Box CEO on the AI Adoption Gap | The a16z Show

1. 0:00 – 0:51

   Intro

   1. ALAaron Levie0:00

      The diffusion of AI capability is gonna take longer than people in Silicon Valley realize.

   2. SSSteven Sinofsky0:04

      It's just absurd-

   3. ALAaron Levie0:05

      Yeah

   4. SSSteven Sinofsky0:05

      ... to think you're gonna vibe code your way to-

   5. ALAaron Levie0:07

      Yeah

   6. SSSteven Sinofsky0:07

      ... like SAP. All of that domain knowledge, it's not just represented in some well-orchestrated data layer.

   7. ALAaron Levie0:13

      The engineering compute budget conversation is gonna be the most wild one in the next couple years.

   8. SSSteven Sinofsky0:18

      The biggest problem right now is everybody is trying to figure out the economics of all of this-

   9. ALAaron Levie0:22

      Yeah

   10. SSSteven Sinofsky0:23

       ... when they're off by at least an order of magnitude on how big the opportunity is.

   11. ALAaron Levie0:27

       Uh huh. If you have 100 or 1,000 times more agents than people-

   12. SSSteven Sinofsky0:31

       Yeah

   13. ALAaron Levie0:31

       ... then your software has to be built for agents.

   14. SSSteven Sinofsky0:33

       People in the abstract-

   15. ALAaron Levie0:34

       Yes

   16. SSSteven Sinofsky0:34

       ... say things like, "Now you're marketing to agents. You're like an API. You've got a good IDL."

   17. ALAaron Levie0:39

       Yeah.

   18. SSSteven Sinofsky0:39

       I actually think that's almost exactly wrong, which is-

   19. ALAaron Levie0:41

       Wow.

   20. SSSteven Sinofsky0:41

       That's-

   21. ALAaron Levie0:42

       This is breaking podcast news. If you start to imagine that we all have to build software for agents, I think we're, like, all clear on that, right?

2. 0:51 – 2:10

   Building software for agents vs. humans

   1. ALAaron Levie0:51

      So, like, that trend is happening, which is, like, we spend as much time now thinking about the agent interface to our tool as we do the human interface.

   2. SSSteven Sinofsky0:59

      Okay. Okay, yeah. So sure.

   3. MCMartin Casado1:00

      Sure.

   4. ALAaron Levie1:00

      Okay?

   5. MCMartin Casado1:01

      Yeah.

   6. ALAaron Levie1:01

      And the reason we're doing that is 'cause our hypothesis would be that if you have 100 or 1,000 times more agents than people-

   7. MCMartin Casado1:07

      Yeah

   8. ALAaron Levie1:07

      ... then your software has to be built for agents. And then what, what is the way that those agents are gonna interact with your system? It's gonna be through an API or a CLI or MCP or whatever. And the, the paradigm that appears to be taking off and is quite successful so far in, in terms of efficacy is what if you give a coding agent access to your SaaS tools and, uh, a coding agent access to, you know, your, your knowledge work sort of workflows and context, and that kinda becomes this superpower, which is it's not just like the agent is not only capable of like, you know, reading some data, understanding some information. It, it can actually code its way or use APIs, you know, through whatever task it's trying to achieve. That appears to be like a, like a paradigm that is, is starting to compound and, you know, that was the, that's the claw- Claude Cowork phenomenon. That's the whatever OpenAI is, is kinda cooking up, um, you know, with the super app, Perplexity Computer, et cetera. And, uh, and I actually think it, it, it kinda makes sense as like the ultimate manifestation of this stuff.

   9. SSSteven Sinofsky2:09

      I

3. 2:10 – 14:31

   Can non-technical workers actually use AI agents?

   1. SSSteven Sinofsky2:10

      think... I mean, I think you're right. It, it, it makes sense in a-

   2. ALAaron Levie2:14

      [laughs]

   3. SSSteven Sinofsky2:14

      ... in a theoretical way.

   4. ALAaron Levie2:16

      Yeah.

   5. SSSteven Sinofsky2:16

      But in a, in a practical way we, we have to be really careful i- in that, that the way to say it is algorithmic thinking-

   6. ALAaron Levie2:26

      Yeah

   7. SSSteven Sinofsky2:27

      ... is really, really, really hard for the vast majority of people who have jobs.

   8. ALAaron Levie2:32

      Yeah.

   9. SSSteven Sinofsky2:33

      And, and so the, the easiest way to think about it is if you were to go into any person and ask them to create a flowchart for a particular thing that they have to go do, they would probably fail-

   10. ALAaron Levie2:44

       Yeah

   11. SSSteven Sinofsky2:44

       ... at producing that flowchart.

