By playing, you agree to uListen's Terms and Privacy Policy and YouTube's Terms of Service.
Sequoia CapitalBox's Aaron Levie: On Reinventing Yourself in the AI Age and Enterprise Diffusion
Summaries, chapters, transcripts, and FAQs on this page are made by uListen, not YouTube.
EVERY SPOKEN WORD
75 min read · 15,003 words- 0:00 – 1:55
Introduction
- ALAaron Levie
I wanna send an emergency alert to, like, everybody who's, like, a sophomore or junior in college and just be like, "Follow these 20 accounts-
- SHSonya Huang
Yeah
- ALAaron Levie
... on Twitter."
- SHSonya Huang
Yeah, I agree.
- ALAaron Levie
And, and also join Twitter because, like, this will just help your career. You're either, like, a year ahead or a year behind-
- SHSonya Huang
Yeah
- ALAaron Levie
... simply based on your feed. And my feed is, like, so, so wired in. I'll still talk to, to 20-year-olds that are like, "Yeah, like, um, you know, I, I, I see some articles." And I'm like, "What do you mean? How do you see articles?"
- SHSonya Huang
[laughs]
- ALAaron Levie
Like, I don't even know what that means. Like-
- SHSonya Huang
Yeah
- ALAaron Levie
... do you just, like, get lucky that somebody emailed you an article? You just have to be wired in. I, I enjoy it. It's a lot of fun. I play with everything. [upbeat music]
- SHSonya Huang
Today I'm excited to welcome Aaron Levie, founder and CEO of Box. Box is building a collaboration platform and content management platform for the enterprise that has now taken on really new life with AI. And I'm excited to chat with you today about Box and AI, about your general thoughts on AI because you are such a thought leader, and then about how founders can reinvent themselves for this AI wave. So, thanks for joining us.
- ALAaron Levie
Thanks for having me. A big fan of, uh, the podcast. Um, I religiously, uh, watch every episode.
- SHSonya Huang
[laughs]
- ALAaron Levie
And, um, and so great work. I am wondering, though, I know this is a different podcast, but when do you talk about root canals in, in this?
- SHSonya Huang
Oh my gosh. [laughs]
- ALAaron Levie
What, what... Is that... Are, have we figured-
- SHSonya Huang
We gotta call in Doug Leone
- ALAaron Levie
... have, have we figured out how to weave that in yet or, or not?
- SHSonya Huang
[laughs]
- ALAaron Levie
I mean, I'm happy to talk about different kinda pain and suffering that, uh-
- SHSonya Huang
We-
- ALAaron Levie
... that I've been through
- SHSonya Huang
... we should just do this pod with a root, with a live root canal. Let's do it.
- ALAaron Levie
[laughs] See how it goes. Actually, that would... It would be like Hot Ones, but you get a root canal.
- SHSonya Huang
[laughs]
- ALAaron Levie
And you have to talk about your strategy while you're in a dentist chair.
- SHSonya Huang
[laughs]
- ALAaron Levie
[laughs] And Doug is just-
- SHSonya Huang
Oh my gosh
- 1:55 – 6:56
Are application companies the hottest neolabs?
- SHSonya Huang
Uh, okay. Let's start with, I'm curious your take on this, application companies are the hottest neolabs.
- ALAaron Levie
Yes.
- SHSonya Huang
Agree or disagree?
- ALAaron Levie
Um, you know, two years ago I think it would've not made that much sense as like, what does that mean? Um, but, but very clearly I, I think this is what, what's playing out in the market. And, um, it's all working out mostly because of open source. But, uh, what's, what's pretty amazing is right now, you know, the, the whole concept of, uh, being a, a sort of LM wrapper or model wrapper is actually working out. Because, uh, what, what, what I think people underappreciated was that in the real world, in the enterprise, uh, what you need is some bridge from the model's capability to the actual workflow that the enterprise has. And that bridge, you b- basically, like, probably a trillion dollars has been bet on, on basically one of, uh, of, of two outcomes. Either that bridge is, is very kind of limited, or that bridge is actually very vast and, and needs to be able to, to, you know, take on, you know, a lot of depth within organizations. Um, and that bet basically looks like are you only long, you know, kind of the, the model itself and super intelligence, or are you long this sort of application tier or, you know, maybe previously would've been just pure neolab. Um, but, uh, but I think it's very clearly playing out that, that actually there's a lot of gap between the model and the workflow, and as you bridge that gap, over time you get to a point where you realize, oh, I should also do the model. And, and then you have enough data, you have enough sort of domain expertise where that becomes its own, you know, sort of flywheel. So I'm very bullish on this, um, uh, on this idea, and what's cool is it's opening up like multiple layers of, of, you know, y- sort of opportunity for startups. 'Cause you could either be the actual applied company itself, i.e. the neolab, or you could be the infrastructure provider to the neolab.
- SHSonya Huang
Yeah.
- ALAaron Levie
And you have like multiple layers of, of, you know, going and, and attacking that space. But, uh, I think a huge update for, for the market and everybody's kind of views of it.
- SHSonya Huang
Box Labs, let's go.
- ALAaron Levie
Uh, we, we already have it. It's a little bit, it's a little bit secret, but, uh, we're, we, we, we pay very close attention. Right now the m- main focus is let's make the agent really, really good on any, uh, model. But over time, obviously, you would peel off certain use cases, um, either on a per customer basis or, or kind of across the, the whole data set.
- SHSonya Huang
Yeah. I mean, it seems like there's two forces that are happening. One is people don't want the fox regarding the henhouse. They don't want the seller of the token to be the one that is also metering and gating, you know, what is the best token for each use case.
- ALAaron Levie
Yeah.
- SHSonya Huang
Uh, and then the second is, you know, there's actually a lot of work to do on that bridge to cross.
- ALAaron Levie
Yeah.
- SHSonya Huang
And there's real research involved in it. And, um, it's pretty bespoke to the exact, uh, the exact workflow and the exact end customer that you have.
- ALAaron Levie
Yeah. I think, I think what, what I tend to see happen in the Valley is... And, and for very good reason, and it's actually why these companies have been so successful, is like everybody is so kind of research pilled, uh, which is again, totally awesome, big fan. It's, it's le- leads to all these breakthroughs. Um, but the, the sort of sense that, okay, the model and the intelligence in the model is kind of the only form factor that matters. And then you go to the real world and you, you sort of see how intelligence actually gets rolled out in people's workflows. And the model could be the most intelligent, you know, super intelligence in the world, but that workflow still requires you to connect up to other data systems.
- SHSonya Huang
Yeah.
