Transcript
0:30 >> I first met Aaron Levie in 2010 at Aim, a content management industry conference, when I was running the Fierce Content Management Newsletter. We met in a random side room off the floor because Box didn't even have a booth at that point. And on that same exhibitor floor, companies were literally selling cardboard boxes to store paper files. But, there were also software companies there, amongst them giants like Microsoft, IBM, and EMC. And here was this guy trying to convince enterprise buyers to take a chance on his startup, convincing them to move their valuable files to this thing called the cloud.
1:13 16 years later, Box is a multi-million dollar public company, and Aaron is leading another major transformation inside his organization, this one involving AI. What else? And investors look skeptically on at SaaS companies like has. I'm Ron Miller, and this is Fast Forward on PPN. >> >> I like that intro. Gets the energy going. Welcome, Aaron. Aaron is CEO of Box, as though he needs any introduction, but in case you don't know who he is.
2:10 I'm going to start off, Aaron, with a little trip down memory lane, if that's okay. you know, we all know by now that you started Box in a dorm room in 2005, and that you you started as a consumer online storage before shifting to enterprise a few years in. And, you know, I can tell you that having been covering all of this way back then, there were a lot of people in IT who were just pretty hostile towards the whole cloud idea in those early days.
2:41 How challenging were those early conversations with enterprise buyers, and how prepared were you as an executive to answer their concerns about SaaS and cloud computing? >> well, yeah, I would say I would say it would they were very challenging in the early days. and we were we were fortunately, probably decently prepared. It doesn't mean that convinced that many people. So, >> >> so we we had our we had our answers. not not everybody went with them, and and there was a tough period, between kind of like '07, '08 to maybe 2012, '13 where where, you know, we we absolutely would lose a number of deals because the customer wanted the system on prem, and we we were so religiously committed to the cloud, that we we were we basically said we are never going to veer off this architecture, and and the reason for it was we wanted all of the leverage gains of being in a in a single cloud platform, so we could just ship new features constantly. Customers wouldn't be on kind of forked or legacy versions of the code base.
3:50 And, and what's interesting is that is it it started to pay off in multiple ways. It paid off when mobile arrived because it meant that instantly you could just pull up your iPhone and you could access your files. And if your data was on prem, you often couldn't do that except for some weird kind of VPN workflows. it meant that we could integrate much more seamlessly with the SaaS ecosystem that was rapidly emerging like products like sales Slack and Salesforce and whatnot. And then now is the kind of ultimate moment where it's paid off, which is if you want agents to be able to work with your data securely, they obviously need really easy APIs to access that data. They're not going to like go and, you know, traverse a VPN and network system to do that. And so so kind of we're finally at the point where that architecture decision has paid off at its most in its most significant and important way. And and that's for a world of AI. So, we're incredibly excited for this moment and it's it's a great example of staying very committed and true to your vision despite losing lots of deals in the early years and not not necessarily convincing every customer to move to our our approach.
4:54 >> But it seemed like the whole industry moved towards your approach. I mean, that was what that was what happened ultimately, right? Even by the following year, everybody was talking cloud, you know, this from by 2011, even though there were still IT people who were saying things like over my dead body will I go to the cloud. >> Yeah, I mean, I I think I I I don't I I don't know if I felt like within a year it that everything had shifted, but but I would say, you know, firmly by 2026, it's pretty clear that the cloud >> >> I I felt like it took another 10 years, but and and you know, there's an interesting analogy here that actually will be that that will be related to AI, which is which is really kind of how long these transitions take. and even though it was obvious to us in 2006 and 7 that this would this had to be the architecture of the future, there was no other way that that that you would want to deploy IT at scale, even though that was super apparent and obvious to us, it still took 20 years for that to really roll out across every every organization. So, so I think we're in for, you know, some amount of diffusion and and kind of rate of diffusion where it will take some time for this for AI to hit most organizations and we have to, you know, be prepared for what that looks like.
6:12 >> You know, and I think I think you had a lot to do with the success because you weren't just selling box. You were evangelizing SAS and I think that was an important distinction and I see you doing the same thing today with AI. You're you're, you know, all over social media. You come on you know, podcast like this one and you're talking about the value of of AI and you're trying to get customers to understand that. But there are a lot of companies who are struggling with that and we keep seeing study after study saying that people are having trouble scaling it and I'm wondering like how do you reconcile this honest idea that you believe in the future of AI but it may not be where people want it to be right now.
6:58 >> Yeah, I would actually say that that is I don't really have to reconcile that. That actually is our business model. So our business model has always been in some way, shape or form take incredible breakthroughs that we get to see in the tech industry and figure out how to translate them into real world organizations within a particular domain of of working with your unstructured data. We obviously don't do this for for every kind of category of work, but when it comes to how do you work with your enterprise information like your contracts and your marketing assets and your financial documents, we our job has always been to bridge breakthrough technology with the real world and AI is no exception. So so I think you can have this duality which is, you know, in some areas of work coding is a great example. or in some types of organizations like AI-native startups, you can see these amazing breakthroughs happen with AI.
