Aaron Levie: Why Enterprise Buys Outcomes, Not Models

1. 0:00 – 1:15

   Intro

   1. ALAaron Levie0:00

      ... wait a second, if we could use AI to automate more, we can build more. If we could build more, we could lower the cost of things. If we can lower the cost of things, then we can actually lift up anybody's lifestyle right now.

   2. GTGarry Tan0:10

      I think that we're in the middle of the revolution, and the revolution does not have to be, uh, Black Mirror. It could be something that is driven by Jevons paradox-

   3. ALAaron Levie0:19

      Yeah.

   4. GTGarry Tan0:20

      ... driven by abundance for everyone, and that's certainly the, uh, the timeline we wanna be on, so...

   5. ALAaron Levie0:25

      That's the future I'm betting on.

   6. GTGarry Tan0:34

      Welcome back to another episode of The Light Cone. I'm Garry. This is Jared, Harj, and Diana. Uh, we're partners at YC, and collectively, we've funded companies worth hundreds of billions of dollars. And today, we have a really awesome guest, Aaron Levie-

   7. ALAaron Levie0:50

      Oh, thank you.

   8. GTGarry Tan0:51

      ... of Box. (laughs)

   9. ALAaron Levie0:53

      (laughs)

   10. GTGarry Tan0:53

       He's-

   11. ALAaron Levie0:53

       Love that intro.

   12. GTGarry Tan0:54

       Yeah. (laughs)

   13. ALAaron Levie0:55

       (laughs)

   14. GTGarry Tan0:55

       Aaron, you're o- one of the best product CEOs out there.

   15. ALAaron Levie0:58

       Thank you.

   16. GTGarry Tan0:58

       Public company as well.

   17. ALAaron Levie0:59

       That's what I w- that's what I write in my Wikipedia.

   18. GTGarry Tan1:01

       Yeah, yeah. Uh, that's how I classify you. We're in the middle of the AI revolution, so, uh, how are you feeling?

   19. ALAaron Levie1:08

       Oh, pretty good.

   20. GTGarry Tan1:09

       Yeah. (laughs)

   21. ALAaron Levie1:09

       (laughs) Um, it's a, it's a good time to be in, uh, in, in software right now. Um, so yeah, feeling pretty good today.

2. 1:15 – 4:44

   Why the GPT wrapper was a bad meme

   1. JFJared Friedman1:15

      Something we've been speaking about for a while, which I think we probably agree on is that the ChatGPT wrapper was like a bad meme, and that actually there's, like, lots of value, and always has been, in building apps on top of these foundation model companies.

   2. GTGarry Tan1:27

      In fact, the opposite might be true.

   3. JFJared Friedman1:28

      Yeah. Which is gonna be more valuable-

   4. ALAaron Levie1:30

      Yeah.

   5. JFJared Friedman1:30

      ... in 10 years time, right?

   6. ALAaron Levie1:31

      Yeah, I think so. So, it's interesting. So, um, there's, there's a l- like pr- 2% truth in, in the meme and then 90% n- not truth. And so, I mean, PG, you know, with the sort of wedge theory, is like actually you do want something that is, is sort of a simple product that finds a little wedge and then you expand from there. You know, in the early days of, of cloud, if you were to be building software that let you manage documents and data, th- you would've been like, "Well, that's a wrapper on Amazon." And, and it was a total misunderstanding of the, the entire scale of software you have to build to make the storage bucket be useful in, in a particular application. So, on the, on the wrapper conversation, the exact same thing is true, which is how much software do you need around the workflow and the proprietary, you know, sort of business logic and the data that the customer brings. That's actually the value, not the, not just like what is the, the set of tokens that are coming out. Where it's a little bit true why, why startups should be, you know, at least think a couple steps ahead is you probably don't want to be something that just ChatGPT would incorporate. So it's less that the model, uh, will incorporate your value proposition, it's more if there is, you know, an a- i- if the model provider also has a consumer scale application. Like you don't want to be right in the way of something that ChatGPT will just fold in directly into its functionality. So in that case, I think you have to be, you know, sensitive to being kind of, you know, a "wrapper".

   7. JFJared Friedman2:46

      How do you separate out stuff that's gonna get incorporated into the model from stuff that won't? 'Cause I feel like the hard part with that often is we don't know what the next models are gonna be capable of.

   8. ALAaron Levie2:56

      Yeah.

   9. JFJared Friedman2:56

      And there's this general sense of, oh, well, like anything could theoretically be-

   10. ALAaron Levie3:00

       Yeah.

   11. JFJared Friedman3:00

       ... incorporated into the next model if it's powerful and, you know, approaches some version of generalized intelligence.

   12. ALAaron Levie3:04

       Yeah, I mean, I, I look at all this through, um, just the B2B lens, which I, I know that then probably you just lost half the podcast listeners as well. (laughs)

   13. JFJared Friedman3:13

       (laughs) No, B2B stuff is cool these days, actually.

   14. ALAaron Levie3:14

       Okay, okay, okay, cool.

   15. JFJared Friedman3:14

       Yeah, yeah.

   16. ALAaron Levie3:15

       On the B2B front, uh, it, to me, it's a l- uh, it's a, it's a simpler question because, um, an enterprise doesn't want a model. It wants an outcome. Uh, it wants an outcome of customer support, you know, conversations being answered or, um, you know, healthcare, um, uh, you know, transcription going into a, an EHR system, uh, or an automated workflow of reading documents and contracts and plugging that into a, a contract workflow. So the model getting more intelligence is actually usually a better thing for anybody building software in those use cases 'cause then you're doing less in terms of, of, of, you know, hacking your way into the model because it's sort of insufficient at, at solving that particular problem. But what the customer actually wants to buy is like, "I need software that will plug into, you know, it my ERP system, that will plug into my support system, that will power the workflow, that, that lets the customer do a password reset." Like that's actually the, what the customer wants to buy and what the model is doing is, is sort of, you know, really abstracted from the ultimate customer value proposition. So I think as long as you're, you're building, uh, you know, software that, that really can deliver that full outcome to the customer and, and, you know, two years ago the initial wave of these, these use cases started to emerge. And I think the, the companies that will do best, you know, are ones that, that realize that, that you need to abstract the model a- away as much as possible from the ultimate value proposition. And, and then you just incorporate all the model updates as quickly as possible for your customer. And again, they just buy the outcome of customer support but it's just getting better and better every time there's a model improvement. So...

