Today, I’m talking with Willem Avé, who’s head of product at Square. You know Square — it was started by billionaire Jack Dorsey, of Twitter fame, more than 15 years ago, and it got big on the back of that little magnetic reader that plugged into the headphone jack of the iPhone and let small businesses accept credit cards.
Now, of course, Square is more than a credit card reader, and sadly, the headphone jack is ancient history. Square itself is part of a parent company called Block, which is made up of a really interesting mix of financial services like Cash App, the digital wallet and payment platform, Buy Now Pay Later service Afterpay, and quite a few other subsidiaries. Block also owns a majority stake in the music streaming service Tidal, for reasons that are still a little opaque.
And that’s where this episode becomes pure Decoder bait right from the jump: as you’ll hear Willem explain, Block just restructured the entire company last year into a functional organization, so that it shares resources across all of its units. For Square, that means it’s sharing engineering resources with Tidal, a company that makes an entirely different product serving an entirely different industry.
Verge subscribers, don’t forget you get exclusive access to ad-free Decoder wherever you get your podcasts. Head here. Not a subscriber? You can sign up here.
So I asked Willem about how all that works, and his answer might surprise you: he says Square, and Block as a whole, are better aligned now that they have a single, shared roadmap — one that, as you might imagine, is pushing very hard into AI and the kinds of back-end, financial automation companies like Square want to bring to business owners.
So sure, Square might not move quite as fast as if it were an independent division working autonomously. But Willem says the new structure is already reaping benefits to how the company develops new products and plans for this next turn, which he refers to as Square 3.0.
You’ll hear Willem and I dig deep here into what it actually means to enable AI for small business owners — what kind of questions might a Square user ask a language model about their sales data, and then how does Square actually turn that into actionable input that won’t hallucinate or just straight up fail to do what you ask.
Willem says the really meaningful shift for this kind of AI is to take the nondeterministic elements of LLMs and agents and link them to deterministic systems. That’s going to be how a company like Square moves beyond the gimmicky, often unreliable era of AI chatbots into meaningful automation for small business owners that helps them be more productive.
It is also a very weird time for money — when you combine what’s happening in decentralized finance and the Trump administration’s interest in crypto with the discontinuation of the US penny and the overall turbulence in the US economy. And Square is pretty deep on all of that, as a company that processes transactions for all kinds of cash businesses every day, while at the same time holding large reserves of Bitcoin and working to enable merchants to actually transact in cryptocurrencies.
It’s been quite a long time since I think anyone, anywhere, has wanted to give up any portion of a Bitcoin to pay for something except well… crime. But you’ll hear Willem make the case that enabling this kind of choice is part of Square’s mission.
By the way, one fun piece of trivia: after we wrapped the recording, Square got back to us and confirmed that it moves at least 16.7 million pennies between its merchants and customers on a weekly basis. That’s all going to go away now that the penny is getting discontinued.
This is a really fun one. I can always tell when I’m talking to a true Decoder fan, and Willem came ultra-prepared to talk about not only his org chart, but also how he makes decisions. You’ll actually hear him use a French winemaking term to describe part of his process, which might be as nerdy as we’ve ever gotten here on Decoder.
This interview has been lightly edited for length and clarity.
Willem Avé, you are the global head of product at Square. Welcome to Decoder.
Thank you. Glad to be here.
I’m excited to talk to you. It’s an interesting time to run businesses connected to the internet. It’s an interesting time for money, which is an odd thing to say, but it is an interesting time for money. I suspect you and I are going to talk about the end of the penny quite a bit in this episode. As somebody who has to accept literal cash at restaurants, it seems like we should talk about the penny.
That’d be great.
Let’s start at the start. I feel like everybody knows what Square is. Everyone’s seen the reader. I have a lot to say about the headphone jack and how it enabled the Square reader and how it went away. I think people who listen to the Vergecast know I have a lot of feelings about headphone jacks on cell phones. But we’ve all seen Square, we’ve all seen the hardware certainly. What do you think of Square as a company, as a product?
A lot of people have seen us in coffee shops or farmer’s markets. The interesting part about Square is that we serve sellers that are very small — farmer’s markets — all the way up through Chase Stadium and everything in between. We’ve been hard at work building this comprehensive commerce platform that helps sellers of all types run their business better with Square. In the very beginning, the innovation was on your mobile device; you could quickly take payment and our focus has really been to help sellers accept any sale. That’s how we got started and how we ended up here today.
All right, I’m going to do it. I’m going to talk about headphone jack. I promised myself I would not get totally sidetracked with what enabled the first Square reader, but it is interesting to me. Accepting any payment in a world where most payments are going digital or at least passing through a credit card and lots of people don’t even have wallets or cash anymore puts you in direct competition with cell phones in a very real way.
The reason I bring up the headphone jack is that first Square reader was just a card swipe and it plugged into the headphone jack and it passed the magnetic data from the stripe on the card through the headphone jack to the Square app and suddenly the phones could read credit cards, and that was an open interconnect, maybe the last great open interconnect on cell phones.
The cell phone companies have a lot of thoughts about me talking about it this way, but it was the last great open interconnect and then obviously that went away and you moved to a Bluetooth reader and now you have dedicated hardware. But on iOS, Android, you can just beep a credit card to it, you can beep a phone to it. There are other kinds of payment transaction services that are available in just apps. Maybe you don’t need an open interconnect anymore. How do you think about Square in that ecosystem?
Yeah, it’s a good question. We’ve always wanted to help sellers get started easily. You can download the app, sign up, and get going within seconds, and as the payment landscape has evolved and the hardware interconnect landscape has evolved, we’ve supported both the Lightning plug or Lightning jack for Apple and then now USB-C, and our hardware and software ecosystem is really meant to support both.
When you want a really pro device that’s all-in-one, running our own Android and our own version of Square point-of-sale, all the way through integrated, beautifully designed hardware, all the way through just bringing your own hardware — this is your cell phone and you can plug it in either to Bluetooth or one of our jack-based card readers. And now tap-to-pay is really blowing up.
So if you get started you can just click a button, get registered with Apple or Google and easily just, boop, take a payment. We’re seeing a huge amount of adoption, particularly in the smaller sellers that value portability and value being able to take payments easily. As far as an open ecosystem, ecosystems evolve and I think it’s important to realize that and be able to lead through those changes.
