Owen Jennings on Block's AI-Driven Restructuring

care of scaling, monitoring, and security. And let me also tell you about Gusto, the unified platform for payroll, benefits, and HR built to evolve with modern small and mediumsiz businesses. And without further ado, we have Owen Jennings from Block. He's the executive officer and business lead. Owen, how are you doing?

Hey, how are you? Thanks for thanks for having me, guys. Big fan of the show.

Appreciate that. Thank you so much. Great to have you on busyus. You guys know each other, but for those who don't, can you kick off with a little bit of an introduction on yourself, how you wound up at Block, how long you've been there, what you do on a day-to-day basis.

Sure. Yeah, I joined Block uh about 12 years ago. I was on the the Square

overnight success.

That's awesome. of the business first and uh I uh I worked on uh scaling the business team um before the the IPO and then in about like 2016 I shifted over to to Cash App. Cash App was like 25 people at the time. it was just peer-to-peer. Um, spent uh spent a while on the on the cash side and then we functionalized the company about two years ago. And so I look after the business team which is product and and a number of our of our operations teams for uh for block overall across our brands which are Square, Cash App, Afterpay, Title, uh so on and so forth.

What what does functionalizing the company mean for those who might not know? So we used to have pretty rigid uh business units where Square and Cash App operated pretty separately and they each had the CEO and and there was kind of duplicative disciplines uh between them. I think part of our core strategy now is connecting the seller side and the consumer side. So having like both sides of the of the transaction.

Um and so about 18 20 months ago we we functionalized quote unquote which just means we're operating at the block level. So the the product teams across all brands uh for all of block roll up into me and same for engineering and design. So it just lets us have centers of excellence um or uh for each discipline and be able to more flexibly move resources uh uh throughout the company versus like really rigid boundaries.

Before we get into the present, tell us about the chaos of that 2016 era scaling cash app. That was like truly insane. hyperrowth from everything that I've heard and I know a lot of a lot of teams are going through that style of hyperrowth right now in consumer AI.

It was it was incredible. Um I mean I look back on it very fondly. I think the the key thing if you go back far enough to like 2015 2016 there was actually um a meaningful question of if we should continue investing in Cash App. Like there were executives at the at the company. At the time it was Square Cache, but there were executives at the company who were saying, "Look, you know, Venmo already has this thing on lock. This doesn't make any sense. You haven't figured out a path to to monetization." Um, and so it was actually Jack uh uh our CEO um who continued to push and and continued to invest. Sarah Frier, our old CFO, actually gave us a um a deadline at which we had to figure out monetization or else we were going to be shut down. and and we were trying to go through the IPO and so I kind of I kind of get the the trade-offs there. Um but then like very very quickly found found market fit. The the key difference for Cash App back then was the way Cash App initially started was you could move money instantly from bank account to bank account. Um that that wasn't something that was available with with any other sort sort of peer-to-peer functionality. And so we found market fit really really quickly and the uh yeah tremendous tremendous growth and like incredibly small team. Um, and just uh, you know, working to make sure we could we could stay up and and we could serve customers every day, especially Friday.

Jacobe Jacobe was on was on.

No way. No way. So So that was fundamentally different because I had I had other uh money transfer apps at the time, but the flow was always take some money out of your bank account, put it into the app, and then send that money from that app to my friend who it sits in their account, and then they can take that to their bank account. You could go bank to bank basically. So we actually it's actually really funny how we how we got here. So we were we were thinking about building a how we build a lowcost payment network for buyers and there was this one type of transaction that was pretty unique and it it was a it was a unlin refund. So for a square merchant it's like how do you refund money back onto the card and then we were like oh we can just process these as unlin refunds. So we worked with Visa we ended up turning into this whole Visa direct thing and now so much is happening on debit. But what we would do is there actually wasn't even an application to start. You would um you would CC cash at squareup.com and uh if your if your debit cards were linked on either side, you could you would move money instantly from Bank of America to JP Morgan Chase or what have you on debit rails and so it was instant and it was and it was free. Things have evolved a lot uh since.

Yeah. Yeah. Of course. Uh well, I mean with that functionalization of the company, is that one lens to view the change in workforce size as just a continuation of that or is this some sort of key strategic pivot? Like there were a lot of different messages going out. I'm wondering how you're thinking about the size of the company, the workforce over the next few years.

Yeah, I can I can um kind of paint the picture on how we how we got here. I mean the primary driver for this decision was around um how AI tools are evolving and how they're flowing through our work. So when I think about uh software development over the past year or so I would say there was pretty meaningful progress from

let's say you know the start of 2025 through about November and it was flowing through our work and kind of changing how we were approaching things. We we've been pretty early here like we we launched Goose a couple of years ago which is our open source agent harness. I think it was the first Asian harness out there uh or or that I know of. We worked on the MCP um with uh with Anthropic and Open AI. So anyway, things were progressing and then in like the last week of November, first week of December, things just fundamentally changed and it and it was with Opus 46 and it was with codeex 53 and we basically crossed the chasm where like I I think the way AI tools were flowing through on the development side before is like it's a useful tool to help a given engineer be be more productive. you can autocomplete, so on and so forth.

