ALISON BEARD: I’m Alison Beard, and this is the HBR IdeaCast. We have a special episode today. In fact, it’s the first in a five-week series featuring conversations from our 2026 HBR Innovation Summit. For that event, we brought together CEOs from a variety of industries to talk about driving innovation in a time of rapid technological change. Adi and I interviewed them, and we also took questions from our live audience of senior leaders around the world. We’re going to kick off this special bonus series with Jack Dorsey, co-founder and chairman of Block, who also carries the title Block Head. Over the past few years, he’s integrated all of Block’s customer seller ecosystems, Square, Cash App, TIDAL, Proto, and more, and he’s now reorganizing the company around intelligence. Here’s our conversation.
Jack, thank you so much for being here.
JACK DORSEY: Thanks for having me.
ALISON BEARD: Yeah. So let’s start by going back a few years. In 2024, you reorganized Block around functions rather than business units. Why was that an important first move to set the stage for the intelligence-driven innovation that you’re pursuing now?
JACK DORSEY: It was more getting back to how we started. We were organized by discipline, by engineering, design, and so on and so forth, when we started the company in 2009. And we didn’t expect to have an entirely new business line in Cash App. We started with Square, Cash App grew quite rapidly after about five years of Square. There were two mentions of it in our S-1 when we went public, but then it just took off and we decided that we would experiment with having a separate business within this larger business. So we had these two separate ecosystems, we ran them as separate companies, but we realized we didn’t want to end up just being a conglomerate of loosely coupled businesses. We wanted to be two ecosystems that actually benefited one another. So it made sense to remove the business structure from them and move back to a discipline base where we can really focus on getting all of our engineers to mastery, getting our designers to mastery, getting our finance and counsel folks to mastery, instead of just serving a pure business function and having differences along the way.
So now we have standards across the company and it made it much easier to deploy similar tool sets and policies and culture. Each one of these businesses back in the day had different cultures and they were using different tools, and there was just a lot of duplication that was unnecessary, especially as we entered more into using AI tools, which really normalized a bunch for us and made it much easier to just move faster because we weren’t choosing all these different vendors and tools.
ALISON BEARD: Yeah. So you acknowledged back then that that initial move was disruptive and uncomfortable, as is the more recent shift to organizing around intelligence, which involved layoffs this year. So how do you mobilize your people to embrace a change that’s going to radically shift their roles and might eliminate some of their work?
JACK DORSEY: I think there’s many times in my career where I had to be comfortable with not seeming credible to folks, and that I’m making decisions that I believe in that may not feel credible or decisions that are wanted throughout the business. Any change is going to be disruptive, and it’s not often the case that change is embraced. But when I feel very strongly about something, I get feedback from a bunch of people that would have differing opinions, and it built a lot more confidence in the action. I’m okay with losing a little bit of credibility if that outcome over the longer term is going to be successful, and it’s going to take time for folks to get there.
And it generally worked out quite well, but it’s hard along the way, but it gives me a sense of empathy of what I have to do, as a lead of the organization and how I should see the organization from everyone’s point of view, and then just make sure that I’m communicating that all the time. It’s really like why are we doing this, why is it important, why does it set us up for the future? And not everyone’s going to agree, not everyone’s going to believe it immediately, but it’s my job to prove it as quickly as I can.
ALISON BEARD: Yeah. Okay, so now the big question. What exactly does it mean to turn a company into an intelligence? How does it practically change the way that the company is structured, how it operates, and most particularly how it innovates?
JACK DORSEY: I mean, I don’t think we often think about it, because it sounds a little strange, but a company itself is a technology. The way we organize, the way we build into departments, how we meet, how we structure our communication, our collaboration. We don’t necessarily see it as a technology because we think of technology as electronics, but it’s a technology that serves distributing information and distributing tasks and accountability. And it’s something that when we started our company, we borrowed a playbook from Google, which borrowed it from Silicon Graphics, which borrowed it from Silicon Graphics, which borrowed it from GE and so on and so forth, all the way back to the railroads and all the way to Roman times.