- ALAaron Levie
Still requires these moments where there's a human in the loop interaction. There's delays in the workflow, and so the, the sort of agent has to sit idle. Um, there's change management of the actual business process. There's legacy systems that haven't been modernized. So you kinda go through these 5 or 10 things that are way sort of, you know, much more operational, much more blocking and tackling than just the pure super intelligence of the model. And the last thing that I think a classic sort of research organization wants to go do is go attack every single one of those things. And this is sort of not a... There's like, this is not the, uh, a sort of a one-off in history. Like we've always had this relationship between infrastructure and application. Like, you know, obviously AWS or GCP or Azure have created, you know, trillions of dollars in value of, of kind of market cap of infrastructure, but guess what? There's also trillions of dollars of value in software-
- SHSonya Huang
Yep
- ALAaron Levie
... that only exists because of that infrastructure. And if you were to go back, you know, 10 years ago and you were to look at what was happening in the data space and you looked at what, what, you know, GCP, ear- early kind of versions of GCP was building or, or AWS was building, I guarantee you would not have predicted Snowflake or Databricks existing. You would've been like, "The infrastructure just already does that. Why would you pay another $10 billion of revenue to all these other products that are just making it so you can work with your data?" The same thing is gonna be true for intelligence, which is the models will be insanely valuable But the application of bringing those models into real workflows in banking, and life sciences, and healthcare, and government, that's just gonna be a lot of software. Now, the sort of challenge for the next, let's say, two to five years is how much do the model providers need to move up that stack and also try and, you know, go and compete at that layer? Or do they either leave open intentionally or accidentally that entire space for the application sort of ecosystem? And actually, to some extent, this is like a, a big question strategically for them because on one hand, you wanna be closer to the customer. On the other hand, you also wanna be able to have an ecosystem so people trust you.
- SHSonya Huang
Yeah.
- ALAaron Levie
So, so that's gonna be, like, a, a really interesting tension over
- 6:56 – 12:34
Will the labs move up the stack?
- ALAaron Levie
the coming years.
- SHSonya Huang
How do you think that'll play out? I feel like that's the question we spend every day-
- ALAaron Levie
Yeah, yeah, yeah
- SHSonya Huang
... wrestling with.
- ALAaron Levie
Yes. Uh, you can... Yeah, I, I, you know, I don't wanna speak for, for Sequoia, but you can sense some of the existential dread of a VC right now is like, "Should I just put another billion dollars into Anthropic? Or, you know, should I attempt to sort of see what, what at the applied layer is-
- SHSonya Huang
Yeah
- ALAaron Levie
... is gonna, is gonna play out?" So nobody's, you know, that envious of, uh, of your guys' position having to, to figure that out, but, but obviously it's even harder for the entrepreneur. Um, uh, but you know, I, I am, uh, you know, I'm pretty long obviously the application layer. Like [chuckles] I, I, but I'm also equally very biased. Like I'm, I'm-- I have a very concentrated bet, uh, with very limited diversification on, uh, on it working out that you still wanna buy technology that, that sort of understands your workflow and can get to the, the core enterprise data. But I don't know that, that there's... I, I just don't see a different event happening than the, than all of history. I mean, m- you know, we only have like 50 or 60 years of, of, of computer software history. Um, but basically, when you go to that law firm and you go to that w- you know, pharma company and you go to that bank, they need something that bridges the core technology to their workflow and their business process, and AI has not meaningfully sort of changed the need or the shape of what that looks like. And the, and the best sort of manifestation of this is, um, sort of this chatbot versus kind of agentic workflow, you know, demarcation. So the chatbot can be totally universal and can be totally horizontal. But you know, y- all of a sudden you kinda look at that and you say, "Well, my workflow kinda needs something to kinda ping me at the right time in the process, or it needs access to a certain kind of data that, that just, you know, the, the chatbot can't natively get access to." So somebody has to go into that organization and like get it set up, and somebody has to go and provide domain expertise to this model so it really understands our particular business process. Um, and, and so then unless you really just underwrite, you know, the big labs at, at, honestly like, I'm not exaggerating, like 100,000 employees. If you don't underwrite that, then, then the diffusion kind of economy is gonna be massive. Because every single one of those companies, whether that's a 50-person firm or certainly a multi-100,000 person firm is gonna need an army of people to go in and help them with that transformation, that change management. And so this is, you know, like whether it's the FDE phenomenon or just again understanding that domain expertise, that, that's gonna be a very big deal. And then you, you kind of alluded to, to this point, um, uh, fox kind of guarding the henhouse and, um, [lip smack] uh, and, and that, that might actually be singularly the biggest sort of reason this has to happen.
- SHSonya Huang
Yeah.
- ALAaron Levie
Which is even, even under all sort of like complete benevolence and like nobody's actually doing something, um, uh, in, in any kind of like, you know, sneaky way, it just stands to reason that if I'm gonna sort of give a task to an, to an a- an agentic system, I just want that task to be cost optimized.
- SHSonya Huang
Yeah.
- ALAaron Levie
Like with accuracy as, as kind of holding constant. And so who can do that? It's, it's the company that doesn't care among, you know, 10 different models which model is, is performing that task. It's like definitionally you would want that to be the, the, the company that does not have a-
- SHSonya Huang
Totally
- ALAaron Levie
... a sort of a preference.
- SHSonya Huang
And then the f- and then the force you have going against that is that the model companies can subsidize their, their models or offer them at different, different rates.
- ALAaron Levie
I don't know how long that lasts though, because when, when these companies become public, I think they will be held to basically the same laws of, of capitalism that everybody else is. So, so the subsidization is working very well up to a certain threshold of spend, and we might have, we may have exceeded that spend when you're at like tens of billions of dollars of, of kind of capital. And in fact, if anything, it might even be worse because, uh, eventually you have to go and, and sort of pay for your training runs as well. So like they, they, like the subsidization of tokens is a-- I, I think it just has to be a temporary phenomenon.
- SHSonya Huang
You mean from a gross margin perspective or from like an antitrust perspective?
- ALAaron Levie
En- entirely gross margin. Like if I-
- SHSonya Huang
But they have such high gross margins on their inference right now and so they can afford-
- ALAaron Levie
But that's-- But then what are they subsidizing? The... Like then, then it's actually then they're, then they're charging actually like, you know, decent rates.
- SHSonya Huang
Yeah. But they can afford to price the API higher than their own first-party products, right?
- ALAaron Levie
So that part, uh, tot- totally fair. You know, there's an interesting kind of, you know, dimension which is well, like if API is the high margin thing that is sort of paying for the subsidization, but you've moved all your customers over to the applied product and there's no API revenue. So like-
- SHSonya Huang
Yeah
- ALAaron Levie
... there is like an equilibrium you have to strike with this.
- SHSonya Huang
Yeah, yeah. Totally.
- ALAaron Levie
Um, and uh, and so then all the while, if you have some sort of either non-economic actors or people just with a totally different game theory in this, you know, Meta being one, maybe SpaceX being one, China certainly being a giant one, even Nvidia being one, like that changes the calculus as well, which is those four kind of cohorts don't necessarily need to make money on, on inference in the, in, in sort of at least the, the same margin structure at like what Anthropic or an OpenAI needs. So they would might be fine to bring down inference to 10% margin because they just wanna basically com- you know, pay for massive compute clusters. And then as long as that happens, and as long as like there's not insane proprietary sort of, you know, you know, sort of closely held secrets, then no matter what you're gonna have kind of cost per token go down on a, on a like for like basis. All of which means more value accrues to, to the application layer. Which is, I think, like a very long-winded way of saying, like, I think there's just value in kind of everybody in the stack. I just don't know that, that the... I, I... The only thing I probably wouldn't bet on is just, okay, one or two labs get 95% of the value creation. I think there's just gonna be a much more dynamic environment. And honestly, if I were one of the, you know, two or three biggest labs, I think I'd prefer this outcome too because back to your antitrust point, like at some point you'll just be nationalized if you're the only thing that, that is, is sort of exists as intelligence. So you kinda want a little bit of healthy competition in this
- 12:34 – 16:50
Box and betting the company on AI
- ALAaron Levie
ecosystem anyway.