The reason I started with open ecosystems is that easy initiation with Square, where I’m a merchant, I download the app, and someone can just tap their phone onto my phone and pay through Square.
That is very powerful. That is the selling point for a lot of small merchants. It’s just that easy and it’s the hardware they already have, and somewhere in the middle of that is whether or not Apple wants you to do that in a way that they could not really stop you from doing it with the iPhone jack. There’s been a lot of controversy around what Apple will and will not let you do with the NFC reader on the phone, and with the other parts of the phone. Has that come up with Square?
Not as much as you’d think. It’s in Apple’s best interest, I believe, to provide APIs and accessibility to let sellers use products on their devices. As far as there’s a lot of news around a lot of other things that they’re doing with the App Store, on the business side of software and the business side of payments, we haven’t really felt a lot of those constraints. Apple’s a fantastic partner of ours as we continue to innovate.
Yeah, that was a good, managed, PR response to Apple being a good partner. I love it. One of the reasons I’m starting here on the ground and in the weeds — because I want to get to the big parts of Square, and how Square deals with the death of the penny — is the move from “This is the thing that can accept payments” to “We are running your business” to “There’s Square AI in your product roadmap and it will custom develop applications for you as you ask about your data,” and that is the future of all businesses and now we’re running stadiums.
That all starts with “we can accept the payments as easily as possible.” How do you think about that transition and that journey for somebody who signs up just to take payments at a farmer’s market?
When you talk to sellers and you listen to the challenges that they have, running a small business is one of the hardest things ever. These are the true definition of entrepreneurs. You’re putting your livelihood on the line many times to start something that you’re passionate about and they reach for software and technology to help them run their business.
So in the beginning it’s just an idea, “I want to sell something,” and you get started small and you start taking payments. Very quickly you struggle with, okay, I need to manage my inventory better or I need to hire my first staff member and manage schedules and manage payouts and all those other good things. And the idea with Square and the ecosystem that we’ve built is that every step along that journey should feel simple.
As you keep getting larger and larger, we want to constrain the complexity so that it still stays easy to use our software. That’s an incredibly hard product and design challenge, but that’s where we’ve built our ecosystem organically rather than via acquisitions and other strategies that companies grow, because then we can maintain a bar for how sellers use these products, how they discover them, and I think that’s really important to be able to grow with the seller.
One thing we say at Square is that a seller should never outgrow Square, and that means we need to meet them where they are with all these different capabilities. You’re right, it starts with just “payments have to be good.” If payments aren’t good, fast, et cetera, then nothing else matters. But as you get larger and larger, most of the problems a seller faces are related to staffing. “Can I get folks to show up on time? How do I manage my staff schedules so that staff are happy? How can I manage my vendors and all the other costs of goods sold?”
As we know, costs are going up across almost every dimension in the industry. So the problems keep happening and I truly believe that you want democratize technology, you want to take technology — and this is really relevant in the AI space — you want to take technology that’s usually locked to just the very large corporations, the Wall Streets, et cetera, and bring them down to Main Street, and I think if you do that successively and make that technology accessible, you can actually help society. You can make things better, you can get neighborhoods vibrant again, you can help these sellers.
Because a lot of these small businesses fail, unfortunately. It’s hard to run a business, and if you can apply technology effectively, you can run a better business, and you can revitalize a lot of the Main Streets across the world.
That’s a big mission. It’s interesting. That implies that you’re going to go from taking payments to helping more efficiently run businesses or helping small business owners more effectively run their businesses. I’ve heard variations of this pitch from… I wouldn’t even call them competitors, but from other companies that have an entry point into small business. Intuit was on the show and its CEO was like, “We’re going to run your whole back office.”
Squarespace has been on the show and the slide for Squarespace goes from, “We set up your website” to “Your website has a booking calendar on it” to “Now we manage your bookings, we should manage your payments, we’re going to do all of your invoicing.” It is very easy for companies like that. It’s very compelling. It’s like, “Oh, yeah. I started a website. Now that you run my front end, you should run my backend, too.”
I get the pitch and it all starts with “We’re going to do one thing and we’ll expand to everything else.” Is that how you think of Square in that same ecosystem, that your competitors or other providers who start with one thing and are trying to expand the entire back office?
I would think in broad strokes, yes. I think the unique part about Square is that because we build and design all of our own hardware from the chip up, we can provide, I would say, better end-to-end experiences than competitors that may be starting in other areas. At the end of the day, for the majority of brick-and-mortar businesses, the point of sale is the artery, so to speak, of quick-serve restaurants, of retail, and even a lot of services that need this help to make a sale.
It’s important to think that if you do the core things that a business really needs done super well, then it gives you the ability and the believability that you can build additional products. And then one of the unique things about Square and then Block is that we’ve also been able to build other business units, like Cash App, and other complete ecosystems, which is pretty interesting.
Do you think of Toast as your competitor? Toast makes hardware, it’s in point-of-sale in restaurants. Is that a big competitor? Is it other hardware companies? How does that work for you?
It’s a good question. While I think about competitors, I don’t obsess over competitors. I think it’s better to obsess over customers. But one of the unique things about Square is that we can effectively serve almost any type of seller — so retail, food and bev, health and beauty services, et cetera.
The reason why that’s really interesting is that when you think about our competitive set, yes, for food and bev at the large end of the market and quick-serve and full-serve restaurants, it’s Toast, Qu, and a couple other folks. In retail, it’s Lightspeed and Shopify. Health and beauty, there are a bunch of scheduling products out there. Services, mainly Intuit, QuickBooks Service, Titan, et cetera.
I just come back to what a lot of these sellers need: a rock-solid platform and software to help them run their business. As you build some of these capabilities, a lot of the needs across these industries rhyme and I think that’s where you can leverage platforms, and our ecosystem especially, on “Help me understand how my business is running, help me grow my business, and help me manage my staff.” There are a lot of very core similar needs so that we can then build capabilities that we can apply across industries.
I’m fascinated by the fact that you said you think the hardware is a big differentiation. You design it from the chips up. That’s a big investment. The cliche is hardware is hard, that it has to run all the time. If the point-of-sale system goes down, that’s the end of that — your sellers are probably moving on. How do you think about managing the hardware roadmap next to the software roadmap next to, “Hey, maybe the future of all these capabilities are on an AI roadmap somewhere else”?