And it really, if you look at what it is, it’s this technology around information flow. And the hierarchy around a company is really to make that information flow as efficient as it possibly can, but all the intelligence of the company is within humans’ heads and it’s all dependent upon these information flows working, and the speed at which people can organize the information, provide context, disagree with it, answer questions, ask the right questions and whatnot.
Now, every single interaction we have in the company usually leaves some sort of written artifact behind. We’re a remote first company, so everything we do has an artifact associated with it, that can be queried and ingested by an intelligence. And now we can organize all the information, all the working artifacts that we have, into one surface where anyone can query it at any time, anyone can act on that information at any time, and it really diminishes the need for the communication hierarchy. And then you get down to, like, okay, so what are our roles? How do we think about our work? And to us, it came down to three significant roles, which is individual contributor, which is someone who builds or makes something or operates something, a directly responsible individual, which is someone who’s strategic and can take accountability and build out a roadmap, and then a coach, which used to be called a manager, and that’s someone who’s tasked and assigned us to an IC or DRI that can help them master their craft.
So we have those folks for engineering, we have them for design, we have them for finance, but it’s no longer a command and control structure, it’s more of a coaching assignment. Because ideally we want everyone in the organization, especially the ICs and the DRIs, individual contributors and the directly responsible individuals, to have a lot of agency. Now they have all the information around the company, they can pretty much build anything they want, and they can be very, very close to the customer, which means they can make context rich decisions and act on them almost immediately. And it turns out that the chain of command and the communication hierarchy, we’re just slowing a bunch of that information down, or making it so lost because someone didn’t want to bring this up or they forgot about this context or forgot about this idea or question. Now we get to see everything and we can make much faster decisions with much richer context with a lot more data.
And it just seems to me that is the company structure in the future. It’s a lot more flat, it’s a lot more organized around assignment rather than command and control, and everyone has access on a level playing field to the very data, the very intelligence of the company. And we can also build systems based on our knowledge of the company that can help us predict the next actions the company should take, and then we get to decide if we actually want to do that or not, based on everything that we’re seeing.
ALISON BEARD: I think that gets at my next question. So with all of this bottom-up innovation, how do you ensure that the work everyone’s doing is adhering to the company’s long-term strategy?
JACK DORSEY: I think humans are the edge, and they’re the ones that are aligning. You have this core central intelligence within the company that represents everything that we’re doing, and around the edge you have the humans, and they’re constantly aligning based on what we want to be, based on our values, based on our purpose, based on what we think is strategic or not. We have customers every day telling us what they want in our products, and it’s up to us to make those decisions, as to whether that fits with our direction, our values, and so on and so forth. And because it’s so easy to build something, what becomes hard is having the right judgment to determine if something is correct to go out to all of our customers or to live on, or to take the resources that we’d have to exert in order to build it or to maintain it.
And that’s us understanding who we are and who our competitors are and who our customers are and what our goals as a business are as well. So in terms of someone building an AI system or a general intelligence, you have humans constantly working to align that to particular directions or sometimes goals or sometimes outcomes. And I don’t think it’s all that different from what you would do with a company as well. Again, a company, if you really look at it, is a technology, it is an intelligence already, but it’s not as legible as what we can make it today. Now we can make it almost perfectly legible, and that gives us a lot of great insight on where we are and where we want to go.
ALISON BEARD: Yeah. Speaking of where you want to go, how are you and your top team staying on top of the latest developments in AI that might be useful for your business? It’s developing so rapidly.
JACK DORSEY: I dedicate three hours in my morning. I open this site called arXiv, which is just research papers, physics, biology, engineering, mathematics, and I read either two to three papers, and sometimes I only get to one because I want to go deeper on the subject. And that just makes me think in a different way, but I also get to use the tools to help me better understand whatever the subject matter is, and I get to see what the tool boundaries are currently. I spend a lot of time building with the tools as well to, again, understand their boundaries and understand their capabilities and what I can ask of our team and what’s changing. I’m a big proponent of open source, so I really appreciate and I’m grateful for the open source models and the iterations on how to build intelligence. We have a significant focus right now as an industry on LLMs, but there are other paths towards intelligence.