- SHSonya Huang
Yep, totally. Um, let's, let's transition talking about Box. I wanna come back to talking about tokens and Jevons paradox and China and all these, all this stuff, but let's, let's talk about Box-
- ALAaron Levie
Sure
- SHSonya Huang
... for a second.
- ALAaron Levie
I love that topic too.
- SHSonya Huang
Um, [laughs] give folks-
- ALAaron Levie
I'm a big fan.
- SHSonya Huang
I, I mean, I'm guessing a lot of people that listen to this podcast use Box, but give, give pe- people, like, a brief kind of explanation of the history of Box and how, how you're reinventing yourself with AI.
- ALAaron Levie
Yeah. We started the company as a way to be able to s- kind of securely store and share data in the cloud. Um, and, um, uh, and it was a very simple idea but, but we, we, we kinda just kinda cracked a, a, a nut or, or struck a nerve, um, and, um, uh, you know, for, for Doug out there, uh, we, uh, we, we were able to sort of scale up quickly. We pivoted rapidly in the enterprise, and the idea was enterprises would be moving from on-premises systems to the cloud, and they would need a, a better way to be able to s- sort of store, share, collaborate, manage their, all this unstructured data, their, their corporate documents, their financial documents, their marketing assets, their research materials in the cloud securely. So that was the, the kinda company. And we, we had been sort of flirting with AI kind of e- products and, and sort of experiences really since like 2015. Um, if you remember like the first kind of rise of the, you know, maybe the, the, at least in modern times, the, the AI winter that happened in like the 2015 to 2018 period, which is like we think it's gonna happen now, and then it didn't. Um, and, uh, uh, that was a period where we were like, okay, these very, you know, sort of early AI models were showing us signs that, okay, if you, if you, you know, looked at an image and you could classify the image, well, that's pretty useful if you're in an enterprise because now maybe you take all of your image data and sort of label it, or maybe you'd OCR something and you'd be able to kind of pull out, you know, kind of o- you know, sort of the, the, the text in there. That's enormously helpful. The problem was is it insanely expensive, and you had to have a model for every single use case required a, you know, kind of a hyper-trained model for each workflow that you wanted to do. So we kinda shelved it. Um, you know, few years later, uh, you know, started paying attention to the GPTs. We had some hackathons where people were like, "Oh, we could do like, you know, type ahead in, in one of our kind of, you know, note-taking products." And, and that was, that was sort of early kind of versions. We did some, some early work in, in sort of text detection and classification, which helped with security use cases. Then obviously ChatGPT moment sort of hits and, um, and that was the, the big sort of, you know, head exploding moment if for no other reason why, you know, then, then they kind of figured out a form factor that opened up everybody's mind to, oh, these could be these interactive systems that you just like ask a question, get an answer back of, of sort of, you know, kind of increasing complexity and length.
- SHSonya Huang
Yep.
- ALAaron Levie
So we looked at that very quickly. We jumped all in. We sort of did the whole company pivot, like, like everything was, was kind of exactly like academically what you should do. We had a team carved out. We, we put, you know, the best people on the team. We, we, you know, met every day, looked at the updates, and then slowly but surely sort of built out what, what today is, is kind of our a- you know, our AI stack and then basically the Box Agent. And, um, and for us you can imagine the use case is, is, you know, very s- straightforward. We sit on hundreds of billions of files. Every single one of those files contains critical information for an enterprise. That could be their contracts, their research files, their marketing assets, their loan documents, like all of this critical information. The problem is they rarely know what's actually inside of it, so they, they, they, unless you literally look at the document and, and kinda search and find it, you just don't know what's inside of it. So now agents can go and basically be farmed out to go and answer questions about that data. They can pre-process it and extract metadata from those documents and turn it into structured data. You can use agents to automate sort of steps in workflows. So we're, we built a platform that basically lets you deploy agents against all of that unstructured data, and that's been the, the kinda core focus.
- SHSonya Huang
Are you using agents to create new content?
- ALAaron Levie
We are. Um, there's a couple modalities where that shows up. One is we have, again, an online sort of collaborative product that, that an agent can just like generate any amount of content in it. And then, um, and then we've done most of the, I, I think probably more exciting work with, with OpenAI and, uh, Anthropic on just how do you do like advanced document creation, PowerPoint creation. We've, we've decided that like their tech is, is, you know, at, at this point gonna always be frontier. So, so we have an agent that goes and interacts with those systems, uh, to produce, you know, a, a high quality PowerPoint, et cetera.
- 16:50 – 18:42
Hero use cases: reading a million contracts and long-running agents
- SHSonya Huang
Awesome. Your tagline, "Your business lives in content. Unleash it with AI."
- ALAaron Levie
Yes.
- SHSonya Huang
So what are the hero home run use cases for how people are unleashing it today?
- ALAaron Levie
Yeah.
- SHSonya Huang
And then if you had to fast-forward a few years, what do you think people will be doing with AI and your product in a few years?
- ALAaron Levie
Yeah. So the, the, probably the easiest hero for, again, more of a traditional enterprise to think about is, is just as simple as you have a million contracts. Why don't you find out what's inside them? Or you have a million research documents. Be able to go and pull out all of the critical structured data, put that into a database, and then be able to query, analyze, automate workflows around that. So that, that's kind of the thing that, that just knocks it out of the park every single time because it's been a long-standing problem that people have never been able to go and, and, and sort of apply human, you know, sort of labor to because it's just too expensive to read every, every contract, every research document. You know, maybe you could do it if you had like a loan document process, but most other data just never gets read at that, at that scale. And then I think the, the stuff that, that we're probably, you know, uh- As much if not more excited by is, is really the equivalent of what we see with, let's say, coding agents or other, other, other complex agents, which is you have long-running agents that are just executing your entire kinda workflow or process. And, um, and this would be in the form of you go to a bank, and you're onboarding at a bank, and they've basically, like, automated every step that is possible to automate, and then sort of jumps out to a person in the, in the steps in the process for extra review or extra verification. But now instead of that sort of one or two-week back and forth, it just happens in, like, an hour. Like, that's the dream state of most of these enterprise workflows, is what if we can onboard a client faster? What if we could discover kinda critical data inside of our research much more quickly? What if we can alert to a security event much more, much more quickly? So to do that, you need these sort of background agents or workflows that are, that are sort of pre-established, uh, for those processes.
- SHSonya Huang
Totally. It's the year of the long-running agen- agent.
- ALAaron Levie
It, it is.
- 18:42 – 24:08
Work slop: why AI code is embraced but AI content isn't
- SHSonya Huang
Yeah.
- ALAaron Levie
It is.
- SHSonya Huang
Yeah, yeah. Um, I'm curious, like, you, you made the analogy to cloud code. It seems to me that in the coding domain, uh, using AI is, like, not only accepted, it's embraced.
- ALAaron Levie
Yes.