3. 4:44 – 8:38

   Enterprise users just care about getting workflow done

   1. SPSpeaker4:44

      I wonder, uh, maybe one good analogy is-When you were building Box for your customers, they didn't care what was the underlying database or cloud-

   2. ALAaron Levie4:53

      Right. Right.

   3. SPSpeaker4:53

      ... or what was the networking year or all the hard drives that were running.

   4. ALAaron Levie4:56

      Yeah.

   5. SPSpeaker4:56

      It was all about the end user experience at the software level.

   6. ALAaron Levie4:59

      Yeah.

   7. SPSpeaker4:59

      And the analogy to today is the end users of B2B AI workflows don't care whatever model is or how it does it-

   8. ALAaron Levie5:06

      Yeah.

   9. SPSpeaker5:07

      ... but that it ultimately does the workflow.

   10. ALAaron Levie5:10

       Yes. Yeah, I think, uh, that- that- that is absolutely the conclusion. Um, you'll- you'll often, you know, get some idiosyncrasies in different organizations where they do care, okay, where- you know, where's your data center hosted or, you know, what- what's the- what- what's your infrastructure provider, but that's of small minority. In AI, I think we're going to go through a temporary period where you do see differences in the models for anybody that- that has, you know, a discerning, you know, s- set of- of- of skills on this front. So, you know, you can t- see this in, like, Cursor, you know, with- with Anthropic, right? Like, people like that combination and they- and they can sense the differences of- of the output. But if you, you know, fast-forward five years, I- I think you'll see a convergence of- of basically models and intelligence to the point where you- you wouldn't really distinguish the- the quality levels that much for 90% of- of business use cases.

   11. JFJared Friedman5:55

       I- it's definitely interesting to see how especially developers have developed different preferences for different models.

   12. ALAaron Levie6:00

       Yeah.

   13. JFJared Friedman6:00

       Like, I remember at our AI retreat a few weeks ago, um, Anthropic and Claude has also emerged as, like, the preferred LLM to, like, orchestrate your agents. Like, if you have multiple agents and you want sort of the LLM to intelligently call the right ones, people seem to prefer Claude for that. What do you think is gonna happen to the model companies themselves in this world, though?

   14. ALAaron Levie6:17

       Probably everybody needs to update their understanding of what- what a model company is-

   15. JFJared Friedman6:22

       Yeah.

   16. ALAaron Levie6:22

       ... just in general. I actually think there's very few model companies. There are sort of AI companies that have model, you know, frontier model labs, but- but increasingly they're selling software to either consumers or businesses. Like, I don't even know who I would consider to be a pure play model company at this point. You know, Anthropic, if you look at their- their software revenue-

   17. JFJared Friedman6:43

       Hm.

   18. ALAaron Levie6:43

       ... um, you know, it's- it's effectively a- an API business for- for enterprises. I'm sure they have, you know, some- some large scale consumer kind of cloud business, but you're really paying for the security, the compliance, the governance, the privacy, the uptime, the SLAs, talking to somebody that- that m- you know, manages your account, and the- the model just continues to sort of switch out underneath all that. Uh, if you look at OpenAI's revenue, anything that's been kind of leaked publicly, it's very clearly a software company at this point that has AI models that- that power its underlying software. Google obviously, you know, is just GCP, um, and then- and then Meta doesn't need to monetize it because they can just open source it. So, maybe xAI is almost the closest thing to now a model company, but that will show up in Grok, et- et cetera. So, you know, what you probably wouldn't want to do right now, uh, is- is start a pure play model company expecting you're gonna have, like, licensing, you know, revenue by just selling your model to- to people to go and- and use if you don't have e- enough other kind of surrounding value proposition that- that, you know, again, lets you get incorporated into- into enterprises or has a large-scale consumer application where you have some degree of kind of, uh, traffic that- that, you know, keeps people within your ecosystem. Um, I think it would be very bad to be just a pure play model company, uh, in, uh, in- at- at this moment just because you have- you have enough different business models that have emerged now in AI where- where it's gonna be pretty hard if- if your pure play business model is just pure AI tokens because, you know, you always have Meta that will- will always create a counterbalance by just open sourcing, you know, a frontier model that- that kind of wipes you out. So...

   19. JFJared Friedman8:17

       And now DeepSeek.

   20. ALAaron Levie8:17

       Yeah. And now DeepSeek.

   21. SPSpeaker8:18

       (laughs)

   22. ALAaron Levie8:18

       And so... And- and- and that- that's what's amazing is like now that n- like, you can basically guarantee Meta has to do anything DeepSeek does because it obviously has to stay in the game on the open source front. And so we always have... Uh, w- there's enough dynamism in this industry th- that basically ensures, to- to Gary's opening point, that, like, the cost of intelligence is gonna go to zero. Like, it's just, like, absolutely guaranteed.

   23. SPSpeaker8:37

       So extrapolating,

4. 8:38 – 12:47

   What does it mean for startups as intelligence becomes a commodity?

   1. SPSpeaker8:38

      what does it mean for startups as intelligence becomes commodity, basically?