I’d love to talk about AI in a little bit. We have an incredible hardware team. Thomas Templeton leads our hardware team and it’s, I think, one of the best in the industry. We partner very closely on making sure that we can form fit seller needs, the form factors that they need to run their business with then the software workflows that work best. It’s a partnership, and the faster that we can iterate and deliver both software and hardware experiences together, the better off all the sellers will be.
I’ll make the comparison to Apple or even to Google. Every year Apple shows off a new iOS and then a few months later it shows off the phone and it sort of embodies whatever new features of iOS happen. Google does the same with Android and the Pixel. Are you on that kind of roadmap where you’re like, “I’ve got a bunch of software ideas and the next version of the point-of-sale terminal will really bring them to life,” or are they separate in your mind?
A lot of the reasons why different hardware companies have different strategies is largely how much the underlying hardware exposes APIs and enables the software behaviors. Ultimately for Square, we can decouple those to a larger degree where we don’t need monolithic hardware-software release cycles, where we can be more nimble on the software side. And then yes, there’s step function improvements and things that get unlocked from a hardware capability side, but it’s not as monolithic as some of the other hardware companies out there.
Yeah, what I’m really pushing on is the idea that it’s the hardware that is the big differentiator here. What’s the evidence for that? Is the customer saying, “Boy, we love these Square terminals”? Is it just that it’s easier and more convenient? Is it that they’re more reliable? What are the factors that make it the big differentiator?
Yeah, I would say this hardware and software working together is a differentiator, but specifically from a hardware perspective… I mean, a point-of-sale is front and center. Every single customer, at least in brick-and-mortar businesses where you’re selling something, sees that hardware and interacts with it both from a consumer perspective as well as a staff and employee perspective. So if we can build workflows that are delightful and simple to use on both sides of the counter, that makes the vibe of the business marginally better.
If we can help with that and we can have customers walking away like, “Oh, that was fast, that was easy, that was delightful,” and all the software and tools around that as well as staff members [saying] ““That’s fast, it’s easy to use, it’s helping customers, making sure technology doesn’t get in the way.” At the broadest strokes, we hope that most of our hardware and technology fades into the background. We want our sellers’ brands to stand front and center. There are a lot of design decisions that we’ve made to make sure that that stays true for a seller.
Do you ever do the Steve Jobs thing where you pick up the pictures of the Blackberry and the Palm Trio and you’re like, “This is why they suck and this is why ours is better”?
[Laughs] No, I haven’t. I think it’d be funny to do, but no, we haven’t.
That’s a big vision: “We’re going to enable sellers of all kinds at every scale, we’re going to have hardware that is a better literal customer experience.” And then I was looking at your public roadmap and it is extraordinarily tactical in places. One of the items is there’s going to be support for combo meals at quick-service restaurants. Where do you think that balance is? That Square has to do the work to support combo meals, not that the restaurant has to decide it wants combo meals and it can make Square do that?
The genesis for the public roadmap, and it’s pretty unique for larger businesses in the space, is that we wanted to build out in the open, we wanted to drive accountability for what we say we’re going to do, and we’re going to do it with speed, velocity, and quality. As far as the granularity and the details of what we show in the roadmap, it’s really a judgment call. My general preference is to show more and make it more raw and utilitarian rather than maybe more polished.
That’s why you see everything, even tiny little features, but also really big things will show up there, specifically for combos and some of those kinds of capabilities. When you run a business, that software has to work for how you want to run it and that’s how you merchandise your sales, how you present those kinds of goods or food or services that you’re creating and present to customers. That’s why it’s important, and sellers tell us this. It’s important to be able to help sellers model their business the right way.
This is going to be a very esoteric idea, but in my mind, all software is a combination of metaphors and interfaces and templates. There are just patterns that repeat and then the patterns match the metaphors. A combo meal fits into that framework really well. You open a restaurant, you download Square, and it says, “Enter your menu,” and then you get to some part of the menu and it asks, “What are your combo meals?” and maybe you’ll just enter them.
Is that the kind of thing you want to do for people? To templatize a restaurant so that they have combo menus? Because the other way of doing it is I run the restaurant, I invent the combo menu, and I just put it into the menu with a price and that’s fine, no metaphor required. The combo meal was made up of these three separate items.
What you need for any software — and I generally agree a lot of software is forms and lists of things, which is interesting as a reflection — but you need the core foundational primitives and constructs. For example, in the hierarchy that those things, how they’re designed and engineered in the backend is important to get right and [it’s] important to be flexible. You can’t offer infinite flexibility.
It’s important for any software product to provide guardrails to a certain degree and take opinions, because if you don’t, it could just end up in a super complicated cockpit of knobs and levers and forms and buttons that are really hard to use and understand.
Yeah, it’s funny, I use the cockpit of knobs and levers all the time, but the thing I was thinking about is, “Okay, I run my little team and if this was a submarine, my managing editor needs to sit at a console with a lot of buttons, and how can I put the buttons in front of her so that she can do her job most effectively?”
She does not like this metaphor and I don’t think it actually works, and I think I’d be bad at driving a submarine because I don’t think that’s how a submarine works either, but that’s what you’re describing — “How do I put the various tools of the business in front of the person who operates the business?” What’s the dynamic between Square saying, “These are the tools we think you need that will be most effective,” and the customer saying, “These are the tools we actually want?”
Yeah, that’s the art of building great products. It’s one of the reasons why Square and Block as a company is a very design-led organization. I think design in its simplest form is helping understand the core needs and designing solutions to help solve those needs.
Time and time again, I’m always surprised and delighted by the creativity that comes out of a team and org working closely with engineering to craft these experiences. If you have too much of a top-down view of exactly what you want to build, that limits the creative process. In almost every time, in my experience, it also limits the end result of what you’re trying to build. It’s an interesting thing to reflect on how different companies have taken different strategies, like very top-down, requirements-based approaches versus more, I would say, bottom-up, creative approaches, and the different results that happen because of it.
Do you find that, depending on how big your customers are, they have different needs in terms of how to even describe their problems? I imagine the person who downloads a Square app does not have a bunch of out-of-the-box ideas about how to design a payments workflow. And then I imagine the stadiums you work with have literal data analysts being like, “This is exactly what we need to maximize our revenue.” How do you manage that with the divergent set of needs or even communication styles?
This is one of the benefits for Square’s open platform. At the end of the day, if you’re a very small seller, you kind of just want to adopt all the features and products within Square, mainly marketing, loyalty, everything really. Then as you start to get larger and larger, you want to pull in tools that you want to use, so for whatever reason it’s a best-in-breed barbershop loyalty program or whatever. And our open platform and open partnership strategy really helps sellers pick and choose the parts of Square that help them run their business best. And there’s a continuum.
