaling intellectual knowledge work for the masses that previously were not actually utilizing those kinds of services and, again, on the human labor side.
Turner Novak:
So I'm super interested when we're talking about just cloud in general. That's how Box got started. I think 2003, you wrote a paper for school?
Aaron Levie:
I basically wrote the paper in 2004 and then we launched in 2005.
Turner Novak:
So what was the paper? What was the thesis?
Aaron Levie:
Well, it wasn't a paper for Box. It was a paper for, I had to analyze a market to basically. I don't remember if it was specifically a SWOT analysis but had to do some analysis of some business category. I don't know how I got to this, but I chose online storage, which is like-
Turner Novak:
Why the heck would you pick that?
Aaron Levie:
It's the weirdest thing of all possible markets. You're supposed to do shoes or what's the market for retail or something in America? I chose literally online storage, and that was a two-fer because I got to then research the actual space.
And as I was researching the space, I was like, "Wait a second. This is crazy." All of the companies in this market are charging way too much money. They hadn't updated their products. They hadn't responded to many of the technology trends that were actually happening at the time. They didn't give you more storage. They had really poor user interfaces. They didn't work on mobile devices.
So in the process of basically doing this class project, a flywheel developed and it became just so obvious that this is crazy that there's this big opportunity out there. Someone is going to seize it. So then I started to develop the initial idea to actually go and take advantage of that opening.
Turner Novak:
Nice. I knew you basically had a roster of some friends that you'd known for a while. It sounds like it took you a little bit, like a year or two for everybody to come together and decide on this.
Aaron Levie:
I couldn't convince everybody at the start, but basically, I had some friends from high school that we had done startups in different business ideas with. Everybody had gone their separate ways for college and things, and so I had been brainstorming ideas with that group.
And then Box was another one of the things I was brainstorming, and then eventually, everybody got bought in and successfully dropped out over about a year period. But as we scaled up, we all dropped out and moved to the Bay Area and then focused on Box full-time.
Turner Novak:
How did you decide to drop out, and then how did you decide to move to the Bay Area? What were those thought processes?
Aaron Levie:
Probably kind of one and the same decision. So what was happening was we were getting enough users, enough usage. At the time, this was now 2005, middle of 2005, fall of 2005. We had just raised some funding. Mark Cuban invested in the company. That got us super excited. That was a huge endorsement to us.
Turner Novak:
And that was before you moved, right? You moved right at the same time?
Aaron Levie:
Yeah. The summer of 2005, we all worked together ... Oh, sorry. Dylan, who's our CFO together in Seattle. And that was the first instantiation of Box really as a business. We had been remotely on different coasts and got together in Seattle where we all grew up, started scaling things up. We raised a little bit of money. Mark Cuban invested later in the summer, early fall. That gave us enough of an indication that, okay, maybe there's something more real here if maybe it has that endorsement.
And then got to a point where honestly, most of my day, and true for Dylan as well, was just spent on Box-related stuff. As simple as just customer emails during the middle of class that I had to answer, the amount of customer support stuff of just the service cost $2.99 and somebody would call and say, "Hey, this didn't work, or I need a refund, or I have this issue or whatever."
And just most of my day was just responding to those types of things, building new features, trying to figure out how to market it more. So it got to a point where really there's a juncture which is I'm going to totally fail in school, just guaranteed, just literally will flunk out or-
Turner Novak:
So why am I still doing that?
Aaron Levie:
Why am I still doing that? And there's this moment right now where the internet was starting to talk about this thing called Web 2.0 and there was this moment which was like, "Hey, something is happening on the internet that feels like '96, '97, '98, a thing is happening." It was this compelling pull of maybe you only get these windows every decade or so. And so when you have one of these windows where there's a big opportunity, there's a tailwind in the market, there's this push for a new web essentially, you got to just grab that and run.