   2. ALAaron Levie8:44

      Well, uh, the good news is we kind of know, uh, we- we kind of know the playbook on this, uh, with one X factor, which is like AGI and- and, like, what is the ultimate-

   3. SPSpeaker8:52

      Yeah. (laughs)

   4. ALAaron Levie8:52

      ... you know, kind of out- out-

   5. SPSpeaker8:53

      X Factor. (laughs)

   6. ALAaron Levie8:53

      Yeah, exactly. A little bit of an X Factor of, like, wiping out all of- all of relevant business models if we don't have money-

   7. JFJared Friedman8:58

      (laughs)

   8. ALAaron Levie8:58

      ... in the future. But, like, if you- if you put that to the side, um, I think it- these- these companies need to look like software companies. And- and it is, it- it's sort of back to basics, which is, you know, we used to have an API into a database. We had an API into storage. We had an API into compute. Now we have an API into intelligence. That intelligence is- i- it should be, you know, basically the cost of that intelligence will go down to the cost of the bare metal. So whatever the underlying c- cost of the GPU is, that's what you're gonna pay. Um, you know, with a little bit of margin from a- a hyperscaler, but the- the cost of the actual tokens, you know, will converge at zero. And so then it's all about, you know, do you build software that takes, you know, this complicated technology and delivers it to customers to solve real world problems? And so, you know, you guys have talked a lot about vertical, you know, uh, AI. I think that's a massive play. I think there's gonna be, uh, certainly a whole layer of AI software that- that kind of stitches together different AI systems so you have horizontal plays, you have vertical plays. I- I think the idea of every single industry and every job function probably will have some degree of new startups and agents that- that get built out for those, uh, th- those- those, like, slots. Um, I don't know if you guys have, like, a whiteboard of every industry and every job-

   9. SPSpeaker10:08

      (laughs)

   10. ALAaron Levie10:08

       ... but, like- but, like, you can just basically play, you know, bingo on that and then- and then until it's fully filled in, like, there's probably still opportunity left within AI. We figured out the first wave of SaaS o- of- of how to do this. Um, uh, you know, YC was obviously a- a big part of- of a n- a number of major kind of category killers in SaaS, and I think we'll see the same playbook happen in AI.

   11. JFJared Friedman10:27

       One of the many interesting things about DeepSeek was specifically it's the first open source reasoning model. Like, in the short term, do you think there's, like, new ideas in the enterprise in particular that are gonna come out because now we have open source reasoning models?

   12. ALAaron Levie10:40

       So what we've seen is, um, uh... So we- we- we do a- a number of benchmarking kind of exercise internally for- for the reasoning models versus- versus, uh, y- you know, kind of, let's say, non-reasoning models and some things they're- they're actually better at it, some things they're weirdly worse at. Um, uh, and, uh, I- I don't think we've even discovered why they're- they're worse at these things. Um, maybe they overthink a problem. In general, I would- I would kind of just...... argue that anything that directionally improves intelligence, you will see B2B use cases- You'll see the value of those use cases go up.

   13. JFJared Friedman11:13

       Yeah.

   14. ALAaron Levie11:13

       Because, um, you know, w- you, you can begin to reasonably chain together more, more agents working together. You can get more agentic workflows happening. Like, anytime we can get the intelligence factor to go up, um, I can now reliably introduce this for a more important business process. And so, you know, in the enterprise, you can almost think about it as there's some probably, you know, either two-by-two or chart. I don't know if anybody's made it, but like, kind of like mission criticality of the workflow, AI level of intelligence, and, and, and, you know, kind of there's an element of, like, you can't introduce it to, you know, closed banking, you know, the, the, like, uh, you know, a banking system's, you know, sort of, you know, end-of-day data, you know, yet because it's, it's not particularly deterministic. It's, it's, you know, we, we don't sort of, you know, know all the, the answers that it's gonna give. But it could write a summary for, for, you know, a new product launch at a bank, um, uh, or it could help answer, you know, banking, uh, you know, uh, product questions if you're a consumer. So, there's a continuum there and as, as we get every degree of intelligence going up, we get more use cases that we can now implement this for. Um, and then there's another axis which is like, how many of those use cases, you know, can you string together to, to complete like the full, the full workflow of that business process and that's yet another, I think, axis that, that, that, uh, we're early in. But like, uh, you know, I was, um, I was in New York a couple weeks ago meeting lots of banks and, and, you know, just generally what you think of, like, the New York industries and, and enterprise, and I would say we're like 10% of the way into the adoption of, let's say, just like general chat, so like assistants, and like 1% of the way into adoption of anything we would all call agents. Like, and that's maybe even inflated

5. 12:47 – 20:19

   Do Fortune 500’s have any interest in underlying models?

   1. ALAaron Levie12:47

      numbers. So, yeah.

   2. JFJared Friedman12:47

      When you're in the room with, like, the banks, the Fortune 500s, all the people making their decisions in enterprise, um, do they really have zero interest in the underlying model? So, like is DeepSeat coming out just like a total nothingburger for them and they just care about what you're pitching them and offering them, or do they, do they have interest in the actual underlying tech?

   3. ALAaron Levie13:05

      There are people like us and the people listening, uh, at every company on the planet.

   4. JFJared Friedman13:09

      Yeah.

   5. ALAaron Levie13:09

      And so those people care. Uh, by the time you get to the, let's say line of business, so I'm the head of wealth management at a bank, they don't care.

   6. JFJared Friedman13:16

      Okay.

   7. ALAaron Levie13:17

      So, but the CTO cares, and the head of AI cares, and the, and the, you know, the, the, the IT folks that, that dabble and they're hanging out on Hacker News, like those people-

   8. JFJared Friedman13:25

      (laughs)

   9. ALAaron Levie13:25

      ... th- they, they care.

   10. JFJared Friedman13:26

       Yeah.

   11. ALAaron Levie13:26

       'Cause they're, they're using Cursor and they're using, they're seeing the differences between, between Anthropic and, and OpenAI, you know, tokens within, within that. But when it goes to, to talking to, you know, an, an executive in the business or the daily end user, they, they have no interest. Uh, it's all, it, it, it's all, you know, a foreign language to them. And I think that will remain the, the, the, the, you know, the way forever. I, I think more of the expectation is, is that, again, these things will converge. And what's amazing about AI is because of the, I don't want to call these, the models fungible, but, but directionally fungible. Because they're somewhat fungible, you will see characteristics that we've seen in, in other areas of compute, which is any, uh, best-in-class model eventually has to be, you know, match the price of anybody that beats them in pricing because, because you can just switch to a slightly inferior model, even if it's like inferior by 1%. The risk is that you could switch to that and be, and find it acceptable for, for, you know, 80% of your use cases, which then by definition means whoever is at the frontier actually still has to match the pricing of somebody just slightly worse than them, uh, because they could just, you know-

   12. JFJared Friedman14:29

       The end user actually doesn't care. It's like, as long as it's good enough.