World models is one I’m particularly fascinated with and it corresponds a lot more to our business, because we’re not predicting the next word in a language, but we’re predicting the next possible action a customer might take. And we have this really unique signal in money and transactions and it’s a very honest signal. So we can build this world model for our customers about how they spend their day and literally spend the money in their day, and the same thing for our sellers. And we can get to fairly accurate predictions on what sorts of actions they would take next, and how we can help them most importantly. And that can lend itself to what actions should the company take next. We can build a world model for the company itself based on our understanding of where the company is and what it’s doing. So by exploring the boundaries of the frontier every single day and pushing myself to use the tools and see something new every single day, I feel fairly forward of where the field is right now.
ALISON BEARD: Predicting the next, not the next word, but the next behavior is pretty cool. So you’ve developed several open source agentic AI tools, goose and Buzz are two of the names. Your teams use them, you’re now bringing them to other companies. How do you know when an internal open source tool is ready to be commercialized?
JACK DORSEY: Well, with goose, within a month, we had a number of companies such as Stripe fork it immediately for their own usage. And if there’s a will to fork it to make it your own, there’s also a potential for us to build a business around it. And it would be the case that the more that people want to pay us to host the tool or to constantly update the tool versus fork the tool and do the development work themselves, is an indicator of our success in building something valuable enough that people would want to pay for. So I think it’s really around that. Open source is a distribution mechanism for us. It allows a lot of people to see what we’re doing, it allows a lot of developers to extend it and then give back to things like goose and Buzz and all of our Bitcoin efforts as well.
But also it’s a signal on is it valued or not? And we can take that and we can say, well, here’s the things that become a little bit harder. Open source, I can host it, but hosting might not be something I want to deal with. Hooking up all the models might not be something I want to deal with, so I’ll pay for inference. Hooking this into my company to deal with all the compliance and security and risk issues and all the systems I use on a daily basis, I don’t want to deal with that, so I’ll pay Block to make those connectors and supply them. So there’s places in the chain which are valuable, which ultimately save customers time, and I think the easiest exchange of value is paying to give me time back. And that’s where I think we can play with our open source offerings, especially these machine intelligence tools.
ALISON BEARD: With all of these initiatives, how do you measure success?
JACK DORSEY: If customers use them, then value them, and they value them if they use them more, and they find new interesting use cases for them. I’m a big fan of Clay Christensen and Jobs to Be Done, and the idea that true innovation comes from you putting something out there and you see a customer using it in a way that you didn’t expect, and then you support that, and you get a little ahead of it. It’s exactly how we started Twitter. We started with updates to the world. We saw that people were @ mentioning each other. They were literally trying to have a conversation. We made that a feature and suddenly it became a conversational platform. That’s not what we planned, that’s not what we expected. We had very similar aspects within Square and also Cash App, that we’ve seen in goose and Buzz. And it’s just seeing those threads and observing what your customers are trying to do but failing, and building for that.
I think that’s ultimately where true innovation comes from, because it pairs a really strong skill of customer observation, looking at patterns, not just asking customers what they want, but looking at patterns across a mass, and inferring what they’re actually trying to do with this thing that you didn’t plan for, you didn’t expect, but you could support, and it may open an entirely new line of business or an entirely new use case that grows the business itself. So obviously I look at our core metrics, but also the potential of here’s something customers are trying to tell us that we should be doing. And again, if you see the company as an intelligence, they’re telling you that every single day, but it’s very hard to see. So now we just have much better insights in terms of what they’re trying to do with our tools and our systems that we can act upon much faster.
ALISON BEARD: So we’ve been talking about all the opportunities presented by recent advances in AI. There are obviously a lot of risks too. There’s been much recent news about rogue agents, cyber attacks. So what is your nightmare scenario? What keeps you up at night for AI going wrong at Block or in the broader industry, and how do you protect against it?