So while we decided to drop out, and it was pretty obvious then at that point that where you do that is the Bay Area. So that decision was quite straightforward. I had an uncle that had some real estate in Berkeley, so he let us live there, and then we got, again, two of our other friends to drop out as well and then went from there.
Turner Novak:
Yeah. And I read in that house you stayed in, you had the boardroom on the second floor and you were hosting big companies you're trying to do deals with and play it as professionally as you could in that house. How did you pull that off?
Aaron Levie:
Yeah. So we went from Berkeley to Palo Alto, and then Palo Alto was this ... I don't even know what it's zoned as, but it's absolutely not zoned as a residential. But the four of us lived there and it was our office. We scaled that to about 20 people or so.
Turner Novak:
Out of the house?
Aaron Levie:
It wasn't a house. So we went from literally a garage in Berkeley to Palo Alto. It was basically one of these buildings on El Camino, but it had two garages and so we retrofitted the garages for having bedrooms basically. So if you just walked into our office, you'd be like, "Oh, cool startup." If you came through the garage, you'd be like, "What a weird fucking place. Why are there beds on the floor?"
Turner Novak:
Okay.
Aaron Levie:
So it was business in the front, party in the back dynamic. So we were juggling everything. We wanted to save as much money as possible. We didn't want to have to have a commute. We didn't want to pay for rent.
So that's basically why we did that. And we scrounged every dollar we could just to focus on growth. Horrible living conditions, but you do what you had to do at the time. This was a very different era for startup funding. I don't think I'd ever do it again, but again, we had to make do with what we had at the time.
Turner Novak:
Yeah. And I think you didn't have the most easiest luck early on. Obviously, you did end up raising money, but why did people not invest originally?
Aaron Levie:
Oh, yeah. We certainly had no luck, for sure. Actually, to be fair, there were some coincidences that maybe if you think about them as luck than that worked out.
For instance, we were at a TechCrunch party in 2006. We ran into Emily Melton, who introduced us to DFJ, and she brought us in. And then Josh Stein invested, which was our Series A. So there were some of these kinds of moments that ended up working out very favorably, but even then, it was $1.5 million Series A. For anybody-
Turner Novak:
That's a small-
Aaron Levie:
For anybody listening, very different times. The Series A was you sold a fourth of your company for a million and a half dollars.
Turner Novak:
Wow, that's crazy.
Aaron Levie:
I appreciate the bootcamp that created and it hardens you very early on. But at that time, first of all, you had definitely, I don't know the actual number, but I'd probably say I'm sure a fifth as much or a 10th as much in venture capital overall as an asset class.
So there were fewer dollars to work with. Nobody had yet really seen what is this new cohort of web software going to return. You had Facebook, which obviously was going gangbusters but-
Turner Novak:
It was a consumer business though. There's no enterprise software.
Aaron Levie:
Yeah. It was a consumer business so obviously, totally different. And nobody had really made that much money in software as a service. Salesforce was the best example, but probably at the time that I'm talking about, I'm going to make this up also, I'm sure Salesforce's valuation was in the eight, nine-figure range at that point so it's not like it was a billion-dollar company. It was probably hundreds of millions of dollars. We should look this up. Or maybe it was a billion dollars or something.
But the fact that the largest company at the time in the enterprise software in the cloud was at best a low billion-dollar company, it wasn't there was a lot of people who could underwrite in terms of where software was going.
So you take that and then you compound the fact that we were doing a pivot, so we were going from consumer to business. All the founders were under 22 at that time. It wasn't exactly the team you would think of as like, okay, these people are going to go build an enterprise software company. I don't think I would have invested based on those conditions. And so I think we were lucky to have gotten any investors, but the early days were wild.
Turner Novak:
Yeah. Do you remember what that was like then when you were pivoting? I know we haven't really talked about this yet, but you started as a consumer storage business and then now you're enterprise storage. What was that whole process like and why did you decide to pivot?