   13. ALAaron Levie14:31

       The end user doesn't care, and their business could evaporate if, if they don't do that. Which means, ironically, you could actually stay on one of the providers, you could just pick a provider, and you kind of know that the, your, your tokens will become as cheap as the second or third, you know, cheapest option because that, that first, you know, whatever that first provider is, the marginal, the next marginal customer doesn't have to choose them. They could choose the second or third player. So, you, which eventually then, you know, you, you, you run that experiment out, you know, 10 years, you converge on the basically the same pricing, which is what we've seen, which is like the d- difference in pricing of, of, you know, uh, storage buckets, you know, between the top three or four hyperscalers are, are not so different to drive business model, you know, fundamentally different business models, um, in the software stack, similar to compute, et cetera. So really you're making a choice based on, you know, some other s- set of reasons, like how much data do I have in the system? What are my workflows that I've built in the system? Um, and, uh, and then I think the, again, the price of the AI eventually becomes largely the same.

   14. SPSpeaker15:27

       Actually, I think what you're saying applies to what we're seeing for startups. I've done a number of Office Hours with AI startups that are selling to enterprises, and a particular story is this company that scaled to 12 million revenue within a year.

   15. ALAaron Levie15:40

       Yeah.

   16. SPSpeaker15:41

       They actually switched models underneath-

   17. ALAaron Levie15:44

       Yeah.

   18. SPSpeaker15:44

       ... a number of times and the end customers, which were these big enterprises, didn't care.

   19. ALAaron Levie15:49

       Yeah.

   20. SPSpeaker15:49

       What they care was that ultimately the contract and the expectation was like, just get the workflow done-

   21. ALAaron Levie15:54

       Yeah.

   22. SPSpeaker15:54

       ... with this level of accuracy, done.

   23. ALAaron Levie15:56

       Yeah.

   24. SPSpeaker15:57

       And as the cost per token has been cheaper, they actually have been increasing their margins. I think when they started launching, I think their margins were around like 30%.

   25. ALAaron Levie16:06

       Yeah.

   26. SPSpeaker16:06

       The next cycle of iterations middle of last year with all the model releases got to 60%, and I think now they're at like 80%.

   27. JFJared Friedman16:13

       That sounds like file storage.

   28. ALAaron Levie16:14

       Yeah, exactly. (laughs) We, we love that. (laughs)

   29. JFJared Friedman16:15

       (laughs)

   30. SPSpeaker16:15

       So that's, that's kind of an-
6. 20:19 – 27:04

   What are enterprise execs thinking about AI right now?

   1. DHDiana Hu20:19

      we go back to your trip to New York City?

   2. ALAaron Levie20:21

      Yes.

   3. JFJared Friedman20:22

      (laughs)

   4. DHDiana Hu20:23

      Can you-

   5. ALAaron Levie20:23

      Which, which was just any of my- any of my trips? (laughs)

   6. JFJared Friedman20:24

      Yeah. Well, no. (laughs)

   7. ALAaron Levie20:25

      (laughs)

   8. DHDiana Hu20:26

      Like, you spend a lot of time talking to senior execs at Fortune 500 companies about their technology and AI strategy, probably more than almost anyone in the world. And I'm, I was really curious what those people are thinking about AI.

   9. ALAaron Levie20:39

      Yeah.

   10. DHDiana Hu20:39

       Are they focused on it? W- what do they think it's gonna mean for their business? Are they building AI initiatives internally or are they trying to buy products from other people? What's happening?

   11. ALAaron Levie20:49

       Yeah, I mean, definitely all the above. Did you see this, this thing that went viral, like, two weeks ago? David Solomon, uh, CEO of Goldman Sachs-

   12. JFJared Friedman20:55

       The S-1, correct? Yeah.

   13. ALAaron Levie20:56

       Yeah, the S-1 prep. He, he basically had this quote at, at a, at an AI event at Cisco. They're doing projects internally where, where AI is writing an S-1 in, like, 10 minutes or something. And it used to be a team of, of six people that work on that, et cetera. The exact same quote, uh, just parallel universe quote 15 years ago, let's say in the early days of cloud, just as a, as a useful kind of comparison, I'll, I'll probably keep coming back to the cloud thing. Th- it probably would have, uh, been a banking CEO saying, "We'll never go to the cloud." Like, "We don't trust the cloud. We, we don't..." And, and now the exact opposite.

   14. JFJared Friedman21:26

       Which happened, right? Didn't Jamie Dimon do that?

   15. ALAaron Levie21:27

       Jamie Dimon did. I think he's, he's sort of evolved his thinking. But you had that ac- uh, kind of across the board. These were these famous moments is like, "We'll never be a cloud company." You know, "We don't trust it. We don't wanna move..." I mean, Amazon, it's a bookstore, like, like that was the-

   16. JFJared Friedman21:38

       (laughs)

   17. ALAaron Levie21:38

       ... that was the refrain. And it made sense. I mean, I, I even said that when I saw S3, like the bookstore is gonna power... Uh, like what? So think about how different of a world it is that the CEO of Goldman Sachs is basically saying like, "This is now what's possible." He wasn't saying that in like a, "We shouldn't do it" way. He was saying that in like a, "We need to open our, our eyes, uh, you know, up to all of the, the potential use cases that AI is gonna have in the business." And he was sa- saying it as a way that they're leaning in and starting to try out all these use cases. So for that to happen for a top five bank in the world at this early in the cycle, you know, the... it only goes kind of... it only goes more aggressive from there, because he's in the most regulated of all, all the businesses in the, in the most important financial market in the world, and he's already leaning in.

   18. JFJared Friedman22:20

       I- is that because he's like a, a particular early adopter or are you just seeing this across the board where-

   19. ALAaron Levie22:25

       Yeah, I mean, uh, also he's like a DJ. So like-

   20. JFJared Friedman22:27

       (laughs) So, so...