   12. ALAaron Levie2:45

       Yep.

   13. SSSteven Sinofsky2:46

       There, so within any organization, you know, say doing a marketing plan and there's 50 marketing people working on a giant product line-

   14. ALAaron Levie2:53

       Yep

   15. SSSteven Sinofsky2:53

       ... one person probably understands and could document the flowchart.

   16. ALAaron Levie2:57

       100%.

   17. SSSteven Sinofsky2:58

       So if you put one of these agents or you put this, this, this tool, this coworking tool-

   18. ALAaron Levie3:04

       Yeah

   19. SSSteven Sinofsky3:04

       ... in front of people to create these things-

   20. ALAaron Levie3:06

       Yeah

   21. SSSteven Sinofsky3:07

       ... their ability to explain to it what to do-

   22. ALAaron Levie3:10

       Yeah

   23. SSSteven Sinofsky3:10

       ... is really, really limited.

   24. ALAaron Levie3:11

       100%.

   25. SSSteven Sinofsky3:12

       So then you're-

   26. ALAaron Levie3:13

       But, but what if that becomes the new, this is the new way you have to interface with computers?

   27. SSSteven Sinofsky3:18

       Well-

   28. ALAaron Levie3:18

       And you, and you just have to cycle that through.

   29. SSSteven Sinofsky3:20

       Well, then you're, what you, you're basically, you, you just go back to, you're basically just developing the next abstraction layer-

   30. ALAaron Levie3:29

       Yes. Yeah
4. 14:31 – 18:39

   CFO/CIO pushback: the real fear of agents doing integration

   1. SSSteven Sinofsky14:31

      Uh, let me get off my lawn.

   2. ALAaron Levie14:33

      Okay. [laughs]

   3. SSSteven Sinofsky14:33

      Okay. So-

   4. MCMartin Casado14:34

      Okay. [laughs]

   5. SSSteven Sinofsky14:34

      So the reason I, I just was in a room filled with a bunch of CFOs and CIOs, and this, they all looked at me when I said something along these lines-

   6. ALAaron Levie14:44

      Yeah

   7. SSSteven Sinofsky14:44

      ... although not as optimistic as you can imagine.

   8. ALAaron Levie14:46

      Oh, yeah, yeah.

   9. SSSteven Sinofsky14:46

      But they just, they, they looked-

   10. ALAaron Levie14:47

       More realism was in the-

   11. SSSteven Sinofsky14:47

       No. It-

   12. ALAaron Levie14:48

       Yeah

   13. SSSteven Sinofsky14:48

       ... it caused like six of them to come running up afterwards and say, "You're insane."

   14. ALAaron Levie14:52

       Uh-huh.

   15. SSSteven Sinofsky14:52

       "You've lost all credibility with me," because it's back to-

   16. ALAaron Levie14:55

       Wait, wait, wait. What, what specifically that the, the agents are gonna do integration on the fly?

   17. SSSteven Sinofsky14:58

       That, that this, that the integration-

   18. ALAaron Levie15:00

       That's too obvious

   19. SSSteven Sinofsky15:00

       ... is a problem that will get a lot easier-

   20. ALAaron Levie15:03

       Yes

   21. SSSteven Sinofsky15:03

       ... a- as we-

   22. ALAaron Levie15:03

       They were against that?

   23. SSSteven Sinofsky15:04

       No. They're, no one's against it.

   24. ALAaron Levie15:06

       Yeah, I know. They, they think it was just awful.

   25. SSSteven Sinofsky15:07

       But their, but-

   26. ALAaron Levie15:07

       They think it's practical.

   27. SSSteven Sinofsky15:07

       But their, their fear-

   28. ALAaron Levie15:09

       Yeah

   29. SSSteven Sinofsky15:10

       ... is like unleashing not just the agents themselves, but humans to do integration.

   30. ALAaron Levie15:14

       Yeah.
5. 18:39 – 27:35

   Treating agents like employees and why it breaks down

   1. ALAaron Levie18:39

      Okay, now can I instantly take, do a takedown of, of, of this, uh, element that we're gonna run into?

   2. SSSteven Sinofsky18:44

      Please.

   3. MCMartin Casado18:44

      Yeah.

   4. ALAaron Levie18:44

      Okay. So that is fantastic for personal productivity.

   5. MCMartin Casado18:47

      Yes.