Aaron Levie:
One probably important maybe definitional thing is I think we weren't consumer as much as we were just agnostic. We didn't think about it in terms of consumers or businesses. We just basically just thought about it as, "Hey, we got storage online. Use it however you want."
Now, what that led to was a bit of schizophrenia where basically, you had consumers asking you for, "I need better features for sharing photos and backing up my family pictures and my computers, and I want you to only charge no more than $5 from that." And on the other hand, you would talk to businesses that were like, "I need the most military grade data security and compliance and I need it to scale to a million documents or 10 million documents and oh, by the way, I'll pay you $10,000."
And so eventually, it became this IQ test, which is like, do you want to go chase the thing that actually seems hard to differentiate on, probably will be out-competed by the big tech companies at some point and at best, at the end of that prize is you get $5? Or do you want to go do the thing where you actually have a lot of room for differentiation? There really is a disruptive dynamic happening in software where you can build something that's cheaper, more scalable, faster than the incumbents. And the prize at the end is you could do six and seven-figure deals with customers really at scale. There's tens and hundreds of thousands of these customers.
So I actually took the longest of the founding team to get religion on this. It eventually got to a point where people were just literally like, "Aaron, you have to change the business model." And eventually, I got convinced somewhere in the '07 range, and we basically pivoted the company.
We effectively burned the boats and just said, "We are an enterprise software company. We're going to do a unique twist, which is we'll be freemium, so you can always adopt us outside of the ... in your prosumer life, you can sign up and use it for free."
And we used that as basically a viral engine to get into enterprises. But other than the freemium component, it let us be 100% focused on the enterprise market with not a single distraction of any other feature we would need to build, any other market or customer we wanted to serve. It was just let's be the best enterprise platform for managing data.
Turner Novak:
Yeah. And I think I remember just reading around this time, I forget the exact number, but it just had millions of users. A lot of people using it. How did you market this thing? You said that was a big focus. So how did you get people to start using this way back in the day?
Aaron Levie:
Freemium did a lot of the work. So we were one of the very few services at the time that gave you literally gigabytes of free storage online. And our angle of attack was if you are a knowledge worker in any company of any size or a small business and you are running into an issue with your current IT services that you're given for managing or sharing data, maybe you'll hear from somebody or somebody will share something with you or you'll go look for something. Let's make it as easy as possible to get you to choose us versus anything else.
And then quickly create evidence to that organization that you have a problem in your enterprise, which is people don't want to use the traditional software. Now, eventually that became a playbook that I think was widely used, Yammer, Slack, Zoom, etc., which is like, how do you democratize technologies so the end user becomes your viral engine? And we really tried to hone that strategy, and it worked. We got tens of millions of users from that approach.
Turner Novak:
Was there any specific product decisions? Was it like a PLG, someone would share a link and someone else would sign up? Was it that simple or anything specific?
Aaron Levie:
It's that simple. Yeah. There's probably no cool lessons other than just honestly, if you're in a category where any part of your product benefits from multiple people using it, then you should tune your product just religiously to make it easy to share, easy to go viral, easy to spread. And so any multiplayer product, you need to be honing in on what is the best user experience that causes that to happen.
Turner Novak:
That makes sense. One question that I had that I don't want to miss because I think we're going criminologically right now. We're hitting all these phases. There's an interesting debate on Twitter. It's probably going to be about two or three weeks ago, considering when this will come out.
It was this associate versus partner debate. You shouldn't talk to associates at venture firms. It actually sounds like you had a pretty pivotal moment with an associate at a venture firm for Box. Can you talk about that?
Aaron Levie:
I understand the essence of the debate, which is for a founder, your time is very scarce. You are in live or die moments all day long. It's hard to even fully express this, but any founder obviously knows what I'm talking about. There are single days where multiple times in the day, you either simultaneously think the company is the coolest thing of all time or you might fail and you're going to have to go get a job. In the same day, you might have that three or four times in rotation.