- SHSonya Huang
Uh, in the content domain, which is, I think, a lot- where a lot of the content in, in Box sits-
- ALAaron Levie
Yep
- SHSonya Huang
... um, using AI to pro- produce content at least, it's just like there's this almost this allergic reaction to it.
- ALAaron Levie
[laughs]
- SHSonya Huang
Like, all the Pangram stuff on Twitter.
- ALAaron Levie
Uh-huh.
- SHSonya Huang
There's the, um, you know, it's like this concept of work slop.
- ALAaron Levie
[laughs]
- SHSonya Huang
I'm curious what you think about work slop.
- ALAaron Levie
Yeah.
- SHSonya Huang
And, like, will this still be a thing in a few years?
- ALAaron Levie
I'll, I'm gonna sort of separate the Box corporate hat and just now kind of maybe riff as a-
- SHSonya Huang
Yeah
- ALAaron Levie
... as a, as a, as a consumer of, of... Gosh, I wish there was a better term, but work slop, as, you know, inside of an enterprise context. Um, uh-
- SHSonya Huang
I get board decks that are en- entirely written-
- ALAaron Levie
I know
- SHSonya Huang
... by AI these days-
- ALAaron Levie
I know
- SHSonya Huang
... and it kills me.
- ALAaron Levie
So here, here's the difference, I think, uh, on the acceptability. Um, so there's probably, like, more, um, symbolism to this actually topic than, than, than just, like, the, the, the slop element, but, like, actually, like, diffusion of AI in general sort of almost ties to this. Code, like, other than, you know, the, the top engineers that we hang out with that, like, are, like, they have, you know, deep taste in the code. Like, like, you know, and, and the judgment is incredible and, and, like, it is to them an, as much an art as it is a science.
- SHSonya Huang
Yep.
- ALAaron Levie
So take that group aside. For most of the world, code is a utility.
- SHSonya Huang
Yep.
- ALAaron Levie
It is just trying to accomplish something. You're, you're just trying to automate something. You're just trying to put a, a, an, a sort of interface up there that somebody presses a button and moves to the next step. So for most of the world, the value creation of code has been to automate things and to be able to have it as a utility. So, so at the end of the day, uh, like s- like we're... And we'll probably still use the term slop for a while because, because there's taste in kinda front-end design, and there's taste in, in sort of systems, and you don't wanna have vulnerabilities in your code. So that's gonna exist for, for a while. But at the end of the day, if you can ha- tell an agent, like, "Please go and generate my entire backend system or my front-end system," like, it's just like it, it's not only acceptable, it's, it's preferable because it's just like that was the thing that was blocking us from moving forward, so we need to go do that. At least the way society functions and the way the world works and our brains work at the moment. Maybe this changes. You know, when you, when you get a presentation from somebody, there's still this association which is like I'm trying to decide if I can trust that person to go and execute on that thing or deliver that result or understand that topic. And so if when you see work slop, you're like, "I, my, I'm losing my ability to, to sorta know for a fact that, like, like how much of the, of the thought process was them versus how much was the AI. How much should I even care about that because I myself am doing the same thing?" So, like, we have this weird, like, it's a, it's this very weird sort of, like, collective issue that we have, which is like, which is like I'm doing work slop for some of my, you know, brainstorms and decisions. But when I get it from somebody else, I'm like, "Hmm, should I trust you?" Uh, and, uh, and I don't know. I, you know, it, I mean, it just might be a thing that as a society we have to kinda keep cranking through over the next kinda three to five years and, and end up at the other, at the other side. Like, like, you know, I hate to use, like, these, like, totally busted analogies but, you know, obviously, uh, you don't care when you see somebody's financial model. You're like, "Yeah, that was generated clearly by, like, a macro," uh, or, or, you know, that was, like, not, you did not personally go and compute all of that. But, but you're showing it to me, and we're talking about it. So why can't the same exist for a strategy deck or, or whatnot? But I think right now we're going through this evolution of, like, what is the person's role? What is the content a proxy for? Is it supposed to be a proxy for how much that person knows? Is it a proxy for what we think that they can go and execute on? I think, I think we're just in this very messy period where we have to kinda figure that out.
- SHSonya Huang
Totally. Um, did you read the Stan Druckenmiller Wall Street Journal article?
- ALAaron Levie
I, I, I read the, uh, I, I read the sort of-
- 24:08 – 27:23
Building Box's agentic harness and the evals that matter
- SHSonya Huang
Let's talk about harnesses.
- ALAaron Levie
Yes.
- SHSonya Huang
How does the, uh, how does the Box agent work?
- ALAaron Levie
Great transition to harnesses. Okay. [laughs]
- SHSonya Huang
[laughs]
- ALAaron Levie
Speaking of calculators, let's talk about harnesses. Okay. Th- this is back to the kind of neolab, uh, kind of applied, uh, applied layer. Uh, there's, there's a bunch of things that, that we know about our file system, permission structures, our search, our, our search engine that, that, you know, certainly, and by all means, we actually would love the, the... We'd love all of the labs to train on our understanding of this, 'cause it would only make external agents, you know, as good as possible.
- SHSonya Huang
That are using Box, yeah.
- ALAaron Levie
Yeah, like, like we, we, we always talk to labs like, "Hey, we'll give you as much data as you want about, about kind of how the system works." But, but let's just say, like that aside, we, we have a lot of depth of understanding of what do people do in Box? How do they search Box? How do they decide when they look through 10 files which is the one to go pick? How do... What, what is the, in their, their internal kind of calculus or heuristic on, on sort of figuring out the most relevant document to look at? So we know all of that, and we basically, you know, have, have built, uh, an agentic harness that, that, you know, attempts to kind of understand that set of domain understanding about our system. Uh, it, it obviously has access to our search system, our file system. It has a, a bunch of kind of mechanisms for just pulling out just the text of a document, just pulling out chunks, uh, from the document, uh, doing embeddings on the document on a fly. So there's a set of kind of tools that it can use, um, and effectively it's a harness for asking questions of a large data set. So in my Box account I have, I don't even know the latest number, but on the order of tens of millions of files. Um, just because it's like, it's every- everything that has ever kind of accumulated over, over 20 years. But I can now ask any question of, of all that data set using the Box agent, and it, it goes around, it does, it does a multiple searches in one, it re-ranks it. It then very quickly sort of pulls out the most relevant information, then it'll in some cases read the full document.
- SHSonya Huang
Yeah.
- ALAaron Levie
You know, do- does all the steps. And, and then we compare that against like, well, what if we just gave, you know, Claude our API or gave OpenAI our API and we see like meaningfully better results on accuracy and latency because, again, we kind of know exactly how to-
- SHSonya Huang
Yeah
- ALAaron Levie
... how to tune it for our workflows. Uh, so that, that's effectively the, the harness that we built out.
- SHSonya Huang
And then what, what evals matter the most to you?
- ALAaron Levie
Um, so I have a couple, I have a couple like, you know, just funny personal ones that I just keep track of-
- SHSonya Huang
Okay
- ALAaron Levie
... of like my own use cases.
- SHSonya Huang
Uh-huh.