   6. ALAaron Levie18:48

      And the question that we're gonna run into is in an enterprise, let's say I have... Let's just make a simple example. I have a 50-person team of something. Should everybody als- Sh- Basically, will we have a hund- Will we have 100 people now collabor- I mean, basically 50 humans-

   7. SSSteven Sinofsky19:04

      And then 50 credit cards, yeah

   8. ALAaron Levie19:05

      ... and then 50, and then 50 agents in that same shared space? And do I have... I obviously have complete oversight over my agent, but what if my agent collaborates with somebody else, uh, and, and then accidentally gets access to some resource because they were sharing with that other person, and I'm not supposed to have access to that resource, and now this autonomous sort of stateful, you know, a- agent is, is running around-

   9. MCMartin Casado19:29

      Right. But, but-

   10. ALAaron Levie19:29

       ... working on somebody else's information

   11. MCMartin Casado19:30

       ... but the default end to end argument is you treat them like human beings and you-

   12. ALAaron Levie19:33

       It, it doesn't work. It, it... So you can't fully treat them like humans because here's the thing, and with regular humans, you don't get to look at the Slack channel of the person that, that is working with you or working for you. You don't get to log in as them.

   13. MCMartin Casado19:45

       Yeah.

   14. ALAaron Levie19:45

       You don't get to oversee them. You are... They are accountable for their own set of execution in the real world. You don't get penalized for what, how they screw up. The agent, you have all the liability of whatever they're doing. You do have complete oversight, and you're probably gonna need to have that complete oversight. They have no right to privacy. So, so there's gonna be these, some of these breakdowns that aren't as clean as just treat them like a person because I need to be able to kind of... I need to be able to give access to something to them, but I also need to be able to like log in as them at some point and be like, "No, no, you fucked up the whole thing-

   15. SSSteven Sinofsky20:18

       Right. Right

   16. ALAaron Levie20:18

       ... and I need to undo it all." But if I can log in as them, how could they have operated in the real world working with other people a- and keeping anything, you know, confidential or secure or whatever? So it really is still an extension of you. It's like almost impossible to get around them being an extension of you. So now the thing that we're thinking through, that we're not gonna be able to do anytime soon-

   17. MCMartin Casado20:35

       I, I, well, I just-

   18. ALAaron Levie20:36

       Yeah

   19. MCMartin Casado20:36

       ... I saw this, this doesn't logically follow.

   20. ALAaron Levie20:38

       Yeah.

   21. MCMartin Casado20:38

       Maybe. But for example, um, for my employees-

   22. ALAaron Levie20:41

       Yeah

   23. MCMartin Casado20:41

       ... I can log in as them. I can get-

   24. ALAaron Levie20:42

       You don't though. You don't, you don't log in-

   25. MCMartin Casado20:44

       I can get access to their email.

   26. ALAaron Levie20:45

       Yeah. No, in like in a... If you get like sued. You're not logging in, you're not logging in a- as them on a regular basis 'cause they sent one email that was wrong.

   27. MCMartin Casado20:52

       But, but isn't the, isn't the right operating model-

   28. ALAaron Levie20:54

       Yeah

   29. MCMartin Casado20:54

       ... with an agent the same thing? It's like-

   30. ALAaron Levie20:55

       The risk is like 1,000 times greater. Like these peop- Like they will just leak your information whenever they want. Like they will happily just go and send some email to somebody 'cause they got prompt injected.
6. 27:35 – 42:53

   Diffusion gap: startups vs. enterprises

   1. ALAaron Levie27:35

      Because what's happening is, like, we see startups that can start from the ground up without any of the risks that we're talking about 'cause they have nothing to blow up. And, and so we look at that as the trajectory that we're on. And then you go to, like, JP Morgan, and you're like, "How are you gonna set up NanoClaw, uh, to, to be able [laughs] to, to actually, like, you know, automate your business anytime soon?" And it's like, oh, okay, there's gonna be, like, a little bit of a gap there.

   2. SPSpeaker27:59

      Yeah.

   3. SSSteven Sinofsky27:59

      Well, what do you guys think... Here's... I, I think that that opens up a pretty interesting problem, which is the split between big and small-

   4. ALAaron Levie28:07

      Yeah

   5. SSSteven Sinofsky28:07

      ... or startup and enterprise, which is just that, that the, the enterpr- the current SaaS vendors, who are all struggling in this SaaSpocalypse weirdness that I, I don't really agree with, but they are struggling with this problem that they, they don't really sell their line of business data. They actually sell this intelligence-

   6. ALAaron Levie28:26

      Yeah

   7. SSSteven Sinofsky28:26

      ... and domain expertise and this whole system. And the agent side of things wants to only buy the data now.

   8. ALAaron Levie28:32

      Mm.

