Jack Dorsey: Every Company Can Now Be a Mini-AGI

0:00Since last year, I've just had this existential dread and also hope and optimism in the same hour, in the same thought process of like, what is even a company going forward? Like, what are these structures going forward? And this is the only durable one structure that I could actually imagine lasting for quite some time. And it was coming from a place of like, wow, is our company just going to be completely irrelevant in the next coming years?

0:46Brian Halligan:all right everybody we had jack dorsey on the pod today founder of twitter and founder of block He's the only founder to have two of his companies end up on the S &P 500. Just learned that today. I thought it was pretty cool. Also, Rulaf Botha, who's a Skoy partner and is on Jack's board at Block. The conversation was super interesting to me. He recently put an article out that you may have seen called From Hierarchy to Intelligence. and it's a manifesto on how to rethink. Basically, we think from the first principles how an organization works. How do you completely eliminate hierarchy and how do you put AI right in the center and transform how your company works?

1:32And I let him go on that

1:34Brian Halligan:and he was very thoughtful and he spoke a lot about it. Block has really changed the way it's organized. Before we started, he said, I'd love to get some feedback from you and from other people on this. So if you get comments on it, let her rip. He wants some feedback on it. He's in the early innings of a big transformation over there. That was about half of it. The second half of it, I just asked him for like CEO advice. I work with lots of CEOs, as you folks know. Like, how do you build an amazing board? He's certainly got some scars around board building. How do you build your second act? Like the cash app versus the original business of Square.

2:12Brian Halligan:is it a good idea to be a CEO of two different companies at the same time being brave and not giving a flock versus going with the flow and when do you really dig in and when do you kind of go with the flow and he had some really interesting thoughts on the job of a modern CEO so tune in I'll be back at the end and give you some further thoughts but I I thought Jack did a really nice job and in Reloft did too let's get into it Jack I read your piece I think is fantastic from hierarchy to intelligence before we get into the meat of what it is can you describe what you think is wrong with the way normal companies hierarchies work companies like hubspot companies like block like what kind of led you to this i don't know if um there's one thing that's ultimately wrong um it's just it's it's recognizing that like what what do we what do we see in the in the pattern what what is the function of the hierarchy and um what we wanted to explore is just like where does it actually come from and why did why does it exist in the first place and if you look at it from first principles it's all about information flow to a broad base of people um so being able to communicate over a breadth of of people and um have that be manageable at a human scale.

3:34So we've gotten into structures that, you know, we've borrowed and iterated on a little bit over 2000 years. Yeah. And now we're, we're facing this, I think, a completely foundational moment in being able to question every element of how we work. And the one that I think is questioned the least is probably the hierarchy and probably about how we manage communication flow around the companies. So if we're in a world where we are today, where Block, for instance, is completely remote, or we're remote first, every single thing that we do creates some sort of artifact, whether it be a Slack message, an email, pull request, code, you know, all these Google document, a meeting that we record, all these things have these artifacts of information about how the company is is working as building as failing is making mistakes all these things and traditionally we've been relying upon humans in a management structure in a hierarchy to go up and down a chain to relay that information instead we can take all of those artifacts and put an intelligence on top of it build a model around it and actually have a conversation with the company about how the company is doing it.

4:59And it's not just me as CEO that can do that, but anyone in the company could have that same sort of access to information and same understanding of like what the company can do. So you get to a point where you can build these world models for companies like treat the company as a mini AGI, for instance, an artificial general intelligence, because it really is. I mean, if you look at the company, it is an intelligence. But it hasn't been structured in the way that's the most efficient or the least lossy in terms of information flow and what people can actually do within the company. So the technology is good enough today that we can actually model the company.

5:42We can have everyone in the company put in intent, which would be strategy or these artifacts. And we can also have everyone in the company query it as well. And it just, it really opens the door for what's possible. Like, you know, Roloff and I have a board meeting every quarter where we construct a bunch of board docs, slides, presentations. We get only so much time for them to have questions. But imagine if every single board member can just query the company and have a conversation with the company's intelligence in real time. And we can make that meeting time that we have every quarter really focused on more of the creative or bigger existential decisions and issues than the day-to-day.