- ALAaron Levie
But we do, we have, I don't know, hundreds of different tests that we do on every single model. Uh, we actually have two evals at the moment. One is, uh, we put out a thing called the Complex Work Eval, which is a, which is a set of domain-specific work in life sciences, financial services, public sector, uh, tech, et cetera. And it's, it's kind of exactly what you'd think of as a, as a document-centric eval. So given these five documents and this set of problems, like what would your answers be? And we test every single model against those with our agent. And then we have a holdback eval, which is, um, uh, actually the, the first one is holdback also, but, but the second one is just like then our Box instance and how Box employees use, use their data, and then we eval every model again on that. Um, so we're able to kind of roughly keep track of, of, of all the incremental progress. Like we, we see when things move by half a point, uh, in terms of model improvement. Um, and then we roll out sort of default models, um, based on different kinda cost and accurate, uh, accuracy thresholds, and then we let customers also choose any model they want from effectively our model garden.
- 27:23 – 29:25
The state of the model race
- SHSonya Huang
What's your current view of the race and like where all the horses are in terms of model performance on your use case?
- ALAaron Levie
They more or less closely correlate code with one exception, which is actually in some of our use cases, Gemini, uh, is disproportionately better than, than what you would see from coding. And, uh, and, and it might be, you know, sort of just better tool use, um, you know, given, given kind of, you know, the, the, the Gemini ecosystem and what they need to build for. Um, it, it, it solves, you know, kind of a, a, a strong set of sort of general knowledge work use cases as well. Um, but I think mostly correlating to, to kind of code. So Fable 5.1 was clearly kind of, you know, state-of-the-art, um, and the best model that, that we've seen. There's obviously rumors about other models, and so we'll see how the kind of race-
- SHSonya Huang
Yep
- ALAaron Levie
... you know, kind of, you know, continues on this, on this front. But, but basically by and large, like when you look at, uh, GDP Val, um, uh, Mercor has their Apex, uh, Eval, uh, these things will all generally follow the, the coding, uh, models. And, and so I think it's, we're just neck and neck on like Groq, Muse, the Fable class and, uh, you know, GPT-5, 6/whatever they're building next. Like these are just like it's a total race right now.
- SHSonya Huang
And do your customers typically express a preference on which model they want to use or do they just use your default?
- ALAaron Levie
So they, they, you know, kind of by volume they use our default because it's just easy and, and it works extremely well. And it's tuned for, uh... There's a, a few ways our, our kind of agent manifests. So like the m- way that you'd most commonly experience it as an end user is you would just be searching and, and kind of, you know, asking questions of your data. The, but by volume, the, the volume of tokens tends to go through more of our workflow agents or data extraction. That's where actually you're, you have customers actually doing evals, and they're basically saying, "Okay. I want to really make sure that at this cost profile I can get 98% accuracy on data extraction." And that's a place where like we'll have an FDE that goes in and, and helps you, you know, helps understand your data environment, tests against five different models-
- SHSonya Huang
Yeah
- ALAaron Levie
... and then you're just basically at the mercy of, of the eval.
- 29:25 – 32:34
Open-weight model adoption in the enterprise
- SHSonya Huang
Yeah. What are you seeing in terms of the adoption of open-weight models-
- ALAaron Levie
Yeah
- SHSonya Huang
... in your customer base?
- ALAaron Levie
Um, so h- probably higher than people think, lower than what enterprises actually want, and much, much, much, much, much lower than what it'll be in five years. So, so like some mix of that would be the kind of message. Like-
- SHSonya Huang
And is it primarily cost that's driving that decision?
- ALAaron Levie
I, I, you know, I have to probably attribute 30-plus percent to just kind of the, the sexiness of, of-
- SHSonya Huang
Yeah
- ALAaron Levie
... of like, I'm-
- SHSonya Huang
Like I wanna try GLM.
- ALAaron Levie
Yeah, yeah.
- SHSonya Huang
Yeah.
- ALAaron Levie
I think there's, I think there's that. Like I've heard-
- SHSonya Huang
Yeah
- ALAaron Levie
... I've heard, you know, CIOs of Fortune 500 companies say, "We're, we're playing with open source here."
- SHSonya Huang
Yeah.
- ALAaron Levie
And, and I look at that and be like, well, I know for a fact like, like Gemini or Muse would've been just fine at that particular cost profile that you're trying to do, or probably even like, you know, y- you know, 5, 6 Luna or Terra or whatever, whichever, you know, one had the, had the crazy discounting they just did. Like it probably would've been totally fine. But you wanna be able to be like, "Okay, I'm a little hedged." Like it's cool, like, like we're that phase still. Um- Over time, I think it stands to reason that, that you'll see meaningful different costs because you'll be able to peel off workloads that, that just only make sense at, at sort of, you know, grinding down to the cost of inference, in which case open, open weights will have the sort of economic advantage. Right now there's, you know, this challenge of like sometimes it's more token inefficient, you know, sometimes like randomly... Like I've heard stories like randomly it'll just like s- speak Chinese like-
- SHSonya Huang
[laughs]
- ALAaron Levie
... like mid, mid chain. So you're like, "Okay, well, that'll be weird for a bank." Um, so.
- SHSonya Huang
That's funny.
- ALAaron Levie
So like we need to like probably work on some of those things.
- SHSonya Huang
Yeah.
- ALAaron Levie
But long term I think it has to be, you know, the case that, that you're gonna, you're gonna peel off those workloads. Uh, you know, one, one of the more interesting posts I think on this that I totally subscribe to is Jesse at, at Decagon. Do... You probably read that post of like, of like this paradox of like we're gonna have... Like you're gonna see Claude just, you know, continue, you know, go exponential. But what's gonna happen is each use case that kind of matures you can peel off to open source, and once you have kind of stability in that use case it starts to make sense to veer it toward an open weights model, assuming one of two things is true. One, that, that it's actually literally cheaper, or two, having some post-training gets you x, x percent more performance.
- SHSonya Huang
Yeah.
- ALAaron Levie
And, and so I think you will just like be in a reality where we will, we will... And this is gonna be very confusing probably for like the press more than people in the Valley 'cause you'll be like, "Wait a second, like the revenue of Anthropic, OpenAI, et cetera, are like off the charts, but somehow open weights is like also growing exponentially." And you're like, "Well, how is this..." Like how is OpenRoute and-
- SHSonya Huang
The pie, because the pie is growing so fast.
- ALAaron Levie
Yeah.
- SHSonya Huang
Yeah.
- ALAaron Levie
And it's like the pie is growing so fast.
- SHSonya Huang
Yeah.
- ALAaron Levie
But what's happening is, is actually there's, there's an interesting duality. It's, it's, it's not even just like, like rising tide lifts all boats. It's like no, no. Like we either use Fable or, or, or Five Six for orchestration and then we farm out all these long tail tasks to a cheaper model. Um, or the opposite is true. Um, like you use, you have some kind of orchestration agent that like by default does the cheaper stuff, but occasionally sort of sees some- something that is just way too hard and then it pops it out to, to one of these heavier models. And so you might have blended 50% spend on each, but 10 times the amount of tokens, you know, on the open weights model. And so like everybody's kind of winning, but there's an, uh, there's an interplay between
- 32:34 – 37:29
Memory, continual learning, and what belongs in the weights
- ALAaron Levie
why they're winning.
- SHSonya Huang
I'm curious about how you think about memory and, uh, customization or personalization and where that's gonna go. 'Cause it seems like today the dominant architecture is kinda like RAG-based system still.