We want to enable our app marketplace and our developers’ partners to be able to build on top of Square and extend capabilities that work well for them, and then that gives both sellers a choice. We never want to hold data hostage, we never want to do those things, because ultimately we want to help sellers run the best business, and if that means using a third party for some part of that, then we need to enable that.
I want to come back to AI in that part of the conversation as well too because it feels like the future of a lot of third-party integrations is letting AI agents run between these databases, but I think it’s actually a good time for the Decoder questions, just to take a step back.
I have a sense of what’s going on. Square is part of a company called Block; Jack Dorsey is the CEO of Block, which has a wide array of interests. There’s the buy-now-pay-later platform Afterpay, there’s music streaming service Tidal, there’s all kinds of stuff. Where does Square fit into Block and what’s your relationship with Jack Dorsey as the CEO of Block as it relates back to Square?
Block is the kind of company that has a bunch of brands in it. I’m responsible for the Square brand from a product perspective. We obviously have sibling teams, like our sister teams in Cash App, et cetera. We functionalized the entire company late last year, and it was a big change.
Right before, we had a divisional GM [general manager] business unit kind of organization and we went to a purely functional [structure], so there’s engineering, product, design at the highest levels, and then within that are brand-focused teams that think about the different areas. Jack is an incredible leader and spends time with each of our brands, both guiding the strategy but also giving feedback on fit form and the actual products and the Pixels that we’re building. It’s been great to work together with him a lot closer over the past year and couple months after we’ve turned into a more functional org.
So Block has a functional structure? There’s one engineering team for Square, Cash App, and Tidal?
Afterpay, et cetera. There’s a single engineering org.
How do you fight for resources with the music streaming services?
Basically my thesis for teams and prioritization is, one, when we functionalized the company, we created a single set of priorities — so a single shared roadmap. If you have a single shared roadmap with clear DRIs for each of the areas, it makes a lot of these decisions a lot clearer. Because you can be like, “This is more important than this,” and then we can staff appropriately. I think that’s one.
Two, though, it’s really important to make sure that the teams and the squads are stable at the ground level that they’re building. They have autonomy, mastery, purpose, and they can go build the things for customers. You also need to synthesize stability at the squad team level with top-down strategic prioritization. That’s kind of happened.
That’s been the biggest change and benefit from going from a divisional business unit structure to a functional structure, is that alignment is super, super important in a large organization. If you create effectively different mini-startups or product areas, which happens in a divisional org, then yes, you can kind of go fast in each of those directions, but you’re usually going fast in 14 different directions versus going fast and making tradeoffs and being able to build together and build something really incredible.
All right, I’m just going to push on this a little bit. I think I understand what you’re saying, but just assume I’m much dumber than maybe people think I am. DRI is a directly responsible individual. You said you had squads. I understand why you would want shared engineering resources for Square and Cash App and maybe even Afterpay. These are payment rails. Why does Tidal need strategy alignment with Square?
That’s a good question. Not everything has to harmonize with exactly the same strategy or exactly the same things that we’re working on. It’s more that when you functionalize a company, you can also make sure that each discipline is operating at the highest levels, both from a performance perspective but also accountability perspective.
I think that’s one of the key points. It’s not like we’re trading off people necessarily between Tidal and Square all the time. It’s Tidal as a team, Square as a team, Cash App as a team, et cetera, and at the margins we might make prioritization trade-offs, but in the broad strokes, we have our strategies and we go execute on them.
This is the argument that I’ve heard for functional organizations, that for companies that have lots of disparate businesses, it’s actually very hard to manage engineers, and actually designers all want to work with other fancy designers and the reason you would functionalize them is to hold all the engineers to the same high standard, and engineering management at a smallish company is too hard to break up. You can’t hire enough good engineers, enough good engineering. Is that the thesis here?
I think it’s less at that level. There are a few components to it. Maybe I’ll just take a step back about where Square has gone over the past 10 years and my journey at Square a little bit. So Square in the beginning, maybe Square 1.0, this was a little white reader we talked about earlier. This is solely focused on growing, helping sellers make any sale.
And then we went to, let’s say, Square 2.0. This is where Cash App grew and scaled as a separate business unit. Within Square there was a [general manager] divisional structure where we really exploded our software ecosystem to help sellers with a bunch of different jobs. There was a GM for the staff tools and there was a GM for the order system, et cetera.
As you know, companies go between divisional versus functional structures all the time for various reasons, and I don’t think there’s a single right one for every company. It depends on the circumstance and the culture and the market and a bunch of other reasons—
This is the underlying thesis of the show. This is why I get to ask the question every time because everyone was like, “I don’t know, I picked one.”
Yeah, totally. They oscillate based on different things. The reason that leads us to Square 3.0 is that there was a realization that the benefit of Block as a company and the benefit of the two great products that we’ve built with Cash App and Square is that their strength really is the connections between the business units. It was just very difficult to do that in a divisional model. The incentives are just completely misaligned.
Divisional models work super well if you have completely separate P&Ls [profit-and-loss statements] and the products rarely have to talk to each other and you can hire and build whatever cultures you want within each of those areas. It doesn’t work super well when you want one plus one to equal three, let’s just say.
That’s the realization for Square 3.0. We wanted to, one, build with velocity across the business units, and then two, really build interesting products at the intersection between all those. That’s the journey I believe Square has gone on over the past couple years and one of the reasons why it is very invigorating to work here. I’ve been at Square for a little over 11 years now, and I think it’s the ability to move fast, make quick decisions, and build these products at the intersection of both the business units that makes it just incredibly exciting and fun.
How many employees does Block have now and how many people at Square?
I’ll talk in generalities; we don’t share them publicly, but Block has about 12,000 and we are capped. We’re not growing past 12,000, and then Square’s I believe roughly 4,000 or so.
And then inside of that, how is Square structured? I know the whole company is functional, but you have a product team. How is Square generally structured?
The biggest challenge with Square, like I mentioned earlier, is that we serve every segment. We serve food, retail, health and beauty services, and a long tail of other types of sellers. The core decision that we had to make when we restructured around a functional versus GM structure was whether we organize by customer or organize by functional surface.