JACK DORSEY: I don’t know if I have a nightmare scenario because I’m an optimist, and I think even in the security cases, all these things are races. You can never build a perfectly secure system, you can only be 10 steps ahead of your adversaries and of adversarial activity. So if you want to build a secure system, use the same tools to constantly upgrade the security, and it’s never going to stop, that work never stops, then you just move. Life is motion, and the more we can move and the more we can build, we can keep the adversaries and any issues at bay. And I think the more eyes on a system, the stronger that potential becomes, which is why I’m such a huge believer in open source and open models and open weights. I think it’s really important that we all have this intelligence at our fingertips so that we can even get at the level where we can build secure enough systems that can be resilient to very capable models attacking them. And I think that’s absolutely possible.
But if it’s locked up in a few labs or we’re pacing that frontier, I think it’s quite dangerous because you don’t know who’s able to use that and why and what they’re trying to do with it. So I guess that would be the only thing I worry about is a significant concentration of power controlling this intelligence, but at the same time, I don’t believe that can happen. It can’t be contained. And I think it’s folly to think that it will be contained. It’s just, how do we deal with this? Knowing that this exists and it will always exist and always grow, what are our policies, going forward? How do we act with it? How do we be good stewards of it? What are we learning from it? And how do we stay ahead of the adversarial energies and activities that will definitely come our way? But we’ve always proven to be able to do that. It’s just a decision to be more decentralized and giving more and more of that power to more people, especially when it comes to intelligence.
ALISON BEARD: And as you build more of block around AI, how much does that strategy depend on cheap, readily available computing power?
JACK DORSEY: I think the LLMs are extremely inefficient right now and they require a lot of power and a lot of compute. I don’t think that’s the only path. There’s a long standing history in machine learning and deep learning around world models that are much more efficient and take much less resource. We released a voice model called Dessa, which is very human-like sounding, and it’s trained entirely on speech sound, has no knowledge of language, it just understands sounds and it tries to predict the next best action and sound. And we train that on very few resources and we can run the whole thing on someone’s laptop and we’ll eventually get it to someone’s phone. So I think it’s more interesting to me to be able to decentralize a lot of this compute and utilize it. And we are not at the efficient frontier of where these models can be, even trained on, and how they can operate. So I don’t believe in you need massive, amounts of compute and power to do something interesting in this space, and Dessa proved it without a doubt.
ALISON BEARD: Great. Well, I want to turn to some audience questions. So Simon, who’s a VP technology advisory at CGI asks, he’s interested in how you think with the forward deployed everything model at the moment, which is very much grounded in command and control. How do you see that evolving to be more intelligence led team pod as Block’s doing?
JACK DORSEY: I think everything we have in a harness like Claude Code or Codex or even forward deployed engineers, I believe is a gap in capability from the model itself. And I think the models will become more and more capable with where those things become less necessary, even the harness itself. So I think it works for now, we need bridges to get to these technologies. The alignment aspect I think is more important. Like, what do I want this thing to do, what do I want it to be, is it aligned with our values and our purpose and whatnot. And being able to make quick, informed context rich decisions based on that, I think becomes a real skill and is very, very hard to do. And it hasn’t necessarily been the most valuable in the past or it’s been watered down.
So I think someone paired with a really, really capable model that has really exceptional opinions and a point of view that’s differentiated, and taste, I think all this comes down to I have a really strong point of view and I have the courage to get it out there, and the model makes it happen.
But I don’t think that requires a forward deployed engineer or a lot of command and control because it’s just so much agency for the individual. And as an organization, it’s like how we align all that energy. Again, it all comes back to alignment for me.
ALISON BEARD: Yeah, so this is along the same lines in terms of individual agency. Reema, who’s the global head of design and strategy at Hitachi, says that it’s great to hear about your discipline specific organizational structure. How do you view that structure when given the AI tools available today? There are discussions that product can do design, design can do engineering, et cetera.
