Square Global Head of Product on How to Build AI Agents People Actually Use | Willem Avé

You're reinventing the company to be a mini AGI. Part of the vision was that basically everybody is reporting to the same person like a super flat organization. So how realistic do you think that is? >> You have to kind of step back and understand why [music] orgs exist in the first place. >> William of a global head of product at Square. One of the unique things about your product is that you have a hardware component as well as software. So I'm curious to know how you integrate those two worlds as part of the same function. >> Hardware must be approachable, must be easy to use, it must be completely delightful. and doing so is [music] not easy. >> How do you make sure that someone who is not super tech-savvy is able to benefit from some of this new technology that you are building? >> Running a business is difficult kind of loneliness that you don't always know [music] where to go or what decisions to make. >> Square starts a payments company and now it seems like it's an entire ecosystem. How you thinking about the larger TAM? >> Well, I actually have a hot take on TAM. My hot take on TAM is that >> Hey, this is Carlos, CEO at Product School and your host on the product podcast. My guest today is William Ave, global head of product at Square. He landed there in 2014 when Square acquired Bookfresh, the scheduling startup he co-founded and ran as CTO. So, he has watched this thing go from a little white card reader to multiple ecosystems inside Block and his customers are not on X all day. They run bakeries. Here's what we'll cover. Why they tore up the business unit model and went fully functional. The DRRi model built to kill the silent veto. What happens when AI encodes the knowledge your org chart used to hold? The chatbot era is over and what replaces it. His hot take that TAM is almost infinite. Let's get into it. Welcome to the product podcast, William. >> Thank you. Thank you for having me. >> William, so you are the global head of product at Square. Square is part of a holding company and maybe we can start there. You can explain a little bit more about how the the companies or the teams are designed. >> Yeah, definitely. So, I've been at Square/Block for many years. I've seen really every evolution of the company from the little white reader all the way through to the I think sophisticated uh multiple ecosystems we have. It's I think unique to be at a company that has not only been able to innovate on one product but then be able to create multiple ecosystems of products throughout its journey. And then maybe I'll just like start kind of like my my journey here started you know my company was acquired I was CTO and I I kind of grew up through the engineering org here. We had a general management model for many years. So this was kind of business unit focused and then about two years ago we reorged the entire company to be more focused on our customers and functionalizing uh everything and one of the key reasons we did that was we believe that functional excellence i.e. excellence within your domain engineering product design etc um was really important and that some of the other org models that we had previously reduced kind of the craft and excellence that we wanted to see within each of our functions. So yeah, the past two or so years we've been in a fully functionalized world where there kind of topline orgs, product design, engineering, etc. And then within that we have kind of brand-based kind of focus areas, Square, Cash App, etc. And I think like from an org design perspective, uh there's many ways we could we could take the the the convo, but I think one of the key things is you want to you want to kind of focus orgs around customer outcomes as much as possible. And the farther away you kind of kind of go from that that principle, I think I think the worse outcomes you get. >> Yeah, I love this topic around uh or design and I've seen so many different variations. Many companies start as functional and then at some point they grow and decide to go their business unit model and then at some point they go back to functional or some sort of hybrid in between. >> Yeah. >> So in your case today uh what is part of your function? >> Yeah. Great. So, it's the Square brand and I lead product for the Square brand. And the way that we've decided to organize the Square org is kind of just leaning into two beliefs that I have for for for building products. One is you want teams as close to customer outcomes as possible, right? So, they can really understand the needs of the customers and be able to build products to kind of serve those needs. Um, and that's kind of led us to a couple broad orgs. First is an audiences org. So, these are all of our verticals. think kind of services, retail, uh, food and bev etc. and then a kind of core platform org. But the core platform org importantly also has product surfaces that they're responsible for. I think it's very easy for platform teams to fall into a kind of one-sizefits-one when they're building software. Kind of thinking about a very generic kind of solution that doesn't solve any need. So I think it's important to kind of marry product and platform together in certain areas. And then then have kind of a a growth team, right? So growth all the whole acquisition flywheel um and then a money team. So money for us is everything from bill pay, payroll, banking, etc. And and then obviously we have kind of our hardware teams which is a sister or to to uh to us. >> Oh, I was literally going to ask about that, right? Because one of the unique things about your product is that you have a hardware component as well as software. So I'm curious to know how you integrate those two worlds as part of the same function. Yeah. So hardware I think hardware is something very special and we have I think one of the best hardware teams on the planet. And what I mean by something special is that hardware shows up in business's real worlds, right? It's it's something you put on the counter. It's something you carry with you during your shift. It's something that you use, right? So hardware must be approachable, must be easy to use. It must be completely delightful. And doing so is not