Do you need a retail team that wakes up every single day living and breathing the retail customers and the retail workflows? Or do you just have more generic pillars like orders and payments and stuff like that? Ultimately, the decision we came to was that we need both. We have key product areas that are focused on food and bev, retail, and local services, and then we have a whole set of, I would say, foundational product capabilities that span all the different verticals and audiences. They work together to prioritize and build the best experiences.
Give me an example of those capabilities.
Banking is one. We have a financial suite team. The team’s job is to think about the cost of goods sold — so things like reporting, accounting, et cetera. Then we have a growth team. Every team needs a growth org that helps with onboarding, conversion, et cetera. We have an identities-and-trust org, so helping sellers take payments safely and effectively is super important.
We have a foundations team. One of the things you need in, I would say, most orgs, either functional or divisional, is a team that’s helpful for ensuring that the general experience — whether that be navigation, design systems, things like that — are cohesive and good. It also acts as a check and a balance to a lot of the product teams that are moving very fast to build capabilities across their areas. And then there are a couple other product areas as well. Customers, for example; staff and payroll. We can keep going on. The ecosystem is broad.
Put that into context in terms of “We’re going to roll out combo meals for quick service restaurants.” Is it that your growth team heard from a bunch of quick-service restaurants that entering combo meals was too arduous and you need a template for it? Was it the team that actually services restaurants that says, “This is the capability we think they need?” Where does that idea come from, how do you implement it, and how do all those teams work together to do it?
It really starts with the strategy and the customer problem. We strategically say that QSRs [quick-service restaurants] and food and bev are a priority for Square, which they are. They’re the core cornerstone to the neighborhood. We start with the customer and the strategy and say, “Okay, what are the key things that they need to do?” Then basically that’s the top-down, this is where we’re going, and then each of the orgs and product areas come up with the key needs.
The combo-specific example, that would be the food and bev team thinking, “Okay, this is really, really important for these types of sellers,” and then working closely with the catalog team and a couple other teams to say, “Okay, we’ve got to prioritize this.” In any sizable org, you always have to make sure dependencies are mapped and different teams can work through the things they need to deliver.
Yeah, this brings me directly to the other question I ask everybody on Decoder, which is about prioritizing the roadmap. That falls on the head of product, that’s your decision. How do you make decisions? What’s your framework?
I’m excited for this question. I think about this in two ways: tactical and strategic decisions. At any org of any size, speed and making sure that the machine doesn’t get jammed up, so to speak, are critical. Part of our decision-making process for the org is that you want to make sure teams, the squads and the teams at the ground level, have rituals so that they can make decisions quickly.
At the team level we do pit stops, which are multiple times per week. They can get unblocked. And then at the leadership level we have what we call unblock meetings, where anyone can sign up and bring a decision to cross-functional leadership to weigh in on a path. We aim to have decisions made within 48 hours or less. It’s very difficult to get to that timeline all the time, but that’s just a stake in the sand that we’ve put.
So that’s the tactical side. On the strategic side, the thing that I’ve been reflecting on is that you need the terroir of decision-making. You need the right ingredients as a leader to be able to make decisions, and it’s steeped in a few things. First is deep customer empathy. It’s hard for me to believe that you can make great decisions if you don’t really understand who you’re building for and why you’re building for them. Second is you have to understand the market, you have to understand the history of the market, the environment that you’re competing and building products in. And the last one is just constraints. You need to understand effectively, intuitively almost, what the different trade-offs and constraints are.
I joke with my team that your gut is the best-trained machine-learning model on the planet because it’s just the synthesization of all the things that you’ve seen over your entire career day to day, in and out. But I think if you have a lot of those components, then you can be “good” at making strategic decisions. Then if you think about actually making them, the biggest thing is, should you make it or should someone on your team make it? If it should be you, great. Then you need to decide how much information you need to make said decision, and in that case it’s usually starting from the customer and working backwards. But that’s a couple of the different ways that I think about making decisions.
Oh, one last note. Decisions are important, but what’s even more important is alignment of the entire org. This is why startups can move fast, because alignment is simple. You’re literally sitting with 20 of your coworkers and you can make a decision and just run immediately. Now, if you think about a remote company or a larger company, alignment becomes more and more difficult as you add more people into the equation.
It’s important that when decisions are made that… A lot of people talk about “disagree and commit” or whatever, great. It’s really just another way to say, “Can we align on this thing and can we go run as fast as possible?” A lot of my job and a lot of my team’s job is to make sure that we have that alignment. At Block and Square, specifically the singular roadmap and top of priorities help with that, and then within that, the DRI model can help make sure those decisions are cascaded down, owned, and driven towards completion.
Nothing makes me happier than when I can tell that someone is a Decoder listener and they came fired up for the Decoder questions. If you do a show like this, you’re like, “I did it.” That’s great. I can tell you’ve been ready.
There’s that, there’s the best laid plans, there’s systems and frameworks. It sounds like you have a lot of authority and you push down a lot of that authority. Then there’s the CEO and you have a famous CEO who at least in public plays a very eccentric character. Is that a stabilizing force? Is that a destabilizing force? What is it like to work for Jack Dorsey?
I think it’s one of the reasons why I’m still here 11 years in. Jack is an incredible visionary leader. I sometimes joke that he lives in the future. He just has a great intuitive sense about what’s important for customers and where markets and things are going.
Jack and his direct reports are really responsible for setting the overarching strategy of “These are the big opportunities where we want to go,” and I think we’ve made some good decisions and we’ve made some bad decisions at the corporate level. With the new focus of these functionalized teams, we’re making better decisions every single day, again, from that alignment, and actually forms for Jack to weigh in and provide guidance and input on things that teams might be struggling with.
Is he the same character in your meetings as he is on stage at various conferences?
He’s very similar, yeah. Definitely.
One of the ideas, besides the beard, which I think is him trying to live in the future as much as possible, like seer from the year 3000, is that he’s big on Bitcoin. Square accepts Bitcoin, Square holds a lot of Bitcoin, and he’s big on AI obviously. There are some AI integrations that I absolutely want to talk about with you. Where is he pushing you the hardest?
He’s pushing the hardest on AI. Automation is one of our top priorities as a company, and it’s not just automating products for sellers, it’s also automating internal processes. That’s everything from support sales all the way through compliance risk, et cetera, and there’s just this fundamental belief that if there is a process, we should try as hard as possible to automate it, and I think it’s exactly the right push just because of how much and how fast this technology is changing and improving.