   21. ALAaron Levie22:27

       (laughs)

   22. JFJared Friedman22:28

       It's an interesting character. (laughs)

   23. ALAaron Levie22:30

       So maybe... (laughs) Yeah, I... uh, maybe he's hanging out with, like, EDM people that are just, like, really into AI music. Okay. So, you know, 10 years ago, we'd host th- these dinners, 15, you know, CIOs from different industries, heavily financial services, let's say, if it's New York. And, and it's like the, the... you know, we're gonna try cloud for this one tiny part of our business. We don't really think we can scale. The idea of being cloud first is like... it would be, like, totally an, you know, an anomaly. Like, you would never... like, a bank would never say they're cloud first 10 or 15 years ago. Uh, today, it is, it is sort of like we're trying this in as many areas as possible. Um-Everybody's still insanely early because you've got privacy counsels, compliance counsels, regulatory bodies that have to look at everything. But everybody understands how, how big of a tidal wave this is going to be in their business, um, on a few dimensions. One, they know that the workforce is going to completely change. They, they, I think there's a recognition, this kind of hit me maybe a year ago in some of these conversations, there's a recognition that, that, that basically if you're entering the workforce today, you've had now a couple years of ChatGPT, of, of like college. Like you don't, like-

   24. HTHarj Taggar23:34

       They're native. That generation is native, yeah.

   25. ALAaron Levie23:35

       They're native. It's an AI native, you know, era of, of the workforce. And you know, we could make some jokes about it maybe two years ago, like, "Oh my God, the writing essays, I can't believe it." But like I, I basically almost don't search the internet anymore. Like I only know how to use AI to find information. And guess what, like I'm, I, I find 10 times more, more information as a result of that. So, so actually many respects, the AI native, you know, people will be smarter on the topics that they decide to go in on, um, than, than the, the prior generation of, of whatever that is.

   26. JFJared Friedman24:03

       And so what does that mean, like these Fortune 500 companies are changing how they hire or?

   27. ALAaron Levie24:07

       So I, I don't necessarily know how they hire but, but it'll become clear that if you don't have AI, uh, if you're not an AI first bank or media company-

   28. HTHarj Taggar24:14

       So what's your AI strategy? (laughs)

   29. ALAaron Levie24:14

       Yeah, literally, what is your AI strategy? Because, because why would it-

   30. JFJared Friedman24:17

       You mean to the customers or internally?
7. 27:04 – 28:17

   Is Box investing in internal AI tools?

   1. ALAaron Levie27:04

      yeah.

   2. JFJared Friedman27:04

      And how... Has Box invested in any internal AI tools to speed up how you run the company?

   3. ALAaron Levie27:08

      Yeah, so, uh, a, a few categories. So one, uh, we're, we're, we've been rolling out AI for, for, on the coding side. Um, and we, we're, we're trying everything, uh-

   4. JFJared Friedman27:18

      Like internal tools for your engineers to-

   5. ALAaron Levie27:20

      Yeah, yeah, so basically just, just how do we make the engineering kind of more productive and that, that's sort of obviously what one of the biggest X factors of our business is, can we output more code that is obviously useful and, and, and aligned to our product roadmap? So, um, we will be, you know, fully AI first in terms of how we develop, um, you know, uh, uh, this year it's, uh, it, it's the, it's sort of the big year for all the change management on that. We are, you know, incrementally rolling out AI for different customer facing things, just again, can we solve the customer ticket problem? Can we, can we improve the, the rate of response? And then as an AI provider, a lot of the knowledge management use cases, we, we sort of now do ourselves with AI. So you know, if, if an employee has a, has an HR question or a benefits question, we have a, a feature that lets you talk to all your HR data-

   6. JFJared Friedman28:00

      Cool.

   7. ALAaron Levie28:00

      ... um, and all of the internal knowledge management. And so what, what became this breakthrough for us was all of a sudden all the things that were inside your documents before become useful for now just in- interrogating with questions as opposed to reading documents. And so there's a lot of just embedded productivity that, that we focus on from that, that standpoint.

8. 28:17 – 34:50

   What will enterprises build internally and what will they buy solutions for?

   1. HTHarj Taggar28:17

      What things do you think they're gonna do internally and what things do you think they're gonna buy solutions for?

   2. ALAaron Levie28:22

      I, I would basically guess that, that again it kind of looks pretty similar to, to kind of historical ways of, of thinking about this. Um, I think Geoff Moore created this but, uh, and if I'm getting it wrong again, please Geoff Moore like, like pop into the-

   3. HTHarj Taggar28:36

      (laughs)

   4. ALAaron Levie28:36

      ... pop into the comments.

   5. JFJared Friedman28:37

      We'll just copy in the right name for you. (laughs)

   6. ALAaron Levie28:38

      Yeah, yeah. Um, and the idea was, was sort of context is all this stuff that is sort of like you have to do, it's necessary, but it's not gonna like, it's not gonna make your business, uh, better than your competitor. Uh, and so that's your HR system, that's your ERP system. Like you, you have to have it, it's important, it has to be done extremely well, but like your version of the HR system, uh-... you know, is not going to be radically different than, than, you know-

   7. GTGarry Tan29:01

      Box checking.

   8. ALAaron Levie29:01

      Yeah, exactly.

   9. GTGarry Tan29:02

      Any of the box checking things.

   10. ALAaron Levie29:02

       Yeah, you have to check the box, but it's got to be a good box and, and, and whatnot. And then there's core, which is this is, like, literally your value proposition. Like, like, you sell, you know, wealth management to, you, you sell wealth management services to, to people and, um, and that, that is something you, you own. If your thing looks exactly like your competitors', then you have no, there's no sort of, you know, profit, you know, margin that, that you'd be able to... You know, you wouldn't be able to get reasonable profits. You have to have something that's unique. And I, I think companies really need to understand which category is which, partly because if you get the, if you put the core in the context, then you'll probably be at a long-term disadvantage. And if you put the context in the core, then you're wasting a tremendous amount of time and energy. And, and this is why I, I, I, I really enjoy the, the, the Klarna announcements or whatever. It's, like, fun to read, but I also think it's, it's sort of misunderstanding the context versus core, core thing. Like, you don't need to build your own HR system. I'm glad they're doing it. I think we- it's provocative. Like, it's fun to see different approaches.