   9. SSSteven Sinofsky28:32

      And they only wanna license the data, and they wanna have unlimited access to the data. But they've actually never really enabled that. Like, that's never been their business, and it's been a long-standing tension point with the likes of Workday and SAP and stuff, like how much API access to have.

   10. ALAaron Levie28:48

       Yeah.

   11. SPSpeaker28:48

       Yeah.

   12. SSSteven Sinofsky28:48

       I mean, Salesforce went through three different massive platform redesigns. You know, it's... I think that-

   13. ALAaron Levie28:54

       Yeah

   14. SSSteven Sinofsky28:54

       ... that's a particularly interesting problem-

   15. ALAaron Levie28:57

       Super interesting

   16. SSSteven Sinofsky28:57

       ... not, not for the same reason that Wall Street does. Wall Street's all wrong about the economics-

   17. ALAaron Levie29:01

       I agree

   18. SSSteven Sinofsky29:01

       ... and the problem and all that stuff, but from a technology perspective, what does system of record mean i- in the face of people wanting to access the data-

   19. ALAaron Levie29:11

       Yeah

   20. SSSteven Sinofsky29:11

       ... when the data-

   21. ALAaron Levie29:11

       For, for, for, for training or for, for-

   22. SSSteven Sinofsky29:13

       Well, they're, they, they are w- they're-

   23. SPSpeaker29:15

       Or are you talking about from like-

   24. ALAaron Levie29:16

       Executing the work load

   25. SPSpeaker29:16

       ... actual day-to-day operations?

   26. SSSteven Sinofsky29:17

       I, I think of it as executing the day-to-day operations.

   27. SPSpeaker29:19

       Okay.

   28. SSSteven Sinofsky29:19

       Their concern is that somebody-

   29. SPSpeaker29:21

       I get it

   30. SSSteven Sinofsky29:22

       ... that they wanna do the training layer-
7. 42:53 – 58:13

   Wall Street's economics are off by an order of magnitude

   1. SSSteven Sinofsky42:53

      this-

   2. ALAaron Levie42:53

      Yeah

   3. SSSteven Sinofsky42:53

      ... when they're off by at least an order of magnitude on how big the opportunity is.

   4. ALAaron Levie42:58

      Uh-huh. Okay.

   5. SSSteven Sinofsky42:58

      Because the, the new models that people will come up with, that nobody knows what they are right now, but they will absolutely come out with new models because that's what happens with every new technology. And the thing that holds back with sort of the discussion now is you basically have a bunch of finance and Wall Street people trying to justify GPUs and tokens and things, like, as if we're in some old world.

   6. ALAaron Levie43:20

      Right.

   7. SSSteven Sinofsky43:20

      And they're, they're, so they're, they're viewing the world of revenue as sort of this linear step fu- literally linear growth curve-

   8. ALAaron Levie43:27

      Right. And so they're thinking too small

   9. SSSteven Sinofsky43:28

      ... trying to justify-

   10. ALAaron Levie43:29

       Yeah, yeah, yeah

   11. SSSteven Sinofsky43:29

       ... all the, all the expe- and when people are gonna create... Like, this was the problem with PCs. People viewed PCs as a finite market because they just viewed the consumption-

   12. ALAaron Levie43:38

       Right

   13. SSSteven Sinofsky43:38

       ... of MIPS as some finite thing.

   14. ALAaron Levie43:40

       Mm-hmm.

   15. SSSteven Sinofsky43:40

       And they didn't think what would happen if we put all those MIPS on every desktop. And in particular, people thought software just came with the MIPS, and nobody thought, "Oh, well, they'll just sell the software."

   16. ALAaron Levie43:50

       Uh-huh.

   17. SSSteven Sinofsky43:51

       One guy did.

   18. ALAaron Levie43:51

       [laughs]

   19. SSSteven Sinofsky43:52

       And, and it turns out that was, like, a really good idea.

   20. ALAaron Levie43:55

       Was it Bill or somebody?

   21. SSSteven Sinofsky43:55

       And the same...

   22. ALAaron Levie43:56

       Yeah, yeah. [laughs]

   23. SSSteven Sinofsky43:57

       Bill and Paul.

   24. ALAaron Levie43:58

       Okay, got it. [laughs]

   25. SSSteven Sinofsky43:58

       And, and the same thing happened, but the same thing happened-

   26. ALAaron Levie44:00

       Yeah

   27. SSSteven Sinofsky44:00

       ... with the cloud, which was people looked at the cloud and they said, "Oh, we're gonna take all of the, the server business-

   28. ALAaron Levie44:06

       Right

   29. SSSteven Sinofsky44:06

       ... which was like literally, like, 60,000 units a year-

   30. ALAaron Levie44:10

       Right

Episode duration: 58:28