So if you think about that level of stress, every single moment matters. Every decision you make matters. All of your time, it needs to be allocated as appropriately as possible. And so I think the essence of this is make sure you're spending time in the highest probability ways for company survival and company success.
So I believe at least, is the origin of this debate of, okay, the associate role of a VC firm, what is their ability to get a deal done and should you spend your time there versus going to principals?
Just I guess two elements for me. One, we empirically know that associates can get deals done. We got rejected by, for our Series B, we pitched basically every firm in the Valley and we got candidly rejected by basically every firm in the Valley other than one; it was USVP. And it was because an associate, at the time, Mamoon, for whatever reason, had this belief that we could actually go and disrupt the enterprise software market for managing data in organizations.
And at least from what I understand, he had to really push the deal through the organization because there was a lot of reasons to not invest, and for the same reason that basically every other firm didn't invest. And so Mamoon basically caused our deal to happen. He was an associate.
Now, many years later, he's one of the core leads of Kleiner and invested in Slack, and Figma, and Intercom, and dozens of other obviously top companies. But he fought for the deal. He got it done. He basically saved our business. I think the story would be completely different if he hadn't actually been there and invested in us.
And so to me, I'm sure people have had bad experiences with certain kinds of investors, but I think it's true up and down the stack. I don't think that there's any perfect path to getting a fundraiser done.
And my view in general is from the VC side, I feel like associates serve a very important function. Which is, you have to have a training ground for who in your team is going to become the next partner, the next principal, the next investor. And so if you can't have them go out and source deals and make decisions and push deals forward, then by definition, we don't really have a model that would grow the next crop of VCs.
So I think that that role serves an important function in this type of business category. And again, I think empirically, they get deals done. So maybe there's certain firms where that's just objectively not true, and so maybe you should avoid those firms, but I think it'd be pretty hard to dismiss it as an entire class.
Turner Novak:
Yeah, that makes sense. From what I read, it sounds like he basically just tried every single product and he was like, "Box has the best product. They're going to win," which I think the number was literally 49 competitors. That's a lot of work. So mad props of like-
Aaron Levie:
And that is the work that an associate does. Now, interestingly, now do the reverse. So theoretically, there might be 20 companies out there that also met with Mamoon that feel like, "Wow, an associate burned us or something."
But that's just literally, that is the challenge of venture capital. There's a certain number of dollars. There's a large number of companies. Many times, you'll get rejected. I wouldn't ascribe that rejection to, or the wasting of the time, to it being an associate. I don't know. Have a better pitch. Build a better product. Find a better associate. There's many, many paths to glory here.
Turner Novak:
Yeah. Actually, I read that you considered selling Box. I don't know how serious this was, but maybe you had an offer from someone that you turned down somewhat around that time, maybe it was a little bit after, and it was a very, very significant decision. Can you talk about that?
Aaron Levie:
Yeah. So it was a couple of years after this particular period. We actually had some better fortune. Growth was a lot faster. We were way post pivot. We had some accelerants due to cloud and mobile. The iPad really was an extra nudge of growth. The iPhone and the enterprise helped quite a bit.
Turner Novak:
And this was basically just more access points for more people to need to use data, and no one had mobile tools, all the incumbents. There's no way to even access your Google Drive or SharePoint, whatever.
Aaron Levie:
Well, there was no Google Drive. This is how old I am. We're talking about a technology stack that 90% of your listeners will never have touched. So basically, these are systems that only worked in your data center and your iPhone didn't connect to them. Your iPad didn't connect to them. And we were just an app that worked seamlessly on those platforms.
Turner Novak:
So it was beautiful then. It was like somebody used the product and you're like, "Oh my, god. This is incredible. I don't need my computer."
Aaron Levie:
Yeah. Again, today, that sounds very quaint as like, wow, there was a time where that happened? But that's just the case.