   11. GTGarry Tan29:58

       Gets the people going.

   12. SPSpeaker29:59

       (laughs)

   13. ALAaron Levie29:59

       It gets people going. You know, it gets our, like, we're like, our juices are flowing, but, like, the average bank is just, like, not, it's, like, not a priority for them to, like, reinvent their HR system now. So, so, so I think, I think whatever, whatever is that for every industry. You know, in life sciences, you probably really want to understand, like, how are you doing drug development? And you should probably have a, a, a very strong AI team working on that problem, 'cause that's something that sounds like it's going to be IP for you. But the automation of the clinical trial process, that's probably context for you, because everybody's going to want to be doing that as quickly as possible. It probably doesn't involve a tremendous amount of proprietary data, uh, and then, and then obviously, your CRM system, your HR system, and so on.

   14. DHDiana Hu30:38

       So, it actually sounds like they're gonna buy a lot of things externally.

   15. ALAaron Levie30:41

       Yes.

   16. DHDiana Hu30:41

       Because most functions in a business are actually context.

   17. ALAaron Levie30:44

       Yes. Yeah, I think most, most of the way AI will show up to a knowledge worker in 2030 will be from a, what we would have thought of as an ISV ten years prior. Um-

   18. GTGarry Tan30:54

       What's an ISV, for the people watching?

   19. ALAaron Levie30:56

       Yeah, it's just basically a software s- you know, provider. The CRM system will still come from Salesforce or, or, or X competitor. Uh, it won't be that they built a homegrown, AI-generated, you know, uh, you know, CRM system. They might talk to that CRM system through also a new vendor or something that they build internally at Centrium. There's actually a lot of chatbots being, uh, built internally by companies right now. I'm sensitive that I think it might be a temporary phenomenon, um, and-

   20. GTGarry Tan31:19

       I like UX better than chat, but-

   21. ALAaron Levie31:20

       Yeah, yeah. Okay, yeah. I think we're gonna, we're gonna see a hybrid of, of these two things. Um, I, I think the GUI is not as dead as, as, as people think. You do see a lot of kind of chat interfaces where you're like, you're like, "I think you just did way more work than just going to the dashboard."

   22. DHDiana Hu31:34

       (laughs)

   23. GTGarry Tan31:34

       (laughs)

   24. ALAaron Levie31:34

       Like, (laughs) like, like, "I'm, I'm 90% sure you just probably took all the savings from AI gains-"

   25. DHDiana Hu31:41

       (laughs)

   26. ALAaron Levie31:41

       "... and efficiency and, and, and then just spent it on figuring out a prompt, um, that, like, a dashboard would have solved." So, but I think we'll have a universe of both those things, but I do think ultimately, most will come from software as, like, a, you know, on a percentage of, like, like, time that a, a, a knowledge worker spends inside of technology. But some of the most valuable things absolutely will be, will be homegrown. The algorithm you use for discovering the thing or personalizing the medicine or personalizing the wealth, you know, data or, um, you know, Netflix's recommendation engine, like, those will be homegrown things. Maybe still using a, a model from, from a proprietary player, but, like, the scaffolding there will be, I think, a lar- largely built by, built, built internally.

   27. GTGarry Tan32:19

       Interesting. So, I guess the mental model for people watching might be that there's inside the house and outside the house. Outside is context, which are just check boxes you have to check, and those might be, you know, end-to-end things that just do the thing like Salesforce.

   28. ALAaron Levie32:34

       Yeah. Yeah, I think, I think that's right, and I think the, maybe another way to reverse engineer it is if you were the customer of this company, would you care? Would you care what technology they used for, for that category of thing? Like-

   29. GTGarry Tan32:44

       Or is it, like, just get, get the box checked, and I'm good?

   30. ALAaron Levie32:46

       Yeah. As a customer of Netflix, I literally don't care what their ERP system is. Like, doesn't matter.
9. 34:50 – 36:16

   Is enterprise concerned with third-parties and security?

   1. DHDiana Hu34:50

      of this cycle, there was a lot of concerns from enterprise about the security implications of using hosted models, like, a whole bunch of companies like ban ChatGPT, use internally. What- I- w- what's happening now? Have they gotten comfortable with the idea of having all of...... their data go to OpenAI and Anthropic, or do they still... Are they sort of w- really worried about that and trying to host open source models and things like that on prem?

   2. ALAaron Levie35:11

      Yeah. I, I, I think you'll see a different, different categories, uh, by industry. So, a lot of times, they'll go into, you know, a bank, and they'll say, um, and they'll be proud of this, and, and as they maybe should be, they'll, they'll say, "You know, we have our own kind of enclave version of, of X model, and then we built out a wrapper on top of that, you know, for, for, uh, deploying it to employees." Um, and I think that, that you'll always have some percentage of the market do that, you know, 10% of, of the market, 'cause you, you still have a lot of people that have on-prem systems for a similar reason, that will always be there. Um, but the comfort level is absolutely increasing. That tends to always happen in, i- if, if the industry or the, the company takes it very seriously. So, OpenAI has taken security, privacy, compliance, you know, regulatory controls very seriously. And so then that builds trust over time, so then more people can, um, can, can go and, uh, can, can experience those, um, you know, those use cases. So, anytime a, a software category matures and, and becomes more enterprise grade or m- you know, military grade, you know, you'll see a corre- you know, correlation of the, the amount of trust in, in putting data into those systems. And I think AI is no different than, than, again, software and, and kind of cloud in that respect.

10. 36:16 – 39:46

    Shout-outs to Aaron!

    1. SPSpeaker36:16

       Maybe part of it is just a lot of the behavior is from all these enterprises, they've gotten a lot more comfortable because of what you pay with cloud, right?

    2. ALAaron Levie36:26

       Yes.

    3. SPSpeaker36:26

       They kind of already have preferences on how to do this-

    4. ALAaron Levie36:28

       Yes.

    5. SPSpeaker36:28

       ... uh, hosted and, "Okay, we know how this looks," so it's a lot harder for you to sell on that first cycle.