- ALAaron Levie
Yeah.
- SHSonya Huang
And like you can get fancy on the RAG, but it's-
- ALAaron Levie
Yeah
- SHSonya Huang
... still context lookup where the weights themselves aren't fundamentally changing.
- ALAaron Levie
Yeah.
- SHSonya Huang
It seems like, I mean, if, if I listen to my friends at the labs, like continual learning and this idea of like the model's weights should adapt as it gets to know you.
- ALAaron Levie
Yeah.
- SHSonya Huang
Um, we had Ngram on the podcast. I don't know if you, if you know Dan, like-
- ALAaron Levie
Uh, I, I just got introduced to him. I listened to the podcast, so-
- SHSonya Huang
Oh, amazing
- ALAaron Levie
... uh, I would have loved to been-
- SHSonya Huang
[laughs]
- ALAaron Levie
... the fourth person in the room or fifth.
- SHSonya Huang
Amazing.
- ALAaron Levie
Yes.
- SHSonya Huang
Yeah, they, they... Like I think they're working with customers that... to, to help basically bake in some of the context into the weights themselves.
- ALAaron Levie
Yeah.
- SHSonya Huang
How... What direction do you think this is gonna go?
- ALAaron Levie
You know, you're catching me at a time right before I'm actually doing my call with Dan, so, [laughs] so I, I wish I, I could have talked to him first and then I'll have like a way more eloquent answer. Um, I'm like extremely fascinated by, by the approach. Like I, I have no reason for not wanting to... it to work and, and exist. We live in a world at Box where we see these sort of the, the, the high degree of complexity on permissions and access controls and data that, that tends to be sort of the rub on a lot of these types of, of approaches. And I'm, I'm gonna put Ngram aside 'cause like I'm sure they've already thought this through, so I'll, I'm gonna talk more generic philo-
- SHSonya Huang
Yeah
- ALAaron Levie
... philosophically. I think sometimes you will talk to a researcher that, you know, sort of imagines the world working the way they work, which is like, "I'm a researcher. I have access to everything, and so if I had a model that was trained just on my world, this would be amazing." And then you're like, "Let me introduce you to a lawyer."
- SHSonya Huang
[laughs]
- ALAaron Levie
And the lawyer like has this tiny little, you know, access point of just like the five projects they're working on because somebody right, you know, one door over is working on the competitive project to, to a, a, another company in the space that they can't have any sort of overlap with, with what they see or what they know and there can't be a single document that passes between those two walls. And they have to be these kind of hard barriers. So, uh, you know, so like sure, like you could still, you could still now train a model just for that one user, but like what happens if every single day they get added or removed from something that, that adds important context to, to sort of what they need to understand? And, and, and, you know, again, like I think there's gonna be probably breakthroughs in continual learning that, that sort of all resolve this. But like this is why previously it was just like, you know, there was no way you could pull this off five years ago 'cause it would be insanely expensive, impossible to kinda wrap your head around, around how those access controls are supposed to work. But, you know, obviously with like, as the cost curve goes down, as open weights, you know, get, you know, cheaper, smaller, faster, better, um, I think this becomes super interesting. One thing on, on the podcast that I found very fascinating, and I just need like a T chart honestly, w- is just like, you know, like what is the decision point of what goes in context and what goes in the weights?
- SHSonya Huang
Yep.
- ALAaron Levie
Like, you, you have to be a little bit thoughtful about like, where is the, the massive performance gain that you get by, by baking in the weights ver- and there's probably some like incredible like, like, um, calculation of like, like when the rate of change of the data is not, you know, so far, but the upside of the, of the weights, you know, you know, dramatically change the accuracy of the model. Like, you know, s- you'd have to kinda land on some sort of, you know, rubric like that.
- SHSonya Huang
I mean, if you could wave a magic wand, it almost seems like, and I think Karpathy has said this in some prior interviews, like if you could almost remove all the memorized information from the models-
- ALAaron Levie
Yeah
- SHSonya Huang
... and just have, have it encapsulate the specific reasoning capabilities, like the ethos of how we do things-
- 37:29 – 44:55
Box Labs and systems of record in a world of agents
- SHSonya Huang
Makes sense. Okay, so you hinted at the beginning that there's a Box Labs. What, what type of work is Box Labs doing?
- ALAaron Levie
Uh, so i- it's the equivalent of Box Labs. Like I don't know if we've used capital L yet. Um, but, uh, but basically, you know, it's our, it's our applied sort of AI team, and we just-
- SHSonya Huang
What research areas are most interesting to your, your te- team right now?
- ALAaron Levie
Yeah. So, so the, the, the biggest areas that, uh, that the most sort of research kind of, or, you know, of the continuum of engineers, there's, you know, some cluster that is sort of more on the research bent. And, uh, of that cluster, the things that we've spent time on still again is at the kind of applied layer, but, but it's a lot around how do you, how do you take agents and make them, you know, another 10 points of accuracy improvement, uh, given X problem? Um, so what, what is the, what, you know, how do you build a map of the, of the problem set g- you know, with a given set of data to sort of best execute on that task? So we spend a lot of time on, on that style of work. We have a team, for instance, working on how do you do effectively at the a- agent level, at the harness level, you know, some form of kind of auto research on, on sort of hill climbing on accuracy of answering questions or set- sets of problems on, on given a, a kind of a, a set of client data. So if you're a bank, and you have a bunch of loan documents coming in, and these are like 100-page documents, like whether you're getting 70% accuracy with an off-the-shelf model or like 97% is like basically obviously a world of difference in can you actually go and automate that process. So somehow you have to hill climb from the base model to the 97%, and there's a lot of work going into the system to, to basically pull that off.
- SHSonya Huang
Maybe zooming out, what do you think of as, and this can be a Box question or a non-Box specific question-
- ALAaron Levie
Sure
- SHSonya Huang
... the role of systems of record in a world with agents?
- ALAaron Levie
[laughs] Yes.
- SHSonya Huang
And I'm sure you saw some of the Twitter discourse on like, you know, every software company's trying to sell me their own agent right now. I don't want another agent from them. I want, I want their system record to work well with my agent.
- ALAaron Levie
Yeah.
- SHSonya Huang
And so like how do you think about that?
- ALAaron Levie
#ClaudeForce.
- SHSonya Huang
Yeah. [laughs]
- ALAaron Levie
Um, so, uh, I, I think-
- SHSonya Huang
Catchy name, by the way. Very catchy.
- ALAaron Levie
I, I mean, literally it's like one of those things where, where like the first three minutes you're like, "Man, that seems funny." And then four minutes later, especially when you see the stock, you're like, "Ah, brilliant move." [laughs] Like, like, "This is great. We're doing this." Um, and then when you, uh, s- I think somebody, somebody actually, uh, said this the best. They were like when they heard Matthew McConaughey say it out loud, that was like really the, that sealed the deal.
- SHSonya Huang
That was the aha moment? Yeah.