As a slight bookmark, I think another general startup lesson is just always put your startup in a position of writing whatever tailwind is in the market and you want to be doing things as much as possible that are not running into just general headwinds.
And so everybody that did on-prem software, mobile was a headwind. Their software wouldn't work on mobile, and they had to re-platform their technologies. Whereas for us, we just rode this mobile wave.
And so when you're thinking about either strategy decisions or architecture decisions, always be sure that you're in a position where you can ride the tailwind as opposed to having to run up against it.
But basically, yeah, so we were growing quickly. There was an offer to buy the company. It was extremely stressful because it was a relatively large number, I think, especially even just given our scale at the time. And so it would have been a life-changing decision. It would have been a great win for obviously early VCs. It would have been a fantastic win, I think, for everybody involved.
So we just got to a point where we had to compare. Do you take the bird in hand or do you swing for the fences? Not even necessarily because of the financial impact because again, past a certain threshold, that's not going to matter as much. It's more about the total impact you could have in building a much larger company and what actually you can go and build generally that basically unfortunately would have likely gone away. That potential would have gone away if we've been acquired just due to the natural forces of you getting incorporated into a bigger company, the team dissipates and you never actually accomplish that vision.
It's funny. Zuck was interviewed I think just a week or two ago about the similar moment with Yahoo and Facebook. And I think to a tee, our feelings were exactly the same, which is you start to, one ... Everybody says, "Oh my, god. Think about how much money you'd have or whatnot." Then you literally, fine, you get this check, then let's say it's the next Monday. You still have to do something. Literally, you're still in the world doing something with your time and so then you start to think. People are like, "Well, imagine all the things you could do." And then very quickly, you come back to like, "Well, I just want to do software again, and I'd probably need to go find a way to have a software platform with lots of users with a lot of exciting things happening around it so I can build amazing products." And you're basically back to where you started. So why not just skip past that and not sell the company?
Now, of course, you have to make sure that corresponds with a financial argument as well. So fortunately, we felt like, okay, there is actually a financial argument that is Box can be 10 times larger than this and that would be a greater return. And lo and behold, we're eight times larger than that at this point. But basically, we had to equally have a corresponding financial argument, the two combined needed. Eventually, I think a clear decision, even though it was very stressful and there was a lot of angst in the process.
Turner Novak:
Yeah. And then I remember you IPO'd at some point. I don't know where we are in the story, but it was eventful. Well, can you take us inside just how the IPO went?
Aaron Levie:
Well, it was eventful for a couple of reasons because there was definitely a very intense IPO process that was prolonged. We really subscribed to this idea that when you have software that has a proven product market set and you've got strong economics, like let's say 75, 80% gross margin, decent LTV, 130% net retention rate. Those happen to be our metrics. But let's say any of those things, plus or minus five or 10% or higher obviously is better. If you have that configuration, probably what you should be doing is spending as much money as possible to grow.
Turner Novak:
Yeah. To the point of the AI earlier, it's like if you've got a good business, the more you spend, the faster you grow, the better the business becomes.
Aaron Levie:
The faster you grow, but importantly, the faster you grow on a compounding asset, which is very different from growing really quickly. If you're Groupon or something where you just know that over time ... Groupon is just a random example.
Basically a business where there's a K-curve on each customer you require, then you just are constantly refilling the top of the funnel and then it's going to one of the funnel. So at least enterprise software, if you're doing your job right and you have a product that works, ideally you're stacking this residual income on top of each other and it stays with you for a long time and it's high gross margin, so that means it's contribution margin positive over time. Then basically, you should be building the largest annuity stream you can, which means either hiring sales reps, building out more features and engineering and scaling the organization.
So that was the playbook we did. We raised money according to that playbook, and then we spent that money according to that playbook, and we got the hyper growth period of we were doubling basically every year at scale. And the one challenge was when we finally revealed that in the form of an SEC filing, everybody was like, "Whoa, whoa, whoa, whoa. We didn't know you were spending that much money. We thought maybe there was something that was more profitable in the near term." Because we were literally spending more money on sales and marketing than our total revenue, which is just ... It was aggressive, and it was because we wanted to get really big really quickly so we could have this dominant position.