    6. ALAaron Levie36:33

       Yes.

    7. SPSpeaker36:33

       This cycle, you kind of compounded it for this next generation of founders, so thank you.

    8. ALAaron Levie36:36

       People need to be really thanking me.

    9. DHDiana Hu36:37

       (laughs)

    10. ALAaron Levie36:38

        Um, I, I am not... Actually, you know, it's funny that you put it that way. I don't think I've ever been thanked for-

    11. DHDiana Hu36:43

        (laughs)

    12. ALAaron Levie36:43

        ... for the hard work.

    13. GTGarry Tan36:44

        Consider yourself thanked.

    14. ALAaron Levie36:45

        The years of, of dinners-

    15. DHDiana Hu36:46

        (laughs)

    16. ALAaron Levie36:47

        ... th- to all the CIOs that made it possible for your AI startup to get sold into the enterprise, I'd like a big thank you.

    17. GTGarry Tan36:52

        (laughs)

    18. ALAaron Levie36:52

        Uh, I-

    19. SPSpeaker36:53

        To cloud, yes.

    20. ALAaron Levie36:54

        ... I have not been respected.

    21. SPSpeaker36:55

        Hat tip.

    22. ALAaron Levie36:56

        Uh, that has not, that has not shown up in, in a lot of ways, but I appreciate it. Is there-

    23. GTGarry Tan36:59

        Every single white hair-

    24. ALAaron Levie37:00

        Yes.

    25. GTGarry Tan37:00

        ... is a dinner-

    26. ALAaron Levie37:01

        Y- actually.

    27. GTGarry Tan37:01

        ... that you had to do to teach people-

    28. ALAaron Levie37:02

        Literally. Uh-

    29. GTGarry Tan37:03

        ... that you should buy this.

    30. ALAaron Levie37:04

        It's, uh, uh, th- this is the only way to get that. Um, I would like a plaque at the office-
11. 39:46 – 48:38

    The transition from cloud to AI

    1. DHDiana Hu39:46

       had this front row seat to the transition from on prem to cloud, and now we're at the dawn-

    2. ALAaron Levie39:51

       I am that old (laughs) .

    3. DHDiana Hu39:53

       ... of the next transition from cloud to AI. W- how do you think that's gonna play out similarly? How do you think it's gonna play out differently?

    4. JFJared Friedman39:58

       And, and how do you think that relates to the TAM for software going forward?

    5. ALAaron Levie40:02

       Yeah, so, uh, if I could, if I could try and merge these, actually, there's, there's a, there's a cool connection point. So, the probably single biggest, bare, uh, reason why people didn't invest in SaaS in the mid-2000s was, was they, they, they thought the market sizes would be basically the same size as the on prem software company. And, and so if it was the same size as the on prem software company but also the in-, the software company that's already there is the incumbent, you know, it's like how do you squeeze out enough, enough money to, to kind of make the business really, really, you know, make, make it really interesting? And what everybody basically got wrong was, it turned out that the TAMs were probably about 10 times larger.

    6. DHDiana Hu40:42

       Okay, and why did the TAM-

    7. GTGarry Tan40:44

       ... grow so much?

    8. ALAaron Levie40:45

       Because, um, just to, like, bore everybody, if you wanted to buy a CRM system in two tho- in 20, uh, sorry, in 1999, you had to be like, "Okay, I'm gonna go to the systems integrator, I'm gonna get a data center, I'm gonna buy a bunch of servers from people, I'm gonna install some software, I have to manage the network of that." And, like, you know, lo and behold, two years later, you might have a CRM system, and you probably spent $5, $10 million on the full project to do that. So, think about who's the market that then can implement a best-in-class CRM system, it's the world's largest enterprises. You know, 5,000 companies, 10,000 companies. Salesforce comes out, and they're like, "For three seats, online, with your credit card, you have a CRM system as good as Siebel." Obviously, there'll be some nuance because it didn't have as much functionality, but, like, for that company, that was as powerful as Siebel, you know, getting started. Now, all of a sudden, your TAM is basically every business on the planet, so you go from a, a market that had maybe 10,000 customers, 20,000 addressable customers to now a market that has five million, 10 million potential customers. It is a totally different scale, like, you know, two or three orders of magnitude more scale that you can now go and serve. We- we had a similar experience, which was, you know, the- the industry we were disrupting was, like, legacy enterprise document management, enterprise content management systems, same exact thing as Siebel in terms of, like, we would read the S1 of- of our biggest incumbent competitor, and they were talking about, like, a thousand customers or a couple thousand customers and- and literally, you know, now we have 115,000. But, like- but, like, but at the time, we had, we had, I don't know, 5,000 or 10,000 when we started thinking about disrupting this, and, like, the- the scale was just completely different. So- so that meant the market sizes were so much larger. Um, you know, ServiceNow, today, uh, I don't know the- the exact latest market cap, maybe it's 150, 175 billion dollars. Their incumbent competitor, when they, when they were first growing and disrupting the market, today is worth maybe about five or 10 billion. So- so if you had looked at this company 20 years ago, you'd be like, "At best, ServiceNow should be a five or $10 billion company if it just was, like, a better version than the current thing." And it turns out it's- it's 15 times larger than- than- th- than what you would have thought. Salesforce did the exact same thing and so on. AI, I think, has a similar dynamic because you're basically increasing the total spend on software. So, it's not so much that a new set of companies will buy software for the first- for the first time, it'll be all companies use software to do things that software didn't do before, and that will take from budget that- that previously was sort of untapped, you know, from software. So, um, the budget will be from a variety of things, um, uh, but, uh, often because now th- software is doing useful work for you, you can now afford to spend even more on that software because the alternative was a much more expensive sort of proposition. (smacks lips) Here's where I think people kind of get it wrong though, they- they think about it as zero sum from, "Well, then- then all you can do is sort of take from the labor side of that spend," but it actually just turns out most companies aren't even spending on the labor side, they're just not doing the thing.

    9. GTGarry Tan43:38

       Hm.

    10. ALAaron Levie43:38

        So- so, you know, most companies are, like, uh, uh, just, like, globally, are not, like, spending time to translate their advertising into a different language. So, it's not that- that, oh, the- the-

    11. GTGarry Tan43:48

        Yeah.