7:50 Conversely, when you go to the, let's say, real world and you go and you say, "Hey, let's just look at a marketing team at a large CPG company." You might not be seeing those same gains today because their systems look very different from that small startup. Their practices look very different from what a coder, you know, ends up doing. And these are just totally different ways of of working. And so, it's going to take time for AI to actually diffuse into those those, you know, parts of organizations and and those ecosystems.
8:18 and this is why I I'm, you know, I I really suggest that people have some degree of patience, where this is going to take many, many years and it's going to take a lot of change management to actually go and and make this happen. So, for me, the reconciliation is just, yes, it's we we live in this kind of crazy dual world, which is, you know, you can right now go and talk to, you know, a super intelligence on your phone and get, you know, legal advice and medical advice and and you could generate code and that's that's this amazing thing that that makes us instantly more productive and effective as individuals. And then, when you go to an enterprise and you say, "How do I get that same level of intelligence attached to my IT systems and that access my data and could be incorporated into my workflows?" That's a massive project and undertaking that companies are really only now just starting. And this is not the same as when you're just like asking a question and chatting with an AI system. This is, how do you have an agent that that has some some kind of goal and it can generate a plan and it can work inside of your environment and it can execute work and incorporate into a broader workflow. That is a much more both meaningful type of problem to solve as an enterprise, but it's a lot harder.
9:26 And and that's why we're still in these very early stages. So, that that's kind of how I look at the data right now and and and how I make sense of it cuz we're seeing it every day with our customers. Like, they are super advanced in some areas, like let's say AI coding, and super early in other areas, like workflow automation across the the general knowledge worker base. >> Yeah, I mean I think that's that's really tough. I can tell you from my experiences trying to build, you know, agents on my own that, you know, this this this seems to be like this continual tweaking cycle that goes on endlessly where you try and communicate to the to the AI what you want. It it it delivers you something and you're like, "Okay, it's not quite what I wanted."
10:08 >> Well, but but this is this is actually why I'm very optimistic about people, by the way, in the in the workflow. AI is like I have almost never generated something of important, you know, of of any kind of important value where in a in a single shot it generated it properly. almost every single time I still had to do two, three, five extra prompts or go in and modify things manually myself for for producing real value. So, if if that sort of sustains as a as a, you know, kind of a component of of how work is going to look, then what's going to happen is in you know, maybe maybe 2 years ago AI could do 20% of our workflow, and and you know, we got very minimal kind of productivity gain as a result of that 10 or 20%. Maybe now it can do 40 or 50 or 60% of our workflow. Maybe in the future it'll do 70 or 80 or 90% of that workflow. But, here's the thing. What ends up happening is as that occurs, so first of all, you still need a human in the loop that entire time because it's just the human in the loop sort of gets to move up a layer of abstraction cuz now you're only doing the final 10%.
11:16 So, so in in none of that kind of process could you fully eliminate the human oversight. But, here's what happened Here Here's what ends up happening that that's super interesting to think about is is what happens next is we actually then raise the bar of of what the work output we want to what we want to have in that in that particular work. And once we raise that bar, that 90% now drops back down to about 40% and we've just jumped another sort of step function because what we do is we expand the we we sort of move the goalposts of what we want AI to automate. It's no longer sufficient to to just have read the contract and provided me a summary accurately. It's no longer sufficient to generate a you know, a file or two of code. We end up raising the bar of what what kinds of projects and tasks we want to take on and then again, human in the loop then is necessary for that next sort of remainder part to go and review the work. So so I think we have to actually get used to this idea that we will always be in the loop.
12:15 AI will almost never generate something perfectly because what's what will happen is whatever we used to think was perfect, we will expand the definition of in the future to the point where it's imperfect again and we are back to having, you know, additional work that we end up having to have oversight on. >> Right, that's a really interesting point that you can you know, you can increase the amount of work that you can do because you know, every business has this list of things that they want to do but budgetary constraints prevent them from doing it and and you know, personnel constraints prevent them from doing it and they kind of fall off the list. If you can start to push those things into the done you know, column, certainly you know, you're going to be a lot more productive. But you know, what one one thing I want to talk about a little bit because you have this very optimistic view of AI.
13:03 But what we're hearing from you know, some of the the heads of the AI labs and and you know, people who are in charge of AI at companies like Microsoft is that it's going to be the end of white-collar jobs. What you're saying is that you know, there's always going to be a human in the loop. You know, that kind of negativity is leading to you know, students booing the mention of AI at a you know commencement this week. so so how do you kind of think about what those people are saying and like this negative you know energy that's being put out there about AI when there's also this positive side and nobody's talking about that.