6:28The same can be said for our earnings call and analysts, giving them fully Reg FD possible information that is on their timeline with their questioning. But you can scale this to any position or any role in the company, which is pretty phenomenal. And we've just never had that ability before. And now we do. And, you know, I think the architecture and the structure of the company is ultimately going to determine its velocity and how well it's roadmapped for customers correct.

6:58Brian Halligan:Mm hmm. And just kind of drilling down on it. Can you describe sort of what the organization looked like before? And granted, you're very early in this, but what's it look like now? Which roles have been eliminated? What are the new roles? Just kind of drill down a little bit into what block looks like today and kind of where it's going. Yeah, we are early in it. I mean, just from a, you know, one measurement of how far along we are would be the depth from me to any other individual in the company. And I would say our max depth right now is probably five folks between me and anyone in the company.

7:44I would want to get that down to two to three this year. And in the most ideal case, you know, there is no layer. Everyone in the company reports to me. And that would be, you know, all 6 ,000 of the company. And that feels somewhat, you know, ridiculous when you consider like the old structure. But when you consider that the majority of our work is going through this intelligence layer, it's a lot more manageable. And that goes into the roles going forward. We want to normalize down to just three roles. The first is an IC, which is a builder or an operator. This is a salesperson, it's an engineer, it's a designer, product person, like, you know, whatever it is, they're actually working with the tools to build or to operate the company.

8:39They're augmented because they have access to agents. So, you know, one person can potentially do the work or explore the breadth that, you know, it would take a team or, you know, 10 people to do in the past. So that's number one. And I think there's a durable human skill that lasts there, which is judgment and taste and creativity. So that's probably the largest part of the population is the builder and the operators, the ICs. The second role is the DRI. And that's someone who can own the customer outcomes. They're putting a strategy together. They're understanding what roadmap allows us to solve customer needs and problems.

9:21And they're assembling a team of these ICs to get something done. But the durable human skill there is ownership and accountability. They're really owning the outcomes and whether something is failing or not. And then the last role would be what we consider managers today, which we're calling a player coach. This is someone who is building the capability and the capacity of other humans and their craft. But instead of telling them how to do it, they're showing them how to do it by doing the work. So these are people who might be ICs or they might be DRIs, but they're also really good at giving coaching skills, at a coaching skill to help the people around them get better and better and master their craft.

10:12And today, that's a management structure where ICs and even DRIs report to a player coach. But I think in the future, it's an assignment. It's not a reporting structure, but I'm assigned to ICs or I'm assigned to DRIs to help them master their craft. And obviously the durable human skill there is building human capacity and coaching. And there's a lot of empathy there and, you know, all the soft skills that we recognize great managers are known for. But not making a requirement that they have to be strategic necessarily. They have to build because they need to show off the skill and teach in that way, but they don't necessarily need to be a DRI.

11:02In very rare cases, one person can take on all three of those roles. I think I can take on all three of those roles. My leadership team is expected to take on all three of those roles. They're expected to build as an IC or operate the company. They're expected to be strategic and actually think about roadmaps and customer outcomes. And they're expected to coach and help raise the skill level around them with the people that they work with, with their direct team.

11:29Brian Halligan:Just a little kind of before and after, and you're mid on this, and let's pretend like it's working great and it's to your channel. and it's extremely flat like what is what is the role of the ceo before and after how's that most of the listeners by the way are ceos um what does like planning look like before and after um what happens to all those people who are directors and managers or the kind of glue people like what are kind of the before and afters in your mind i think um my my my job as ceo in the past I've thought about in three ways. Number one is to ensure that we have the right principles and the right team dynamic.

12:11And that's like hiring and firing and setting values and setting the culture and setting the tone. The second is to ensure that decisions are being made in context of our customers and industry trends and competition. And the third would be to raise the bar on our execution, like to constantly increase our capability and to push ourselves and to doing things that are uncomfortable so that we're growing all the time. So that's how I thought about my job up until this point. I think in the future, you have those elements, certainly, but I think it's more about like the architecture of how the company as an intelligence works.

12:53And like, if we're building the company as an intelligence, our job is as humans. And my job as CEO is to constantly align it to where we think the right outcome is. And I, I see like, you know, visually I see the, the company, the intelligence, the world model for the company in the center, and then on the edge of the humans who are just constantly aligning this towards customer outcomes. But even that changes, I believe, because I think a company's ultimate limiting factor is its own roadmap. And I think what these technologies point to are that our customers are going to, they're going to have the expectation that they can ask for a feature that doesn't exist on our roadmap and that it just is served to them.