There hasn’t been a technology in, let’s say, the past 20 years that improves by an order of magnitude every 12 months. There just really hasn’t been something like that in a long period of time, and we can debate on a lot of the impacts of that, but you have to skate to where the puck is going. If I take a step back, remember when mobile was a big thing and the iPhone came out?
[Laughs] Yes, I remember the genesis of my career.
A lot of companies at that stage had a little mobile team off to the side building a little app, and that was the strategy for a lot of companies. A lot of those companies didn’t make it.
The companies that made it were the ones that reimagined how they should work with this new way of reaching customers in this new technology. Uber reinvented transportation in a mobile device. Square I would argue reinvented payments on a mobile device.
We could debate if AI is another one of those kinds of seismic shifts. I believe it is, and I think companies that embrace AI and really ingrain it with both the products that they’re building, and the internal processes they have, are the ones that are going to make it, and the ones that don’t might not.
All right, let’s debate it. I’m ready. I’ve been talking about this a lot. The idea that a business is just a collection of databases and that you should connect those databases and run your business more efficiently, that’s the metaphor. I see it everywhere now. Everyone that I talk to about AI is like, “Finally, I can send my agent off into some database that previously had some weird API and it was very brittle, and I can set up the MCP server and it can finally tell me all the information that I’ve always needed to know to sell widgets more effectively.”
I say that not with derision, I just say it because I hear it so much. But this is kind of what Square’s AI roadmap looks like, right? You’re going to have a bunch of information in your business and you as a small business owner are never going to set up all the tools to get value from that data. So you’ll just ask the AI. The AI will build the tools and then you’ll have it.
That is a very big vision of the future. I see that. I see why you’re describing it as a big vision of the future. That is not a big user interface shift. That is a software architecture shift that now the databases can just talk to each other like old friends or whatever they’re going to do, but it’s not “We need phones to make Uber because the internet is with you all the time.”
How do you bridge the gap? Because that is the thing that I see most of all. We started off by talking about how Square’s differentiation is hardware and now we have to re-architect Square for the AI era, and I can’t quite connect those two things together.
Just quickly, I don’t think Square’s only differentiator is hardware. It’s one of the key things. But I want to come back to the customer. For these small businesses, whether you’re running one location or 10 locations, making good decisions is actually one of the hardest things about running a small business. You just don’t have a ton of time.
You’re sweating making sure that you’ve got your food delivery, the ingredients showing up that day, your staff showing up that day, that you’re doing enough on marketing to make sure that you can generate interest for your business. At the same time, you have access to data, but it’s very difficult to investigate and interrogate that data in a meaningful way for your specific business. So if you just think about the customer, it’s really the first time that the interface for accessing decisions and data is not some complicated UI with a bunch of drop-downs and buttons and charts that — even if you build them beautifully — are just hard to use. Getting access to data across all the different dimensions of your business, sales customers, et cetera.
The intuitive thing about, “Okay, I want to look at my sales on a rainy day versus not a rainy day over the past summer,” or, “How are my tips different on rainy days versus not?” is that these are pretty complex questions. For the first time, agentic systems give sellers a natural language way to answer these questions about their business, letting their curiosity run free.
That’s super unique in a set of technologies. You’re talking about, “Hey, the interface isn’t so different.” But I would argue that the interface is completely different now for two reasons. One, almost every interface in, let’s say, the Web 2.0 world was synchronous. You had to sit there and you had to push a button to get an action. Now for the first time you can have asynchronous action. It can do something on your behalf and it’ll check in with you for approvals and to make sure it can go do that.
That is combined with the fact that it’s more natural, with both voice and text. Voice is really good and honestly, I know Siri has had some struggles, but voice is getting really, really good as an agentic interface. When you combine both of those and you combine just that these sellers are busy and now they can meaningfully know how their business is working and make better decisions, what I think happens is that we see better business outcomes for these small businesses.
Can I sit in the rainy day tips example for one second to make this real? It strikes me that even knowing to ask if I get better or worse tips on rainy days requires a level of insight. Why do you think your sellers are going to ask questions like that?
There are two things that we’re thinking about. One is that sellers intuitively know their business far better than we do. They wake up and they’re in it every single day. They know the things that are keeping them up at night. “Should I stay open an extra 30 minutes and what does that do to my tips on Thursdays?” Or whatever it is. I don’t want to discount how well they know their businesses.
There are very specific things they want to find out to help them run a better business. The interesting next step is how we can be more proactive and suggest things that they may not be thinking about. When you combine these large language models with tool use, you can actually get very grounded correct things to help them understand what’s happening, and that’s a really key investment area for both us and most of the industry.
I’m going to force you to sit in that example, and I know you were just off the cuff, but I want to sit next to it. It is useful for me to think about real examples. Does tip revenue increase on rainy days? I’m going to ask Square’s AI that question. Does Square have a database of historical weather patterns? Do you have to go out and get it?
Just foundationally, that’s the first task. We need to know what the weather was like on Sundays. We need to understand whether it was raining or not, maybe how much it was raining, we need to synthesize that with a database of sales that has both the core revenue and then the tip revenue in it, and then we need to plot that out.
I have asked various LLMs to do tasks like this, and sometimes it works. It works to a degree that is high enough that I understand why the entire industry is betting on it and sometimes it just crashes out, man. It just doesn’t work at all. First tell me, do you have a database of weather that you’ve been tracking against sellers? This is just the first problem. And then how do you make sure it works every time?
This is part of the art of building products in today’s age. There are a few things. When ChatGPT first came out and a lot of these large language models arrived, there was a lot of hallucination. You would ask questions and the answers would just be blatantly untrue. The real innovation in my opinion is, first, they’ve gotten better at pre-training and better data inputs, but the second thing is that we’ve basically taught these large language models to do better agent loops. And this is one of the reasons why Gemini and ChatGPT are much better now — they integrate search data. They can enrich their context with actual data from tools.
If you think about this specific example on rainy day tips, what we do is we have an MCP [model context protocol] ecosystem that we connect our AI tools to, and there’s a weather MCP effectively and there’s a data warehouse that these language models actually write real SQL against. It’s as deterministic as possible. Then basically you could reason through it, get the rainy days — those are actual days — I’m going to use the days as the “where” clause or whatever in the SQL statement and then pull back the data and render it appropriately to the seller, to the customer.