So we did a filing. The market basically told us, pre-IPO, "Hey, this is not the right financial structure to have." There was a temporary correction in the SaaS markets where SaaS went down by 15% or something in a matter of weeks. So we were like, "Okay, let's pause. Let's get a bit more efficient." And so we basically took another nine to 12 months, refiled with better metrics and then we eventually got the IPO done.
But it was a wild journey, and there's a lot of drama at the time because we stayed on file. The SEC then doesn't let you really talk about the business publicly while you're on file. There's a lot of press stories that we then couldn't respond to, but again, eventually got through the journey and survived to tell the tale.
Turner Novak:
Yeah. You had another pretty amazing story of just survival. You had, I think this was throughout COVID, an activist investor that's traditionally very successful, basically tried to take control of the company and make a lot of changes. And I think you were the first to beat them in nine years.
What happened and how did you do it?
Aaron Levie:
We got to a point as a public company where growth was slowing at not an aggressive rate, but over time you can extrapolate the slowdown. We weren't profitable. We were basically breakeven. And you got to a point where if your growth is slowing and you're breakeven, it's a hard story for investors. How do you make those numbers work?
And so we, as a business and as a board, we're talking about that pretty frequently of like, "Okay, we probably need to change the trajectory. We either need to grow way faster or we need to get profitable." So that was starting to happen as a dialogue at the board and we were starting to make decisions around that.
And through that period, a few investors showed up and also gave us that feedback. Now, some did it in a way that was like, "Hey, we'd love to talk to you folks and encourage you to push more on profitability." And those were great conversations. And then one investor in particular, the firm you're referring to, came in and said, "Hey, we need you to do this a lot faster and we're going to ratchet up the pressure in the form of more of an activist type of campaign."
And initially, we were pretty aligned with how they approached it. We brought on some great board members as a settlements offer. Those board members have actually been incredibly fantastic for the company and helped us continue to scale.
But we eventually got to a point where the things they wanted us to do were either too short-term oriented or things that frankly we were already doing and thus, didn't feel the need for more involvement from them. And so we had to get to a point where unfortunately, it was then more of a shareholder decision of, do you want to go with their route and their approach or do you want to go with our route and our approach? And we made the case, and investors basically voted for it, our particular slate of board members.
But yeah, it was a dramatic year-long process because you're constantly in the news for no reason that you want to be. The activist was putting out press releases and trying to drum up lots of drama, whereas we wanted to just focus on execution, improving the business, and we were actually more or less aligned on most dimensions on the ultimate outcome, which is we want Box to be more profitable. We want to re-accelerate growth. Here's our view of how we do that. We were obviously closer to the business, so we felt we had more of a strong position to argue from on what we should be doing. And the activist just didn't happen to want to listen to our view on that so it actually created the divide. But yeah, so fortunately, we won that contest and since then, we've just been executing on that strategy.
Turner Novak:
Yeah. Now you have a whole new runway with all the AI stuff that's happening. It's exciting times.
Aaron Levie:
Which is a massive just boost of, I think, energy and excitement for, again, software and technology in general. I certainly, in the 20, 25 years that I've been doing technology, this is easily the most exciting moment. From the early days of just building websites in high school and middle school to today, there's just simply never been a better move to either be building a startup, to be leveraging technology, using technology. So I think we're in a very exciting moment in the tech industry right now.
Turner Novak:
Yeah, I think a lot of people would agree. Well, this was an awesome conversation. Thanks for taking the time to chat.
Aaron Levie:
Hey, thanks for having me. Appreciate it.
Stream the full episode on Apple, Spotify, or YouTube.
Find transcripts of all other episodes here.