13:44 >> Yeah, I mean I think we're shooting ourselves in the foot in the tech industry. you know if you if you actually do believe that that we're in a race with China the tech industry is is almost single-handedly doing everything possible to lose that race. because if we end up scaring people about the power and the value of AI, absolutely people aren't going to go for it. they they will totally you know vote this thing out of existence if if it's not something that feels like it's helping us and making us more productive and making work actually more interesting as opposed to less interesting. So so I I I I don't subscribe to the rhetoric that that you just referred to. I from from some of the the companies out there partly because I'm not even seeing it.
14:32 I you know the the work that we have put most AI on I'm more excited about hiring more people for those types of jobs because AI is making them even more productive which makes it even more interesting to go and be able to get higher velocity and output of that work. So so I'm I'm just I'm I'm seeing the other side of this which is if I can automate something in the organization and make an an actual human more productive assuming it's an area where there's still more demand to do that type of work and and in most of our organizations we have we still have more demand than we can kind of meet.
15:07 then actually I want to hire more people to to do that and I think that's the part that that a lot of the the labs and some of the maybe more doomer mindset kind of misses is is this idea that that we're just going to take the current fixed amount of work that we do and AI is just going to eat into that fixed amount of work and then obviously labor is that remainder that goes away. But what I tend to see is that the work just expands, you know, based on the kind of tool capabilities that we have and I've seen no evidence at at any kind of macro scale where that that doesn't happen.
15:38 I have I have friends that are hiring engineers for the first time ever because because now AI is letting them go and develop software in a way that they wouldn't have before. So I think what you're going to see is just a shift of of talent maybe from some organizations that aren't, you know, kind of approaching AI in that way to the kind of organizations where where there is growth and there is demand. One more kind of small note, you know, we we tech tech is sort of obviously the epicenter of this conversation because AI coding is sort of the most powerful, you know, kind of form of AI at the moment and and what's interesting is we kind of myopically assume in in the tech industry that tech jobs equal tech industry and and like and and that that we are the we are the the house of, you know, we we are the the home of all the tech jobs. But what we don't realize is that like every farmer company on the planet would love to to have access to amazing computer scientists that can go and automate drug discovery and and go work in life sciences for for clinical, you know, trial automation or health care providers would be able to love to streamline their workflows to be able to get more throughput in their health care system or manufacturing companies and industrial equipment makers are leaning into AI because they know that that, you know, industrial you know, processes in the future are going to be AI enabled. Well, guess what? That's going to produce a ton of jobs for engineers that that, you know, are going to be working on incredibly important, very mission-critical work in in the future. And then so yes, you might not go and work on a mobile game app like you would have before nothing wrong with mobile games. It's just an example of the kind of tech job that people kind of think is like classically tech and maybe that same engineer is now going to go work at John Deere or Caterpillar or Eli Lilly or or Pfizer and they're going to be working on incredibly important work. It's just a very different domain of of the applied use of that technology. And so so this is again where the doomers maybe don't understand that actually no, there's actually a lot of demand for this work elsewhere. There's just going to be a shifting of how do you get access to that talent pool in those organizations and that's that's the kind of change that we're about to embark on.
17:42 >> So speaking of engineering jobs, you you've been you've been writing a lot about the forward deployed engineering movement lately and you see it as positive. I wrote about it and I wrote about it as negative because my point of view was that it adds a lot of cost because these these engineers are highly trained. They're basically like consultants and if you look at like the AI or the open AI deployment company, it's basically open AI trying to create, you know, like a mini Accenture or or Deloitte to go out into the to these large enterprises and help them, you know, implement AI at scale because they can't do it themselves. It's kind of a tacit admission in my view that whatever is being delivered isn't isn't working.
18:28 but you see it as positive and I'm wondering why. >> Yeah. I I yeah, I I mean I have I have sort of the other other kind of take on that. I I think that now maybe maybe from where an AI sort of super accelerationist would have thought five years ago, maybe it's a maybe this is a a disappointing outcome where you thought the agent would enter the enterprise and like automate everything and it was just going to all work amazingly well, but for anybody practical in the enterprise, you just instantly know like these are really really difficult systems to to implement securely, to get them the right data that they need to be able to to to work, to do the change management of the workflow that the that the people are involved in, all of those things require a significant amount of change both in technical implementation and kind of human and process change management. So, I don't see that as a failure by any means. I I see that as a natural occurrence in organizations, which is which is why do organizations ever hire a consultant or professional services firm? It's actually a really straightforward concept. And and every every you know, sort of core business, you know, management science sort of piece of literature that studies this, you know, can explain it very easily, which is firms get really really good at at some core competence. And if you're you don't sit around all day long as as a mid-size life sciences company thinking about the core competence of doing IT change management for AI. You just don't think about it. What you're thinking about all day long is how do I go and develop and design drugs?