JACK DORSEY: All these disciplines have different ways of thinking. An engineering mindset is very different from a design mindset. While they can do tasks that are not in their discipline, the problem solving abilities and the way they approach problems is quite different and I think complimentary. And I think going deeper within that, how you face problems and how you approach them with these tools is great, because I could be a great engineer, using a model to build something and asking it to design an interface for me, and it may look great to me, but I don’t have a lot of experience really tuning the feel of it to provide the right experience or to provide an intuitive interface that people don’t even have to question, they look at and they know exactly what it does. That takes a lot of thinking, a lot of experience with tools that you don’t like using, and the ones that you love and being able to articulate why.
And to be able to do that for technologies and also interfaces at the same time is quite rare. There are people that can do all those things at once, but it’s quite rare. So I appreciate the depth of each discipline because it’s really how they face problems and solve problems. And now that they have tools to see different facets of the same work, but mastery of those different facets is a very different thing. And I don’t think that that goes away. I want to be very deep in one discipline and know enough to be dangerous in everything else.
ALISON BEARD: And do you think that younger employees have the ability to develop mastery now that they do have-
JACK DORSEY: Oh, yeah.
ALISON BEARD: … access to these tools, and make those jobs easier? Okay.
JACK DORSEY: Oh yeah, because the ability to see more. In the past, in my career, if I’m facing a problem, imagine a tree, the roots of a tree, and I pick one path and it takes me a very long time to first pick that path, and I have to be more or less correct about what that path is because I’m going to go deep on it. And then I realize, “Oh wait, wrong direction.” I have to go back up, start all back over again. Now with these tools, at the root, I can see a superficial level of all the potential paths, at least 10 of them or 20 of them or infinite amount. And then I can really think about and pick one and I can go a little bit deep. And if I like it, I can go deeper, or I can pull back up and go deeper on another one.
So the cost of experiencing more diminishes completely. And I think true taste and judgment comes from you experiencing a lot and forming an opinion on it, and developing that point of view above all else. That’s the superpower to me.
ALISON BEARD: I like that visual analogy, that helps me understand it. Switching tacks, based on your startup experiences, what advice do you have for a COO who runs two portfolio companies for a family office, who’s used to conservative traditional tactics to drive the business instead of blue ocean strategies that can catapult to achieve 10 times growth? How do you get traditional business people to adopt the kind of mindset that you have?
JACK DORSEY: I mean, I don’t think you can ask them to do it, I think you show them what’s possible. And the only way I’ve found a changed mindset is to put tools in front of people and ask them to do the simplest thing they can imagine with it. Just get them a taste of the power. And then if they decide to go deeper with it, they’re going to own it. And not everyone’s going to want to do that, and that’s fine. You’re not going to be able to force it then, because if you force it, it’s just going to create a lot of friction for everyone. So they have to want it and they have to be curious enough to own it.
But being able to take that first step is really important, and just making it as small as possible so they’re not overwhelmed. I think people, they just try to start with too much. An analogy, I’m very familiar with meditation, and people start with a practice of meditation, they’re start like, “I’m going to do 30 minutes of meditation every single day for the year and that’s my resolution.” And they do it once and they forget to do it the next day and it’s too much trouble the next day and something’s going on. If instead you start with, “I’m going to meditate for one minute every day for two weeks,” it’s no investment, but it gets the behavior in, and you can replicate and build on the behavior. So then after that two weeks, you never go beyond the one minute for two weeks, and after two weeks, you can expand that to five.
And again, it’s just like it’s getting the behavior and the habit built in, and then you can expand to 10 and 15 and an hour, whatever you want. So all skills to me are like that. Spend one minute exploring this tool for two weeks, force yourself no more than a minute. And then after two weeks you spend 15 minutes and whatever that ends up being. But getting that habit, getting that behavior ingrained, I think is the most critical part first.
ALISON BEARD: That reminds me of James Clear’s Atomic Habits. Good advice.
JACK DORSEY: That’s where I got that practice from as well. He laid that out so well. I was forgetting the name of the book, but thank you.
ALISON BEARD: Yeah, yeah, yeah. So Andrew, who’s the director of transformation at PRS for music, wants to dig a little deeper on the human in the loop idea or the human decision maker idea. So he’s asking, “With flatter organizations, how do you make sure that every decision doesn’t have to go to the most senior management, to encourage the pace of innovation and embracing risk?” Where are you slotting in those decision makers?