easy. And I think the the magic really happens when you integrate hardware and software together in a way that you can kind of build delightful experiences. Obviously, everyone's almost everyone's used an iPhone. I think they they've kind of set the bar for a lot of the kind of software hardware integration. But I think the the the key to that is making sure that when you're building new hardware products, you're kind of marrying the software form factor with the hardware form factor. You can't think of them as completely separate, which is why I think it's difficult for companies to use kind of off-the-shelf component commodity hardware and build truly excellent experiences. >> One of the risks I see with some works that have both hardware and software is that they're just so different that it's really hard to find a leader that can understand both worlds in depth. Similar to what you were saying about general managers, sometimes it's just hard to find someone who really gets business and product, right? So in your case, I'm curious to know like how you actually find the the right level of support to ensure that both hardware and software are at the same level of priority. >> Yeah, it's a good question. So I mean the the biggest difference is the timelines are usually just different, right? Building hardware takes longer than building software and there's different steps that you go through while building um hardware and software. I think from making sure you get really great outcomes. I think you first start you need incredible people, right? Two, you need to make sure that the process you go for building doesn't have a ton of ton of burdensome red tape where decisions get mired either in bureaucratic process or slow decision-m because ultimately both building hardware and software is one of iteration, right? You you have you have a vision of what you want to build, but the path there takes a bunch of different kind of turns, right? As you figure out what you know, what defect rates you're targeting, what kind of reliability rates, etc. you know, battery life, you know, there's a ton of ton of dimensions to it. And then I think the best software is also built via kind of an iterative learning loop. So I think a lot of that if you kind of combine both of those together, I think you can have both an org as well as a process that leads to incredible outcomes. >> Block or your your CEO Jack Dorsey has been really a pioneer in terms of how to think about or design. You recently came up with this concept of DRRI. Would love for you to to expand on it. Yeah. Um, so I think it's it's an it's a it's an incredibly interesting concept and I think one that we've leaned into uh very hard and I think it's interesting and and pretty amazing is one the best products you have to kind of have an arc from idea to kind of scaling it, right? And that arc is a crossunctional effort, right? It's not just building an incredible product. It's how do you go to market with it and then how do you keep iterating and and building it. And a lot of times what happens is that products slow down because of decision-making, right? Either it's like silent vetos throughout the org where one team doesn't want to do something that another team wants to do, right? Or it's lack of alignment between what we're doing or where we're going. So the DRI model is really meant to cut through decision-m and effectively empower a person and an individual to see that decision-m through the entire life cycle of building building that that that product or solution. Now it it's it's a it's a unique position right because you kind of have to be able to synthesize product technology and business and then be able to work very well across the entire arc of building that software right everything from kind of ideation I think the best ideation happens in a design and engineering fashion where they can kind of come up and explore the entire space and then all the way through kind of like go to market and scale where you need to be able to tell the world about what you're building and why you're building it. So I think the the DRRi really helps shape both technical strategy but also kind of keep the teams accountable to execution and then break down decision-m so they can cut through any sort of roadblocks or challenges that that uh that the teams may have. >> I was doing my my research on on this model. I I I found a quote that said that uh you're reinventing the company to be a mini AGI and uh part of the vision was that basically everybody is reporting to the same person like a super flat organization. So how realistic do you think that is? >> Yeah, it's a good question. So I think um you have to kind of step back and understand why orgs exist in the first place, right? And at least historically orgs existed in the because of information flow, right? And that there needed to be a way to communicate between different orgs and teams and align them on a on a singular outcome, right? Uh whatever you're building. I think you know modern technology and AI has actually you know uh amplified the communication ability between a lot of the the leaf nodes of a tree so to speak and I think the interesting thing is that whether it's everyone reporting to one person or there's you know for sure kind of evolutions of that I think it's the key goal is to have small autonomous teams that are able to kind of take ownership of an outcome and build that as quickly as possible And typically orgs are very good at like inventing process and inventing red tape because it you know meets the needs of whatever org they've created or whatever goals uh that org has created. So I think you know and and a lot of times those are for good reasons right you have a sev or a problem and then you kind of create some process to stop that from happening again. I think though AI gives us the ability to encode that knowledge at another layer. It doesn't have to be encoded in the org entirely. It can be encoded in data that agents have access to and that really democratizes decision-m it allows teams right small teams I truly believe that the best teams are kind of small fastmoving fully autonomous and enabled teams and it kind of enables that really to happen you you hear a lot around uh right now around