13:42and that's where like you really get into the the layers of like okay so what do we actually build we build up capabilities which are effectively our tools like we can issue cards we card acceptance peer-to-peer lending all these things we do as a financial technology company we have these interfaces like square which has a register and a dashboard we have cash app we have title we have bitkey we have proto these are interfaces these actually touch the real world touch humans and we We can deliver our capabilities in these interfaces. Today they're built with like these very specific navigations that are our roadmap and our understanding of what our customers want.

14:21When you move to the third layer, you have proactive intelligence. We have all this understanding of our customers. We're moving money. Money is the most honest signal in the world. You can lie about literally everything, but when a transaction occurs, like that's something that like really tells the truth about your life or your business's life. and based on that we can actually prompt our customers instead of waiting for them to prompt us or having the right question to ask so we can do the very simple but very valuable thing of like how do we protect our customers cash flow like we have people using us as a bank account how do we make sure that they can pay the rent and they can pay their spotify bill and they can pay their kid's allowance and and and this is all sequenced in a way that allows them to never go to zero never go negative and have some cushion that allows them to even think about getting to saving or building wealth and not just you know peace of mind i think is the most critical aspect here um so if if we're enabling a customer if we're able to prompt a customer and Also, they can ask like as a business, hey, I have this inventory thing that I've been using, but it's missing this feature.

15:37Can I have this feature? We should just be able to build that and compose it in real time based on our capabilities. If we can't and it points to a deficiency in capabilities or a gap, that's our roadmap. So our customers just by using and talking with our systems are telling us what our roadmap should be. And then it's up to us to give the judgment. And then the final layer is the world model. It's the customer world model. It's the company world model, our deep understanding of ourselves and also our customers. But I think like if I had to say one thing that, you know, myself or anyone in the company has to do, it's I guess it's this overused phrase of judgment.

16:20but it is judgment against like what we intend to build in the world and is it on is it aligned with that judgment is it aligned with the values is aligned with the taste that we have and is it unique or is it not and i i guess in for my part i'm the um extra checkpoint on like is the alignment circle of humans the edge of humans actually working correctly in this i see how this works

16:45Brian Halligan:exceptionally well for block you have a lot of signal how does this work for i don't know sierra workday somebody who doesn't have the quality signal the frequency of signal does it apply or it doesn't work as well i think it probably does it i think it it goes down to like are you building a are you building a business that like understands something of human nature deeply and It gets deeper every single time. And that's a real tangible signal that just doesn't go away. And if you are, then I think you can build your company as an intelligence. If you're not, then it's probably an add-on to something else.

17:25And I think most of the industry is thinking about AI as like a co-pilot, as something that is augmented onto, rather than like, how do you just rebuild this, our whole company with this as the core? Like, and if it doesn't make sense for your business to do that, and you end up being or looking very similar or rhyming too closely with the frontier labs, then I think it's going to be a very, very challenging to differentiate and, and, and, and survive. And that's kind of what's been leading me to all this is like, I just, you know, since January of 2024, which is when these tools really came to bear, like Goose, which is an agent coding harness, was one month before Cloud Code in January of 2024.

18:16And Cloud Code came out that following month and then like was really put out of beta in May of that year. and you know that whole year i just spent every single day for three hours every morning just pushing myself like can i get it to do something that i didn't think it was capable of i didn't think i was capable of and every single day it worked like every single day i was surprised and it's you know it's it i'm sorry it wasn't 2024 it's 2025 it's only been a year of those tools like one year it just the the compounding nature of this um is pretty incredible so you know being able to like uh see that understand it and then shift your company to be ahead of it i think is absolutely critical right now and i don't think people are feeling it enough if they're just living in this abstraction like oh yeah like these tools will make everyone in our company 10x more productive i don't think this is a productivity thing i think it's a structural thing that needs to shift i think you're right

19:23Brian Halligan:i think people are thinking of as co-pilots and individual productivity versus your i think your idea is complete business transformation and i think it's super interesting and compelling can i add one yes um thought i've had as i've listened to this so far One of my favorite pieces of writing ever is Adam Smith, Wealth of Nations. And this idea that if you have the right signals, you can rely on the self-interested behavior of many small participants in the system to actually lead to optimal outcomes. You just need to have the right framework and the right system. I think the way that many companies work today is a little bit more command and control.