20:06 And so, all of a sudden, boom, AI shows up, and you're like, "Oh my gosh, how do I go implement this thing to actually go and turbocharge my workflows?" You have a choice. Do you want to implement that yourself and learn all of the the ways of doing that, and including many of the mistakes that that that that are out there, or do you theoretically, and I'm I'm I'm I'm presupposing that you're going with a good vendor, do you want to work with somebody that has seen it 20 times or 50 times, and they can kind of cross-check the way that they've done it across 10 other customers, and they can bring you best practices to go and implement this?
20:40 So, it's actually just like it's just like capitalism is is like doing exactly what it should do. The that life sciences company is not in the business of doing IT automation. There are IT companies and services providers that are in that business, and they can bring those best practices to the life sciences company. Now, at a certain scale, some companies build this competence internally, but most organizations just don't have the capacity for doing all of this work themselves. So, so I would say if anything, it's a it's a bullish sign that our industry in tech and AI is actually growing up realizing the the real path forward for how these systems are going to be implemented.
21:16 >> I mean I think anybody that uses this and I I I I use it I know these tools a lot because I want to figure out how they work and if I'm going to report on them I have to understand them and I find them incredibly frustrating and and I I'm sure that you know people and not that I'm super technical but people who are less comfortable with technology than I am must you know really have a lot of a lot of trouble with it and you know I don't I think it's like a super big organizational problem. If you go back to you know digital transformation 10 years ago when everybody was you know looking at companies like yours and saying like how do we digitally transform our systems it's an even bigger transformation than that isn't it and and it's like companies have never been good at those big transformations.
22:08 >> Yeah I think it's it's much bigger than any digital transformation because when you think about digital transformation you know in general the thing that went digital was often the customer experience. So the customer experience went from you know an analog maybe in-person based process to you know e-commerce or mobile apps or you know kind of better real-time feedback and and analytics. That was kind of digital transformation in the enterprise. That only really affects you know 10 20% of the organization. Maybe it affected the product team maybe it affected the go-to-market team. AI affects every single you know crevice of the organization. It affects every workflow in the company and so this is a this is a transformation far larger than than the scale of of digital transformation or cloud transformation or even kind of moving companies to the internet. and it's going to This is actually why it's actually very easy to be bullish on Accenture and McKinsey and Open AI deployment company because actually there's so much work ahead for companies on going through this this transformation journey.
23:17 >> So, I mean, we're already at the bottom of the hour. So, >> >> I had so much more I wanted to talk to you about, but I'm going to ask you a personal question to close things out. >> Yeah. >> you've been at this for 21 years. You were You were a kid in a dorm room when you came up with the idea. same company, same mission with, you know, things changing over the years. so so how do you stay passionate and engaged in you obviously are? And do you ever long for a challenge outside of Box because that's been your whole professional life, your whole adult life?
23:48 >> so, I think the reason why I get so excited is because I'm just a I'm a I'm a kind of technology fanatic. I've I've loved technology since I was 13 maybe is when I got, you know, my first computer and and you know, we had a computer in the house, but 13 I had my own computer and and this was like a very big, you know, moment to to be able to build websites and build internet products. And I had friends that were were were engineers. And so so, so I love technology. I buy every new you know, tool that comes out and use every tool and buy every VR headset and try and quickly see do I like it or not and and you know, what what thing is exciting and not. So, I just love technology. And AI is is another amazing breakthrough that I'm always wowed by every day.
24:37 and and so that that's what keeps me extremely excited and and kind of, you know, really on top of this. As it relates to Box, I think what's interesting is is that everything happening in AI right now, I think actually reinforces and makes the value proposition of Box even more maybe, you know, substantial and and important because what agents really need is they need access to your most important data if they're going to be useful. So, so, you know, for right now, I mean, we we see ourselves as becoming directly in the in the very center of of what AI is going to need to be productive in the enterprise. So, this is kind of this incredible moment where it combines 20 years of of building some degree of expertise in the space with with now a a complete technological revolution that needs exactly the thing that we've built. So, So, I I have not found an opportunity to be bored in the past you know, number of years on this.
25:31 >> Well, I'm I'm sorry we didn't get to geek out on headless content management or talk about token pricing models or any of that that kind of stuff because things just went by too fast, but I want to thank you, Aaron Levie, for joining me. >> Thanks, Ryan. >> I want to thank everybody who tuned in for for watching. If you like what you heard, you know, please subscribe to the Fast Forward newsletter. you'll see the QR code on the outro.
25:56 Thanks again to everyone. You've been listening to Fast Forward on PPN. >>