like enterprise context graphs and things like that for AI um I think it's really just another way of saying is like you have to encode knowledge in tools that you can have individuals and teams take advantage of that. So they don't you don't have to play like telephone to get an answer, right? You can get an answer from an agent and then make a decision and move forward. Um so I think you know the DRRi model um combined with um kind of I think smaller flatter uh kind of orgs do allow you to move a lot faster that doesn't diminish the need of management right people grow in their careers people have aspirations in their careers you need to um also have incredible managers like people managers right that can help people through their kind of career growth and trajectories um so I think both of those you kind of have to marry together right you need like great managers and then you need great DR eyes and people that can cut through decisions so that you can deliver uh great products. I think this is the time I spend the most discussing or geeking out on or design because probably the best example of someone who is pioneering something like like this and and I love it. I'm also going to try to shift gears into another unique thing that I I think is in embedded into your product which is you're building for the real economy that your users are not always you know like say revops managers it's also like the person that is working at the bakery right so tell me more about how you're thinking about that in terms of bringing AI capabilities to people who are probably not on X every single >> [laughter] >> Yeah. Uh the the the the real the real world is definitely not um on X every day. Um though some are. Um but I think so small business is fascinating because they are like I think the most pure instantiation of an entrepreneur. They literally put their their livelihood on the line for their passion. And that passion could be everything from like candlestick making to bakeries to to really anything. And I think that's like endlessly fascinating and I respect it a lot. I think two is that small small business in my my opinion really forms the the like the artery right or the core of um the economy right I think the main streets of the world are unique they're um really important social areas for for different economies and neighborhoods and I think what um Square and Block has always stood for is economic empowerment and what that means for Square is making sure neighborhoods can can thrive And we kind of say we want Main Street to stand as tall as Wall Street, right? And I think this is like it's really important because um to to do so they they need technology, right? You can't compete today um in the business landscape without you know help you know without automating parts of your business with with software and now increasingly with artificial intelligence. And I I firmly believe that if you take advanced technology and democratize that, right, give that to as many people as possible, like as many small business as possible, that's that's net good for the economy and that's then net good for society, right? So I think those are some of the things that that I spend a lot of time and it's like it's not easy, right? Because each of these businesses are endlessly unique, right? You build you have and they need very specific workflows. Like a bakery is just different than a flower shop and that's different from a burger joint. and like they just run, you know, they have different people, they have different teams, they have different workflows. So, it's all it's all different. Um, and I think that's one of the one of the awesome um things that we've been able to do at at Square. >> Well, let's try to unpack some of those use cases. Uh, because I can imagine that the level of AI adoption is different, right? So how do you make sure that someone who is not super techsavvy is able to benefit from some of this new technology that you are building for them? >> Yeah, so it's a good question. Running a business is is is difficult and it's not difficult for like the most obvious reason. Yes, obviously it's you know you have to hire and kind of do your craft whether that be food or whatever but but it's also it it's one of kind of loneliness that you don't always know where to go or what decisions to make, right? So, for example, like what should you price a menu? What should you price your burger? Sure, you could calculate how much it costs you to make the burger, right? But should you mark it up by 10% or 20%. Right? Um that there's an endless set of decisions, right? Sometimes you hear kind of small business owners, they they start working the second, you know, while they're while their head's still on the pillow in their bed, right? And they keep working like, you know, all the way all the way through through the evening because there's always something there's always some decision that they have to make. And um I think the the advantage or the attraction to AI for business is not is not just a chatbot, right? Um it's really to have this intelligent thought partner that lets that lets businesses make better decisions, right? And then eventually um kind of offload and delegate full tasks to things to kind of uh parts of their business. So, a couple examples could be, you know, inventory intake or kind of um spotting sales trends um for your business and then suggesting a marketing campaign or kind of a number of other um kind of kind of unique workflows. And I think the the interesting part is that I I think like the the days of just a standard question answer chatbot are kind of over. And I I think they're over for for for a few reasons, but first is that like basically we've already solved that problem. And what people are expecting now is to do real work with AI, right? And real work is far more than just like answering questions and getting some answers back in a thread. Real work looks like synthesizing data across all of your data sources. It looks like kind of creating kind of artifacts specifically for your business, right? That could be something like a, you know, par inventory workflow for for the bakery that we talking about. Or it could mean um kind of a like a sales leadership board, right? so that you can