    12. ALAaron Levie43:48

        ... the market for the translation services are this big and we're just gonna digitize it. It's like, no, 100 times more people will do translation. Um, th- you know, in- in our business, like, we have, you know, software now that reads your contract and pulls out the critical data from that so you can automate a contract workflow, and, like, the number of people globally that are reviewing contracts and pulling out that data, maybe it's, you know, 10,000, 50,000 people, I don't know the exact number, but a very small percentage of companies are doing that with their contracts. So, they will now decide to prioritize automating a thing that they didn't automate before. You know, Cursor, the- you know, the- the- as just a, you know, back to that example, a Replit or Devin or whatever, there's probably not a single dollar that's being spent on that technology that comes from- from- takes away from what people are currently doing. It's purely additive because now it's expanding the use cases that- that software can- can- can kind of tie into. So, I think we're in for a potential scenario where the- the size of software, now you have to include AI in that, could be five times larger, uh, in- in the next decade because it just- it- it supplies the actual underlying work that you actually bought the software for in the first place. Um, and that just changes everything because now we're- we're- you're gonna be paying for work as a paying for- for a tool that enables other people to do work.

    13. GTGarry Tan45:01

        I think that's such a powerful AI white pill, actually.

    14. ALAaron Levie45:03

        Yeah. Yeah, it is, actually. (laughs)

    15. SPSpeaker45:04

        Yeah.

    16. GTGarry Tan45:04

        It's not merely, you know, zero sum, we're converting payroll into software revenue-

    17. ALAaron Levie45:09

        Yeah.

    18. GTGarry Tan45:09

        ... and ha ha, that's it.

    19. ALAaron Levie45:10

        Right.

    20. GTGarry Tan45:10

        It's actually, we're gonna do things that enterprises should be doing, would have been doing-

    21. ALAaron Levie45:15

        They never got around.

    22. GTGarry Tan45:16

        ... and then, you know, actually, the people who are the consumers on the other end, they're gonna have better products, better services.

    23. ALAaron Levie45:22

        Yes.

    24. GTGarry Tan45:22

        Like, the thing will actually just be better for-

    25. ALAaron Levie45:24

        I haven't yet read, like, the full, like, economic study on this, but where economists always get this stuff wrong is, you know, they- they do, probably by default, tend to have a kind of a zero- I mean, you wouldn't have, like, you know, Jevons paradox if- if economists always knew, you know, how to anticipate these things. But, like, what I think they often get wrong is they look at the total amount of market labor in a category and they're like, "Well, shoot, if AI automates that, that's now gone, look how many jobs that is." You know, I- I think we should debate it and we should talk about it 'cause it's a very, you know, serious thing, but what they don't actually ever think about is the micro- the- the more microeconomic impact of this. So, if I'm a company and, uh, bo- it could be Box or- or it could be a 20-person company, and I use AI to, let's say, code faster. Okay, well, why am I coding faster? Because I'm gonna- I want to build a better product for my customers. If I'm building a better product for my customer, my revenue should be growing faster. If my revenue's growing faster, I probably then am hiring people to go and do things to drive that revenue growth. Maybe it's people selling the software, maybe it's customer support, maybe it's HR people to help scale the operation. And eventually, I'll get to a point where I say, "Should I reuse those dollars that helped me grow faster to hire more engineers to grow even faster and to build more of that roadmap?" And- and if- if we were in a- a market that wasn't competitive, maybe you could say, "You know what? I actually just want to take the profit and be happy." But we're in a competitive market, so if you're the one company that decides to sit on the profits that AI generated and just- and just live off of higher profit margin-... 20%, 30%, 40% profit margin as a company, you'll just have somebody come into the space and say, "No, actually, we'll just... I'm fine to have 20% profits and do that same thing," and then that company will, will eat into your lunch. So you actually then reinvest those dollars-

    26. GTGarry Tan47:03

        Yeah.

    27. ALAaron Levie47:03

        ... back into the things that are helping you grow faster, and that's actually like the microeconomic outcome of, of automation is, is you decide that you take that efficiency gain and you redeploy it into the business in something that will make you more competitive or grow faster or better serve your customers-

    28. GTGarry Tan47:18

        Mm-hmm.

    29. ALAaron Levie47:18

        ... because you're in a competitive ecosystem. And that's why I think as, uh, you know, over time, you'll... yes, you'll have some displacement in different categories, but over time, this is why it generally just looks like an upgrade to just how we tend to work, you know, in the world. Like, we just tend to use tools to, to work faster, to make better decisions, to build better customer products, the customer gets a better result out of that, but we reinvest those dollars back into the businesses because we're in competitive ecosystems.

    30. GTGarry Tan47:44

        And then the ultimate winner is the consumer and society.
12. 48:38 – 49:28

    Outro

    1. ALAaron Levie48:38

       state is we use this automation to, to actually deliver better outcomes for, for the world, um, and that will require tons of jobs, uh, as a result of that.

    2. GTGarry Tan48:46

       Yup, we can be a society again-

    3. ALAaron Levie48:48

       Yeah.

    4. GTGarry Tan48:48

       ... because of AI.

    5. ALAaron Levie48:49

       There you go.

    6. GTGarry Tan48:50

       Aaron, (laughs) thank you so much.

    7. ALAaron Levie48:52

       Thank you.

    8. GTGarry Tan48:52

       Thank you so much for being with us.

    9. ALAaron Levie48:54

       Great to be here.

    10. GTGarry Tan48:54

        Uh, I think that that's a great place to end just because, you know, to be continued. Like, you know, I, I think that we're in the middle of the revolution, and the revolution does not have to be, uh, Black Mirror. It could be something that is driven by Jevons paradox-

    11. ALAaron Levie49:08

        Yeah.

    12. GTGarry Tan49:08

        ... driven by abundance for everyone, and that's certainly the, uh, the timeline we want to be on, so let's do it.

    13. ALAaron Levie49:14

        Uh, that, that's the future I'm betting on. (music)

Episode duration: 49:28