This hybridization of “We’re going to make it write real SQL code, so it’s actually doing deterministic calculations” — this feels like the turn that the whole industry is making. We’re not just going to let the models run off and do their thing, we’re going to use the models to accept natural language input and then we’re going to make sure we do real computers in the back.
That turn feels very big. It feels like not just we’re figuring out how to use the models better, but maybe we’re acknowledging that the models can actually do the work that everyone thinks they can do. Do you see that as a limitation or do you see that as an extension of capability?
I think it’s an extension and just a realization of how to use them. For many years, I don’t think we totally knew how to apply these things to real problems. There was this very cool technology and there was this capability overhang that we just didn’t know how to apply this perfectly. Tool use and context window expansion have to be done together because you have to put all this data somewhere.
Those are really the two big unlocks, and then the ability for these agents to work over a longer time horizon. So 30 seconds, one minute, two minutes, five minutes. It’d be totally reasonable for proactive insights for an agent to work for a couple of minutes or more, collecting data, testing it, making sure it’s right before it builds a comprehensive report, kind of like [OpenAI’s] deep research and others. Those two innovations have really unlocked, I would think, far more tangible business value than the earlier, chat-based conversational interactions that were kind of the very beginning of this whole era.
I guess I’m focusing on you saying “as deterministic as possible,” because the idea that you definitely need a deterministic traditional computer in the background of the interface seems important to me. Is that as important to you as it feels to me?
My mental model now is more that these large language models are becoming more like the CPU of a lot of these different products that we’re building, and you need to give the CPU different things. You need to give it access to data, you need to teach it how to use sandboxes and different tools so that it can effectively learn and get stuff done, you need some working memory, you need all these constructs. Even if you ask someone to go do something, they sometimes do the job wrong some percentage of time. Humans using computers are also not deterministic, even if clicking the button is 100 percent deterministic.
This is where there are really interesting UI challenges and things that we’re working on and going to be releasing in our spring release next year around how we can have these large language models prompt the user with actual generative interface. I’ll give you an example. Say I want to bulk update all the prices of my menu. Let’s say increase prices by 5 percent. In one world you would just have the agent go do that and commit those changes and those would go live on your point of sale and your buyer surfaces. That’s pretty risky in my opinion.
Another path is that you could say, “Okay, I’m going to generate a form and a list of all the things I want to change,” and then the user, the customer, needs to go click “Okay,” boop, and then that actually goes and commits those changes. I think that’s the next frontier after we’ve done a lot of work as an industry on this agentic tool use. It’s about how we have more natural interactions in a generative UI construct to help sellers use these tools.
The realization behind all of that was that if you look at most UI, it’s really forms and lists of things. So it’s pretty straightforward to have these large language models generate those UIs. For the longest time, the builders of tools, like the software teams, were specifying and designing the user interfaces. For the first time ever, the interface is generated by the customer or the user. That’s a pretty profound way to think about what the future over the next couple of years is going to be with how these products are going to be built.
That’s the tension that I find the most fascinating in all of this. I am a proponent that no UI should be hidden. I grew up in the Windows and mice era. We just put all the UI in front of your face, we moved from the command line to “here are all of the buttons,” and then over time we have abstracted everything away and now the kids don’t know how to use file systems.
There’s a real push and pull there that has some real consequences. The iPad kids do not know what files and folders are. Now we’re going to abstract more of the UI away, but you can just ask the magic box to do anything and it will at least tell you that it’s doing it. How do you see that transition? Because there’s a part of me that says people won’t actually know the full capabilities of the tools you’re giving them.
It’s a very interesting philosophical discussion. Back in the day, I used computers with file systems and dragging too. And as we’ve built more software on top of those things, we’ve been able to do more with the tools, and honestly… I don’t know if Gmail has files and folders behind it. I don’t really care. I can do way more with Gmail than I can do with all those things in the past.
Part of me thinks that AI tools need to build trust with the customers and people using it. That’s a bedrock belief that we have when building these tools, and that means there’s some sort of explainability about what’s happening and our strategy is to be able to tell sellers what’s going to happen and let them take some control over what’s going to happen. As far as how this is going to impact my kids or the next generation of people using computers, I’m not totally sure we’ve figured that one out yet.
There’s just a part of me that says if you don’t show people what they can do, they won’t even try, and then what scrambles that with AI is you can ask it to do anything and it will just confidently say it’s happening, whether or not it actually is happening. There’s some new dynamic that needs to be invented there.
I want to just talk about databases for one more second and how AI might interface with them, and then I want to talk about just money because it is a weird time for money and I want to make sure we talk about that a little bit.
I’ve been talking a lot about what I call the DoorDash problem, which is you turn every business into a database and then the AIs can just come and use that database and services like Uber or DoorDash might lose the value add. They might all just become wholesalers or commodity providers of sandwiches. Small businesses have that same problem in different ways.
Square’s sitting on top of business or running a business. So having a bunch of AI features that can talk to a bunch of other AI agents that might run around and do things, creates that problem at scale for a bunch of small businesses in a neighborhood. Burrito delivery might change and maybe we actually bring the power back to the quick-service restaurant in the neighborhood to just be the provider of burritos without a DoorDash sitting on top of them and extracting fees. Is that on your mind as a potential solution or change that AI might bring about?
I think everyone is going to want to expose capabilities to these agents and let those agents do certain things on their behalf. At the end of the day, we want to make sure that sellers and customers can build some sort of relationship. I firmly believe that the Main Streets and the neighborhoods of the world are better because of interactions that you have at those small businesses.
DoorDash is notorious for trying to separate the customer from the seller and have the DoorDash tax, as you call it. What we want to do is just support end-to-end seller interactions, and if we can have agents enable that and make that better, then that’s definitely something we’re interested in.
As you’re saying this, Square has done a big integration with Uber Eats, DoorDash, and Grubhub because that is the audience funnel. That’s where the customers come from for a lot of these restaurants. Is your intention to enable more direct customer relationships?
At the end of the day, we have to help sellers sell where they want to sell. We can’t control that, and we shouldn’t control that. We should enable that in whatever ways. The goal for us is to provide enough tools to enable sellers to create direct connections with customers. It’s one of the reasons why we launched this new Neighborhoods product built into Cash App, because we want sellers to maintain and enhance the interactions that these sellers have with customers mainly around loyalty programs and re-engagement. I think the more that we can do that, the better it is for the seller and also, over time, the better it is for customers.