train and incentivize your your staff, right? There's a bunch of these different artifacts. And I think that's the that that's the that's the challenge is that it's it's it's very easy to build a chatbot today. I think it's like very hard to build a dependable, reliable agent, if you will, and and that's what we've been working on and really really trying to solve. >> I would love to see it in action. >> Sure. Yeah, we can we can we can take it for a little little spin. So this is managerbot uh kind of a home screen where you can kind of come in and we AI generates a bunch of ideas for for for sellers pretty um kind of either either either daily or just in time. This is like an example of one where like you know add missing item descriptions and like why is this important? Well, if you want to get found online your your online catalog needs to have great descriptions, right? So, ManagerBot can notice these things um and kind of like suggest descriptions and then you can kind of go ahead and review it in chat. I'm not going to do that right right this second, but I'm going to go over to um computer and kind of show what I mean by uh doing real work, right? So, um this is kind of an example of, you know, maybe I want to work on some menu engineering. Um, and rather than just a quick answer, what manager bot gives you is a kind of a full-on artifact, right? It does a bunch of work and it looks at both your data. It can also use a lot of other tools and, you know, it can kind of say like, okay, here are all the different items that have happened. You know, this is, you know, uh, what you should do with each of those. These are the different types of items that are doing well or not not doing super well. And then you can kind of like ask questions about this. Okay, great. um you know what are some suggestions based on this menu do some research and come back with an opinion right and it says like okay this is what I would do I would make [snorts] you know the halfp pound burger a hero not just an item dot dot dot right so I think when when someone is thinking through a problem right they don't just want like okay here is you know here's the answer to this they want a richer artifact that allows them to really dig deep I think similarly if you look at you know like labor forecasts right so if you look at you know this is the labor forecast is is all uh kind of demo demo account. Um you can kind of see this is where the costs are. It can actually generate very rich rich artifacts for you. And I think this this is kind of what I'm talking about like when sellers want to get work done. They need rich sophisticated artifacts, right? They need real things that they can go and build themselves. Yes, that interaction might be, you know, textbased or voice, but you know, it's it's much richer than just a simple answer. Um, and I think this is when, you know, you kind of can see sellers be like, okay, this is really valuable, right? I don't I have to put less work into the system to get some very meaningful outcomes. And then, you know, I kind of mentioned this whole ideas uh area where, you know, ideas are things from like sales or customer, right? Top customer relapse, two of your top customers haven't visited in 30 days. Um, you know, there's like staff insights, inventory insights, etc. So this whole system working together we feel really helps sellers both understand their business but then also be able to take action, improve things and then get real work done on on their computer. >> Thank you for for going for it and showing this and in action because I think that what you mentioned around delegating to AI or getting stuff done is is is critical. Sometimes AI can be more work especially if you have to be switching windows from chatpt to uh PowerPoint presentation to something else. It seems like here you kind of create that environment for them. They don't have to worry about memory MD or skills or connectors or things that maybe sound more common to to the to the AI geeks and so they can just continue using what they understand in a more sophisticated way. >> That's right. That's exactly right. I mean, I think a few things are true. One is I love that you said you don't want sellers or kind of customers to do more work. And I I think it's a very important insight in that my experience and I think a lot of people's experience using AI is that you just get like a like a bunch of word vomit effectively like you get just a bunch of data and you have to like parse that and that's actually work. Sometimes it's very useful, right? If you're doing some research or whatnot, but other times it's like please shorten that for me or just like create me a little website or something, right? Um, and I think the the more that you can steer more that you can steer these experiences to very high value that doesn't cause customers to do a lot of work to use that, right? That builds trust with the system. And then it allows you to delegate and take actions. I didn't I didn't show this in in in the demo, but you can do bulk actions, you can create campaigns, you can do all that other great stuff kind of within within the same experience. But I think one of the interesting things at least for small business is a lot of times if people don't just wake up and like oh I have to change the price of X. Usually you have to do like some investigation right so you want to kind of like get some questions an you know kind of like get some detailed analyses and then you have enough trust in the system to be like okay now I'll go delegate some work to it to go actually execute on inventory counts or price changes or whatnot. So I'm thinking about your overall TAM, right? Because in a way square started as a payments company >> and now it seems like it's an entire ecosystem. So how are you thinking about that and the larger TAM? >> Yeah. Um well I actually have a hot take on TAM. Um and my my hot take on TAM is that I think with the right team with the right idea and the right kind of learning loop your TAM is almost infinite. What I mean by that is that you can you can create you can create markets right with incredible products and I think that that's important for Square in that like we have effectively endless TAM and you know because it's