20:02Brian Halligan:Very much. And you have hierarchies and you have political actions and people, you know, jockeying for position. And it's not always clear what is actually true. And I think part of what we're envisioning here is you have a system that's just ground truth. And you do away with all the layers. You get back to the kind of productivity that founders often long for. They long for the days when there were 100 people and not 500 people because they were so much more productive. Why? Because you had, you know, lots of transparency, limited hierarchy. And so I think the possibility here, as Jack was saying, is instead of just focusing on individual productivity, you reimagine how we work together as humans.

20:41And you can do it with far smaller number of people that are far more productive, yes, but there's a different way of working together where you get the right signals. and it's much more similar to a capitalist system where the signals tell you what to build. It's not somebody who pounds the table harder who gets their product approved, which is, listen, more customers want this than that. That's how we're going to decide what to build next. So for me, there's something quite magical in that realization.

21:07Brian Halligan:Yeah. About four years into building HubSpot, I eliminated no org charts, no titles, and everyone lost their mind. They didn't like the idea. And we ran that for about nine months and I got worn down and I brought it back. The thing we didn't have was an intelligent system that had all those signals that could help us make decisions. So it's exciting to see this year. I have a theory I want to run by you, Rulof. I sort of think of, like, there's manager mode. It's a pyramid. The VPs make most of the decisions. And there's founder mode, kind of flat. The founder makes a lot of decisions. and then there's, I'm just going to call it Dorsey mode.

21:49Brian Halligan:It's a circle and the AI makes most of the decisions. Do you buy that? Oh, I don't think the AI makes most of the decisions. I think that, and Jack should correct me here, I think the AI helps with communicating the alignment and the management team or the inner core helps set the framework. What is the objective function? Are we optimizing for growth rate, gross profit per employee, net promoter score, probably some combination of those variables. And the humans at the edges perform an incredibly valuable function of correcting and informing and sort of steering that. And I think one of the phrases that Jack has, which I absolutely love and I've stolen so many times, is companies have multiple founding moments.

22:30There's so many smart people in your company that have clever ideas that every day inflect a product, introduce something new. And so this idea that there's just one person who is the brilliant person who comes up with everything, I just don't think, I don't believe in hero worship or the converse to that where you sort of scapegoat people. I think it's harnessing the best of your team to really advance the company. So I subscribe to the circle idea. Yeah, I also don't think the AI is making the majority of the decisions. I think it's facilitating a more context-rich decision. I think ultimately, like in the most ideal case, our customers are actually making the majority of the decisions because they're just based on their queries and what they're trying to do with the system.

23:18Deliniates like where our roadmap should go. And then it's up to our judgment as to whether that's consistent with what we want to be and what we think is most strategic or not. but we just you know we weren't able to get to that level of data fidelity before because we had to infer it we had to do customer research we had to do interviews we had to do look at our customer support you know product feedback on on on twitter all these other things but when your interface is a conversation with your customer instead of like this visual navigation, you suddenly get like this amazing fidelity of like, what do our customers actually care about?

23:59What do they actually want? And it's up to us then to decide if that's consistent with what we want to be as a company, or if they should be going elsewhere to do that. I think all these things are going to blur. I mean, that's the craziest thing is like, you know, again, since last year, I've just had this existential dread and also like hope and optimism at in the same hour in the same thought process of like what is it what is even a company going forward like what like what are these structures going forward and this is the only like durable one structure that i could actually imagine lasting for quite some time yeah and it was coming from a place of like wow is our company just going to be completely irrelevant the you know in the next coming years or or even even sooner like what do we what do we actually differentiate on what what do we have a moat around and what what do we need to be to defend and and also to to grow that um and like what are the and all that follows from customer expectation like it the most amazing thing about open claw to me was that um people wanted to take this thing and contain it into a Mac mini and make it very tangibly theirs and have all this agency around what they did with it.

25:25And we're seeing Square sellers do stuff with it like that, interface with the Square APIs. And we're seeing Cash App customers. And these are like not tech people. These are just like people I want a bot to help me manage my life. that agency, independent of your thoughts on like how good of a system open call is right now, it'll get better and better. But the intent behind that is, is agency. I want to tangibly control this intelligence and, and for it to, to better me. And what it makes possible for me is incredible. And that expectation floor has just risen dramatically. And, and, and that leads me again to like, yeah, our limiting factor as a company is our roadmap.