JACK DORSEY: I think that kind of started with maybe three years into running Twitter. I wasn’t a CEO before, and it really forced me to think about what is my job? What am I supposed to do here? I was an engineer and I was used to building things. So I answered it with, I believe I have three jobs. One is to set up the environment for people to work in. And setting up that environment means we have clear purpose, we have clear mission, we have a culture we can articulate. We have how we behave with each other and why these things are important. Number two, which gets at the question, which is, ensure that decisions are being made. And not that I have to make every decision, that’s a failure because I’m a single point of failure in that case, but that decisions are being made and they’re being made in the right context, they’re being made in consideration of our customers, industry trends, competitive trends, whatnot.
That I can see someone who’s making a decision has set that decision up correctly and is using all that available information and it’s contextual. And then third is to raise the bar on what we believe is possible. So an organization will always bend towards some steady state, and it’s my job to step change that a bit and question some things and question what we believe made us successful in the past. Not hold it sacred, not take it for granted, change it completely, remove it, delete it. And I don’t think that changes with these tools, and I don’t think it changes as the organization gets flatter, it gets even easier and better.
You can tell how people are making decisions and whether those decisions were successful and how quickly they realized that it was the wrong decision, and what they used to correct that action after the decision, and you just have so much more legible information around it. And it’s really a decision to pay attention to that or not, right? Oftentimes in the organizations of the past, we don’t pay attention to it. Great organizations do retrospectives and they’re really like, okay, we made this decision, and here’s why and here’s what happened and here’s what we could have done better and here’s what went well. And that’s a very, very healthy organization, but it’s extremely rare. And if you see that in an organization, that practice is usually only after something bad happens, not when something good happens, not when you’re doing everyday decisions. So these practices become stronger with these tools because you can provide even more context that’s company-wide.
ALISON BEARD: What advice do you have for leaders, maybe especially those outside the technology industry, about how to most effectively put AI to work in their organizations right now?
JACK DORSEY: Got to start with yourself. Put it to work for you. Build something, build a system that gives you time back, build a system that helps you learn something, but build a system. Not just ask the questions, but actually build with it, create with it. Have something that you can take away and show other people and be proud of. If you have kids, build with your kids, because you can see how they learn through their eyes, and I think that one of the most important things is to see that, because it changes how we think and gets us shaken out of our ways as we become older and more precious about what we have and not questioning things and taking everything for granted. But you’re not going to be able to see this within your business no matter what role you are, unless you can see it within yourself and how it changes you, if it changes you and whether you allow it to.
And again, these are all decisions you can make, and you can be diminished by it, you can ignore it, you can be amplified by it. It’s all a decision, it’s all your agency.
ALISON BEARD: Terrific. Well, Jack, thank you so much for joining us, I really appreciate it.
JACK DORSEY: Thank you all. Appreciate it.
ALISON BEARD: That was Jack Dorsey, Block Head chairman and co-founder of Block. Tune in next week for Adi’s interview with Aarron Atkinson-Toal about just how much influence companies have over what we buy, followed by the second episode in our special HBR Innovation Summit series. Adi will talk with the president and CEO of FedEx, Raj Subramaniam, on putting AI to work at scale, and navigating geopolitical and economic uncertainty. If you found this episode helpful, please share it with a colleague and be sure to subscribe and rate IdeaCast in Apple Podcast, Spotify, or wherever you listen. If you want to help leaders move the world forward, consider subscribing to Harvard Business Review. You’ll get access to the HBR mobile app, the weekly exclusive insider newsletter, and unlimited access to HBR online. Just head to hbr.org/subscribe.
Thanks to our team, senior producers, Mary Dooe and Rachel Swaby and senior production editor Kristin Murphy Romano.
And thanks to the HBR event production team, Dave Di lulio, Elie Honein, Simona Sparane, and Julia Butler. And thanks to you for listening to the HBR IdeaCast. We’ll be back twice a week for the next few weeks. I’m Alison Beard.