the it's the small medium even enterprise economy right which is trillions of dollars and I think I think that the two important things for building kind of products and software uh in in these like very large TAM markets is one you you need some forcing function to help teams and people make decisions. So we we at Square believe in local economies and neighborhoods and our northstar is to build what we call a neighborhood network and a neighborhood network is uh effectively all the economic action that happens in in a neighborhood right and that's small business owners that's their staff and that's the customers coming into those businesses. The interesting thing around focusing on like local neighborhoods is it is it's actually refocuses the TAM away from some things like internetonly direct to consumer um websites, right? That that's not part of a local economy. Um and what you can actually start to do is look at look at Main Street, look at your local neighborhoods and kind of see what kinds of businesses that exist there, right? So then you have you have kind of a you know an ideal customer, right? these are the sets of customers that I want to go build for. And then you can actually help teams make decisions about, okay, great. We're going to solve those needs and we're going to solve those needs across um kind of commerce tools, right? Both online and offline um financial tools, right? Banking, checking, kind of, you know, credit, all those those tools. And then intelligence tools, right? Because intelligence and autonomy is a core part of building software and helping um local neighborhoods uh be better. Because today I see the current SMB to have a lot of fragmented point solutions. They might have their website builder. They might have their Yelp page or something similar where they can see reviews, right? They have their reservation or point of sales technology and maybe [snorts] some others in between. So he, if I understand correctly, you're talking about like a full-on integration that allows the the small business owner to have access to technology without having to worry too much about like these these smaller pieces of point solutions. >> That's right. So I think the challenge for that is like it's it's uh at least you know it used to be very difficult to be able to build software for each of those quadrants. I think AI software engineering has kind of lowered that bar to a certain degree. I think the second point is that a lot of customers and sellers have software solutions today that they like right so for us to be successful we need both great firstparty tools right that's point of sale kind of staff management tools etc but also have a very open platform to allow all of our incredible partners to build alongside us right and with us so that we can help reduce like you know a bunch of tabs and copying data between this tool and that tool to help you answer a question. And then I think the third is that AI at least you know kind of agentic AI really allows sellers to interact with a bunch of different software products at the same time because a lot of software is is kind of able to be used very well by agentic tools. You see that today with cloud and codeex etc. Um like it can just use these tools exceedingly well. So, it's it's helping solve a lot of this multi, you know, multi multioftware challenges that that customers have. >> Maybe it's because I'm on vacation this week. I'm in Turkey and it's crazy. I I had to make a hair appointment the other day and they sent me to WhatsApp and I had to message the hairdresser literally on WhatsApp and then the other day I wanted to make a dinner reservation and the you know, Open Table and all of those websites are not very popular here. So, I had to message them on Instagram. So I'm also very curious about how you're thinking about the communication tools. >> Yeah. >> Especially for emerging markets. >> Yeah. No, it's it's a great question. Um so so I think a few things like business at its core is a a kind of communication/team sport, right? Like you want to communicate with your customers, you need to communicate with your staff. Today a lot of that communication happens in very segmented siloed tools, right? uh like you just mentioned, that's going to be WhatsApp, Instagram, email, like I got to go um you know, I need a uh a contractor to go fix something and it's usually email, maybe it's text. So, I think it's like and and and when you talk to business owners, they are drowning in these like many different communication tools, right? They're like, "My SMS inbox is like hundreds of unreads." Like, it's it's crazy, right? Just I mean, just your examples. I mean, imagine that business's WhatsApp thing, right? It's right. It's probably crazy. So, I think I think one is no one's really solved this yet. I think just to be honest, um I think two I'm pretty interested in trying to solve it for small businesses. And I think there's there's some very cool collaboration platforms that I think we're building. One is Buzz. Um that I think actually is pretty unique and pretty interesting to think about how how maybe not that exact product but how that model could help sellers right manage both their staff but also manage a lot of the communication inbounds across a lot of these different channels that they get. >> Yeah, we were also very curious to know because that is a very very new product that Block announced and it seems to be going so many different angles, right? Like one is is what you just described. Another one could be potentially replacement of Slack for for different types of businesses. But it seems like in a way you are trying to create limited time. >> Yeah. I think um I mean I'll just go back to like you know every business and every person has a communication challenge, right? Not not like actually talking, right? But actually like managing all of the different channels. And I don't think we have a great solution yet. I think I'm pretty excited about the Buzz technology uh for a lot of for a lot of good reasons. I think um I think it's something that takes we're kind of excited to work on. >> William, it's been a pleasure to have you on the podcast. Thank you so much for your time. >> Yeah, thank you for having me. It's been great.