26:12Like we need to remove that from the equation. We need to ensure that our customers are truly building alongside us. Um, and, and that they, you know, are seeing us as a, as a series of capabilities that makes their desires fast and easy and, um, and valuable. So like, it really goes back to the capability set that we have. and then like the intuition of the interface that we have and then like how intelligent our world models can be to be helpful and to compose UI in real time.

26:48Brian Halligan:I spent this morning at a company called Rogo here in New York City. They read your article. Let's say you're advising the CEO of a 100-person company. They've already got their hierarchy. What should they do? Should they start with their system? Should they start with data? Should they start with the org? like how should people go about you know running the dorsey playbook uh i don't know about this dorsey playbook thing but um but we don't have it all figured out i think there's an important dose of humility that we have that we're endeavoring on the path we believe it's right but we know there's a lot to figure out yeah it's more like if if you were to at 100 people or even you know just starting today if you were to build your company as an agi as an intelligence what would it look like and what would you need to really differentiate like if i were starting a company today i would be so excited about how quickly i could build things and how quickly i could prototype and get things out to customers i would be um in this like valley of dread about distribution and attention because there is so much noise out there and it's so hard to get to the actual signal of like who's building something interesting that will actually fundamentally change something and will be around for more than a year yeah it just like i think distribution really becomes a differentiator and i think there's some event horizon where the the way we think about distribution today closes off you know it like there's there's um the you know apps, for instance, and websites and like the traditional retail.

28:29There's a number of things that will change. And if you don't have the distribution today, it's going to be very hard to fight for that. But there'll be new areas of distribution that are probably more important. And I don't know what those look like. But I would say like, I would assume that a company of 100 probably is no more than two to three layers deep, hopefully. and now would be the time to just like really question like do I do I need a hierarchy like Brian a year into square I also we removed titles we normalized everyone to lead yeah we did it for like we were talking with all these banks all the time and you would have these EVPs and VPs and they were looking for the same on the business card and there's this whole business card culture so we ripped up all the business cards we normalized down to a title of like you're a lead of what and the longer that is behind lead is probably the farther down you are in the organization um and we've kept it like we we you know we don't have titles we we have like you know what do you lead um going back to like the dri thing what are you ultimately responsible for and i i think it's helped us a lot um but this is another step like if if you're starting today or you're 100 today like what is actually fundamental to solving your customers problems and where is the hierarchy getting in the way of that?

29:54And look at all the tools you're using. Like look at all the information you're generating just by doing your work. Like just putting that into an intelligence and being able to query it will give you an understanding of the company that like is two to three times more than you had before ever because you're relying on people telling you things. and you know that doesn't always happen for various reasons that you know roloff spoke to in terms of agenda or politics or emotions or empathy all all these things if imagine if your company was entirely legible like entirely legible every aspect of it and we're not far off from that from a data perspective it's putting the intelligence on top of it and making it useful and then making it proactive like that's that's the hardest bet is like we can determine causal getting to predictability for these world models around the company and customer is is still right now very much a research but um but like it's it's one that you know it's imminently solvable can you guys just take me behind the scenes like jack this was a really bold move you laid off 40 of your employees just like for ceos listening what was that debate

31:10Brian Halligan:inside like how big it should it be how bold should it be you know ruport's got this good line if you're gonna if you're gonna eat a shit sandwich don't nibble and you seem like you took a big bite what was that debate behind the scenes then ruloff like how did he pitch it to you guys and what was the board's reaction just like behind the scenes on that yeah so um it was something you know, December of last year is when the is the when the models really got a like noticeable upgrade from being able to be really good at building prototypes and grid field efforts to like understanding like large code bases and legacy code bases like our own.

31:56There was, you know, hallucination wasn't much of a topic in terms of like the coding ability and the tool harnesses, we became suddenly very mature, like just in that month. And it just like everyone went home for the holidays and everyone played with these tools and they were surprised at how capable they were and what they could do with them. And, you know, we came back and the conversation was like just going around the table. Would you build the company this way? if you had these tools today? Like, what would the company look like? And everyone around the table and my team just said, like, it would not look like this.