- ALAaron Levie
That was the aha moment, and it's like, man, he can sell software. Like it's actually incredible. Like his voice is so good for selling-
- SHSonya Huang
[laughs]
- ALAaron Levie
... systems of record and agents. Um, if you're in like, like our sort of, you know, contemporary group, uh, y- of like you, you built a SaaS platform, and you have some set of data and workflow that, that, you know, your customers kinda operate in, there's, there's effectively two things you just have to do, and I, I think anybody attempting to do one over the other is just gonna lose. Um, you have to build an agent that is insanely great at your product. Like that agent has to be, you have to provably be like- 10 or 20 points better than an off-the-shelf agent at, at using your system. Not because it's hobbled the other side, it's just like you are so eval maxed and you're, like, so tuned to your particular workflow that, that you can improve your system. You have to have that. And you probably, because you understand your domain, unless you're, like, totally asleep at the wheel, you probably have use cases that no one has, has sort of thought to build products around because you, you talk to customers every day and you see what they run into, and you're like, "Oh, we could just, like, have our agent go do that for you." Like, I've had at least a dozen ... I mean, so I probably talk to a couple hundred customers a year in a, in variety of capacities. I've had at least two dozen times where the customer has a use case that is, like, a breakthrough moment for me of like, "Shit, that would be actually totally insane." Like, uh-
- SHSonya Huang
What's an example?
- ALAaron Levie
Um, unfortunately, since you put me on the spot-
- SHSonya Huang
[laughs]
- ALAaron Levie
... I don't know if my example will pay off the, the level of excitement I just had, uh [laughs] because it'll be-
- SHSonya Huang
Sorry for putting you on the spot.
- ALAaron Levie
No, no. I, I ... It's just like I think [laughs] like the thing I was thinking of is just, like, I think it's gonna be a womp womp for the podcast. But, um-
- SHSonya Huang
[laughs]
- ALAaron Levie
... uh, there was [laughs] there was basically, uh, a customer had this idea of they wanted an agent in the background trying to sort of figure out when documents sort of met their governance policies, and, like, does something need to go into some kind of archive or something into some kind of legal hold or whatnot.
- SHSonya Huang
Oh, cool.
- ALAaron Levie
And, and [laughs] see? Exactly.
- 44:55 – 48:00
Will chat be the dominant UI for enterprise AI?
- SHSonya Huang
Super interesting. Um, maybe related, let's talk about product UI.
- ALAaron Levie
Yep.
- SHSonya Huang
Big, you know, generic chatbot, chat box agent. Um, is that gonna be the dominant UI for how people are using AI, AI in the future, especially when it comes to the application layer?
- ALAaron Levie
Th- this is where, why I think the applied layer, you know, has so much ru- uh, uh, room to run, is because, uh, probably the, the sort of, you know, universal chat system that you ask a question to, you get an answer back or it does sort of some work in the background, I think it's gonna be, obviously, like, that's gonna be a mainstay. That, that, that UI will always exist. It'll be incredibly powerful. The horizontal products will have it. The vertical products will have it. Everybody will have it. It's just like your product has a search box. Obviously, it does.
- SHSonya Huang
Yeah.
- ALAaron Levie
So, so that, that's always gonna be here for these sort of one-off asks of an agent or go find this thing or answer this question or produce something for me on demand. But most of the enterprise, uh, is sort of made up of these processes and workflows that are kind of just happening behind the scenes. Sometimes they're happening with computers, and computers are running these things, or sometimes they're happening with other people that are doing these things, or sometimes they should be happening with people, but you could never afford to have them happen with other people, so they just didn't happen. And so that's, that sort of is a slightly different kind of m- you know, metaphor than, than a chatbot where you ask a question and it, and it comes back wi- with an answer. That's like, okay, I, I kind of want agents in the background to do things for me, read every contract, look at every log, triage every security incid- incident. And then instead of me chatting, maybe I'll chat as a, as a means of doing kind of a catch-up. I want a dashboard. I want a workflow. I want a queue. I want a task list. So then the, the, the, the challenge becomes, well, does the horizontal product kind of take on every one of those, those components and e- and manifest every one of those experiences in one? In which case, I think you'll start to be like, "Man, that thing is like really like a... That's pretty heavy," and like, and then we'll start to like be like, "Oh, this is no longer this simple, easy, delightful thing anymore." So then the vertical players actually like, like they actually sort of understand the process and can manifest all the right buttons and tabs and the names of the things for that particular workflow. So I think it puts-- I think as you have agents that are doing more work in the background, doing more, you know, async work that is just like I farmed out a bunch of agents to review things as they happen or whatnot, that leans more toward the applied companies that can understand those workflows, that can understand those processes. I think you're gonna have these in every field. You're gonna h- certainly we, we already know that, you know, w- how they're gonna look in legal with Harvey, Lagora, et cetera. Um, we are seeing them start to emerge in areas like security. We've seen them start to emerge in the long-running kind of coding agents with cognition and factory. Um, so I think that, that will be, you know, one of the bigger kind of applied AI sort of use cases. And ultimately, like in five years from now, I would bet like 90% of all tokens in the enterprise are things that a user never kicked off and they just see a result. They just s- they see a task show up, and they have to go review it, and it's just like, it's
- 48:00 – 54:31
Why coding diffused fast and the rest of knowledge work hasn't
- ALAaron Levie
just happening.
- SHSonya Huang
Yeah. Yeah. Makes sense. Okay, let's talk about AI diffusion. Um, coding agents, it was like, you know, boom, January 1st, 2026 happened, and like the fastest diffusion of anything into the economy-
- ALAaron Levie
Yes
- SHSonya Huang
... we've ever seen has happened. Um, the diffusion of, you know, the rest of the AI magic into the rest of our jobs seems like it's been a lot slower.
- ALAaron Levie
Yeah.
- SHSonya Huang
What are your thoughts on that, and where are the areas where you think we're gonna see faster diffusion, and how is that gonna happen?
- ALAaron Levie
Yeah. So you always have to kind of compare and contrast coding versus everything else to really understand the, the dissimilarity. Um, so in coding, essent- and this is back to the kinda utility point on, on, you know, slop. Like the utility of code is almost 100% represented by the amount of text that you can generate. Like, like all -- obviously an insane amount of value went into the text, but, but like, and then like knowledge and expertise in meetings and everything, but like ultimately like the text is the thing that produces the, the program that is actually the thing that you're trying to do. You know, if you could have the world's greatest programmer, like, and never had to sleep, never had to eat, they, they could intuit what to build, and they could just sit on a computer all day long, your value creation would be 100% correlated with how many hours they could sit at that computer and type code. Like lines of code, ideally, ideally good code, is the most thing that will be correlated to whether you produced software that, that people wanted. So it's all text. The models are hyper trained on these. Everybody in AI labs treat coding as, as a competitive benchmark to constantly try and exceed. They, they get to do their own evals on it every single day because they are the ones coding the, the models themselves.
- SHSonya Huang
Yep.