The weirdest way to think about databases, in my opinion, is to think of money as one giant weird database. The banks are all just databases with fancy faces on top of them. Credit card companies are just databases. Jack Dorsey is big into crypto, which is a decentralized database of money. Bitcoin is just an openly readable database that you can transact on.
How do you think about that interaction? Block has big investments in crypto and Bitcoin, and those numbers go up and down. That might change the company’s balance sheet day-to-day, and then obviously you can transact in alternative currencies. How do you think about that relationship? It is like the last database to conquer?
The whole space is fascinating, and just from a Square perspective, we always want to help sellers accept any sort of payment and just make the sale. That’s how we were founded and that’s one of our core principles. Sellers can’t control how customers want to pay, which is why we truly believe helping sellers accept any type of payment is super important. Back to my point about democratizing technology, there’s a lot of talk around Wall Street having Bitcoin treasuries and all the other kinds of tools that they have. I truly believe providing those as options to small businesses and guardrails and guidance and education is important.
We should give sellers choice. I believe that they’re far more intelligent and can make those decisions for themselves, and if you see about all the investments and all the tools that we’re building within the Bitcoin space, it’s really that it’s helping sellers accept Bitcoin and they can settle in fiat. They don’t have to take Bitcoin. That’s fine. Bitcoin as a payment rail is super interesting. Our goal is to make Bitcoin everyday money. That doesn’t mean you have to hold Bitcoin. You can choose to hold fiat if you’d like to, and again, giving sellers that choice I believe is super important.
Is it fast enough to transact in Bitcoin today to be usable on a payment terminal?
The Bitcoin main net, the Bitcoin network, isn’t fast enough to transact. However, Lightning is fast enough, and Block has made, I would say, monumental investments in the Lightning network. We’re one of the largest Lightning nodes, and if you combine those capabilities, you get all the benefits of the Bitcoin network with the speed that Lightning provides.
This is a very philosophical crypto question, but here we are. The argument against Lightning is that it centralizes Bitcoin in specific ways. People actually hold Bitcoin, and then now you’ve just built a database again that some centralized providers hold and you’ve reduced the actual decentralized nature of Bitcoin. Do you buy into that?
I think it’s important to have technology be as usable as possible.
[Laughs] That is a very polite way of answering that question.
It’s important that we can enable those technologies. As far as the specifics of Lightning versus Bitcoin and all the trade-offs, I think that’s for other people to talk through.
When I think about crypto generally and why you might transact in it, why you might not transact in it, the idea that “number go up” is just the first thing that comes to my mind, and actually today you and I are talking and Bitcoin is down substantially, so maybe the answer to this is it feels different today than on every other day.
But if I had Bitcoin, I can’t think of a good reason to spend it. To me, the real problem is that the culture of Bitcoin is that you should just hold it as long as possible because “number go up.” Are you seeing significant transaction volume in Bitcoin on your network?
Not significant, no, but what we want to see over time is giving both sellers and customers an option. At the end of the day, there are not a lot of options for digital currencies or digital ways to do transactions, let’s just say, that aren’t Visa and MasterCard. You have ACH, which is kind of slow, you have other payment mechanisms and payment rails, so we just believe that it’s important to offer options.
Again, sellers and customers will choose technologies that are both more cost-effective and better user experiences, and as long as we keep investing in that, I believe good things will happen over time.
Are you seeing increased Bitcoin transactions or is it just flat?
I can’t comment specifically on the number of transactions. We’re seeing a lot of adoption and interest in just the Bitcoin product portfolio. So conversions are taking off a ton, which is basically taking a percentage of your sales and converting those to Bitcoin, and there’s a ton of excitement from a lot of sellers and a lot of customers on our very recent Bitcoin payment rail launch a couple of weeks ago.
This is what I mean. I can’t think of a reason to sell Bitcoin. By the way, I’m not allowed to hold Bitcoin. It’s just like holding stocks. It’s a thing we can affect the price of. But I understand why you would hoard Bitcoin, why a seller would be like, “Oh, I can just convert my revenue to Bitcoin. This is great. It’s very easy.” Can you think of a reason why you should use a Bitcoin if you’re holding one?
I mean–
That isn’t crimes. I’ll set aside crimes, but can you think of a reason?
Ultimately, people need to make decisions for themselves. I want to build the technology so that people can make those decisions and they don’t have gatekeepers that are making decisions for them. As long as I can do that, I think it’s up to those people to make the best decisions for themselves, whether that be to sell, transact, et cetera, and that’s how I think about both prioritizing products, what we build, and how we build it.
I’m just going to say, I’ve been asking for years why anyone would spend Bitcoin and no one has ever given me a reason. We’ve talked about the future of money. Let me just drag us into the past of money to close this out.
Sure.
You run point-of-sale terminals, you run hardware, you have to accept cash. The United States government just decided to round up all cash to five cents. We’re just getting rid of pennies. How has that affected your business? Did you know that was coming? Do you have to recode the point-of-sale terminals to be done with pennies?
First, many countries, like Canada and Australia, already do this. We’re live in Canada and Australia, and so we’ve had to build the features and capabilities to let sellers do roundups and all that other stuff.
We’re excited to just enable that for US sellers. Now, there’s always a transition period where it’s a little bumpy to start and then it gets better over time, and from just a pure point of sale helping sellers run a business, we’re ready and set for them to make this transition.
Did you know it was coming, or did you just wake up one day and say, “Oh, do the Canada thing?”
We got a little bit of heads up, but not a ton, no.
Does it feel like a lot of work or does it feel like it will change transaction volumes or average pricing? How’s that going to work?
I would guess that there would be very nominal changes in this. If you think about it, a lot of companies already do roundups to help charities and other local initiatives, so I believe that the elimination of the penny and rounding up a little bit isn’t going to meaningfully change the behaviors we see just in the industry. Time will only tell, though.
Do you know how many pennies point-of-sale terminals handle today?
That’s a fascinating question. I don’t know offhand, but you know what? I’m going to go check it out right after this now.
Yeah, I’m dying to know. Willem, you’ve given us so much time. Come back to us with the penny answer, we’ll put it in. I’ll mention it at the top. But this has been great. You’ve got to come back. I know you’ve got some AI features and new products to launch, so come back on and let us know all about them.
Cool. Yeah, thanks so much for having me. This has been a great convo. I appreciate it.
Questions or comments about this episode? Hit us up at [email protected]. We really do read every email!