- ALAaron Levie
And it's the most technical audience of all time, where when they deploy an agentic system and they run into a, uh, either a bug or a problem or like some MCP server comes back with like connection invalid-
- SHSonya Huang
They fix it themselves
- ALAaron Levie
... they know how to triage the problem. They don't call IT. They're just like, "Oh, yeah. No, I didn't open up that port. Sorry. I'll go fix it." So that's like five things. Oh, and maybe like the sixth, like it's just like a very, very high-paying vertical that, that like it's just like automatically valuable if you could get 10% or 20% productivity gain, let alone five X productivity gain. So, so, so take those five or six things that coding has as, as sort of beneficial properties to sort of automation, then compare that to, you know, you know, every other form of knowledge work, and you'd probably have like a histogram. And I don't know if anybody's published this. Maybe you can. Like the similarity to coding and like, like what are, what are the domains that like, like start to sort of, sort of, you know, look closer, like, like less and less like coding as you, as you kind of scale out. And it's like, okay, well, lo and behold, legal is kind of interesting because, because like there's a lot of value creation to somebody sitting at a computer reviewing legal documents, writing legal documents, like processing large amounts of information. Okay, so that's kind of blowing up. And then you kind of go down the list. Now let's take something like s- like a sales rep, okay? So much farther down the list in terms of, of sort of likeness. The sales rep's value creation is basically convincing an external customer to buy software or technology or a Caterpillar truck, you know, uh, fr- from them. That is the value creation to the, to the economy of the sales rep.
- SHSonya Huang
Yep.
- ALAaron Levie
And so let's say we brought the world's best automation to them. Like first of all, again, they'd have to like figure out how to technically wire it up. They'd have to make sure they give it all their data, all these kind of things. But no matter what, the, they're, they're still rate limited and constrained by like, did the customer respond to them? Do they wanna meet? Can they meet next Tuesday or j- can they meet today? Like, does the customer have budget? All of these other things. So that, that's maybe the entire continuum right there is like one is like ev- I mean, of knowledge work. Like obviously this is not even, you know, touching the, the sort of working with, with atoms. So that's a continuum, which is on one end you have somebody rate limited by so many external factors. On the other end, you have somebody who could sit at a computer all day long and just type, type text, and that is your ability to basically automate things is, is that continuum. So for the real world, we have to basically bring intelligence to these workflows in ways that are, that sort of somewhat feel like the shape of their workflow, somewhat feel like the shape of their work, and then find a way to deliver the change management, deliver the implementation, get data into a format and into an environment that actually works with these systems. Like Asterisk, like one of the other big things is like if you go to most engineers in 2026, like maybe like minus two months ago given the, given the latest phenomenon, but like, like the code's in GitHub. You just like connect it to GitHub. Like, like remember there's this period where like your... when you launched a coding agent, like there was no like sign up or register. It was just like, give us your GitHub. That doesn't exist in knowledge work. There's no like, give us your GitHub for knowledge work. There's like-
- SHSonya Huang
There's give us your Box.
- ALAaron Levie
Well, w- Box customers have a, have a much easier time with all this. Unfortunately, like we're only 1.3 billion in revenue run rate, so like, like that means there's a lot of people not using Box, and so what are they using? Their data is in on-premises systems, legacy file shares, legacy infrastructure, you know, e- enterprise environments that don't talk to agents particularly well. So just think about that sort of distinction between a, a, you know, implementing coding agents versus everything else. Um, even, even something again you'll sort of fall asleep about is like access controls in the enterprise are totally different. I actually totally forgot that, that point about coding. In coding, you get access to basically most of the stuff ever relevant for your job. In knowledge work, you're like, you're like, "Hey Sally, can you open up that, that sort of file share for me? Can you open up that, that, you know, sort of project because I didn't get access to it?" How do you make sure the agent has access to those set of things? All of that work has to get done. So the thing I think we have to prepare for is two things. One, Silicon Valley has to prepare for diffusion taking a lot longer than they think or the, than, than, I'll say we but I really think they because, because I like, I know how long it'll take. And then the second thing is, is the good news is this is sort of all correlated to the applied layer-
- SHSonya Huang
Yeah
- ALAaron Levie
... value creation. Like the, because the, the companies that will just have the patience, the, the sort of the j- the full sort of domain expertise, the, the just the sheer work ethic because it's not like everybody's just f- from the floodgates. Like you have to go and, and just, you know, pound pavement and get out there. That will be the applied layer. So I think this actually represents, you know, a trillion dollars of applied layer AI value is, is actually how do you go get the technology to the lawyer or to the sales rep or to the life sciences researcher or to the person that runs the customer support team. That's all sort of opportunity right now
- 54:31 – 1:05:20
Staying wired in, making a company AI-first, and what it takes to win
- ALAaron Levie
that exists.
- SHSonya Huang
Awesome. I'm gonna close by asking some advice for other-
- ALAaron Levie
Yes
- SHSonya Huang
... founders. Um, maybe let's start with founder advice and then company building advice. On the founder side, seems like you are in every AI cap table, you know every cool new company, like you know, you know ngram. You know, um, how did you kinda get yourself in the middle of the AI conversation?
- ALAaron Levie
There's probably two, two parts. One, I was just very well primed for it. Like working with unstructured data for 20 years, you just like can instantly see the, the, the benefit of agents on that. So like it honestly like took longer than I would've wanted l- that, that we, we got to have this conversation 'cause we, we tried to have this conversation, you know, eight, nine, 10 years ago.
- SHSonya Huang
Yeah.
- ALAaron Levie
And now it's finally happening. So, so first of all, just super well primed. Our product sort of, you know, shape and, and what people do with our product already lends itself extremely well to agents, so like obviously we, we had to bet the company on that. And, and then lo and behold, we had the positive feedback loop of customers actually saying, "Yeah, that would actually be very powerful if I could go read every document and answer any questions." So that, that's the first and certainly by, you know, biggest by kind of a factor of 10. And then the other is just like I'm extremely fascinated by the technology, um, and it's just fun. It's like it-
- SHSonya Huang
Where do you learn about it?
- ALAaron Levie
Um, your podcast.
- SHSonya Huang
[laughs] It's a daily one.
- ALAaron Levie
Um, uh, Thor Kesh's podcast. Um, uh, the-
- SHSonya Huang
Twitter.
- ALAaron Levie
Yeah, I, I mean, un- uh, probably unfortunately for brain cells, like it's probably 95% Twitter.
- SHSonya Huang
[laughs]
- ALAaron Levie
Um, I have a routine where like at the end of each night I just go through the feed and it's just like I do, I like, I, I look like some sad meme probably of just like I'm just scrolling and scrolling and scrolling and like attempting to like triangulate all the information.
- SHSonya Huang
That's amazing 'cause it is, it is the Global Town Square for AI.
- ALAaron Levie
It is. It is. And um, I wanna send an emergency alert to like everybody who's like a sophomore or junior in college and just be like, "Follow these 20 accounts-
- SHSonya Huang
Yeah
- ALAaron Levie
... on Twitter."
- SHSonya Huang
Yeah.
- ALAaron Levie
And, and also join Twitter because like this will just help your career. You're either like a year ahead or a year behind-
- SHSonya Huang
Yeah
- ALAaron Levie
... simply based on your feed, and my feed is like so, so wired in.
- SHSonya Huang
This is the number one advice I give to people-
- ALAaron Levie
Oh, really?
- SHSonya Huang
... when they're asking how to-
- ALAaron Levie
Yes
- SHSonya Huang
... get current on AI.
- ALAaron Levie
Yeah.
- SHSonya Huang
It's like follow, follow these 100 accounts.
Episode duration: 1:05:21
In the uListen app, the transcript lights up word by word as you listen. Get uListen
Transcript of episode NE4CLThMPGU