In like late November with the with the model changes, we got to the point where um these agentic systems were actually able to write uh the code autonomously and the code was good enough to ship into prod and that that was like a huge change. And then so so we spent you know I think many of us probably spent December and the holidays like playing with the uh playing with the tools playing with cloud code playing with playing with goose under and then and then we sp we spent Q1 thinking okay well now how does this flow through to a to a technology company and a software development company because everything has fundamentally changed over the past four months. Um and so the org changes were were mostly a reflection of of that. I I think that um just the fundamental nature of like the shape of the organization that you need, the size of the organization that you need, the workflow for an engineer, a designer, a product manager, other disciplines, all fundamentally changed. So, you know, we we we've continued to to kind of refine the the size and shape of our org, but but this was definitely a reaction to uh agentic development and what that means for for technology companies. Uh how have expectations around just productivity for developers at Square changed? Like how do you how do you try to understand if somebody is actually using the tools to the full ex fullest extent? Uh like what's the general kind of philosophy?

I think ba I think I think basically uh expectations around productivity are increasing across the board. Uh I think it's it's the clearest on the development side and that's also why like when you when you look at the the reduction in force it actually overindexed to product and design and engineering relative to to other other roles. Um, and so we track uh all sorts of the normal things that you would track like PR PR throughput and we have like a internal concept of like time to time to customer value and all of those thing like like time to customer value is um is compressing um the number of PRs that that we're submitting are is increasing massively. PRs per engineer the main bottleneck now is is code review because uh shipping a PR is is becoming um you know more and more trivial especially for for simpler features. But I I guess I would say we're we're kind of in the in the in the belly of the beast right now. Uh I think a lot of software development companies are where I think expect as high as expectations are, I don't think that they're high enough. like we're we're seeing just a fundamental shift here where the I mean the look the ba the basic way that development used to work is that you would have uh a product manager, a designer, six to eight server engineers, two to four client engineers, and you would kind of sequentially work on a feature in kind of a waterfall fashion. A medium-sized feature would take, you know, four, six weeks. You would ship that, move on to the next one. That's it's a fundamentally different now. Now we have really small squads of two people, three people who are way more full stack. They're all on the tools as we say as we say internally. And uh what you're able to do is is just like orders of magnitude different. Um obviously there's like the okay are you running five or 10 instances of cloud code at once and how does that flow through to to your day-to-day? I I think the bigger changes are just like the fully agentic systems. So we we have um builderbot internally which is somewhat similar to cloud code. It's built on top of goose and so now

anyone I love all your code names. We got goose and we got builderbot. What else you got? What else you got?

Well I think goose was named goose too because uh uh because of the Yeah, exactly. Top Gun. Um little cheeky uh co-pilot shout out there. the the

so for builderbot um there's a we have a slack integration so I can just at builderbot and say like hey fix the spacing on this screen in cash app and then like you know put my computer down go have a a slice of pizza with my kids and then come back and there's a beautiful PR that's that's uh that's waiting um and that's just like a complete paradigm change and it's happening across across everything not just the the development side

I don't know how much you've been tracking like the AWS story but it feels Feels like Amazon might have had some uh you know uh uh uptime problems or some backlash to the amount of vibe code that they were pushing to production. How are you thinking about avoiding the downside scenario where more and more of the system becomes vibe coded. Fewer and fewer people understand everything and something goes down and no one really knows how to fix it. Uptime starts dropping. There's a variety of reasons why we're seeing uptimes fall off. Some of it is just CPU shortage. But uh how are you thinking about like systems reliability in this new paradigm?

I think there's a couple there's a couple of different pieces. So one for like the for the reduction in force itself the number one principle was stability and reliability. So we like blank sheet of paper. We were like okay that's a P 0. We're not going to sacrifice that and let's like build bu start building back the team with that in mind. Um, I think more generally how how we think about this is that there's just a spectrum of like where you can be more risk-seeking and where you want to be more risk averse. Uh, and so there's a bunch of examples right now where um, you know, a product manager on the Bitcoin team at Block has built a a fully functional app um, uh, related to to Bitcoin and stable coins as kind of a proof of concept prototype um, and has done that like really without without an IDE, without looking at code. Um, and that sort of thing can be incredibly helpful. Um similarly like for an internal tool you're probably way more risk-seeking in terms of okay let's uh you know uh we don't need to you don't need to ask for permission for everything and you can push this change and then obviously like the further down the stack and the more loadbearing something is we're way more riskaverse and I think that makes sense like our our financial platform team for instance they run you know reconciliation and treasury and card issuance and uh you know our cloud platform team I I don't It's not the case that we're that we're being um you know willy-nilly in terms of of how we're approaching it. It's just that like those who are experts and like our system architects and our principal engineers themselves who have deep context are able to become you know 2x 3x 5x more more productive but um uh you know we we also have humans reviewing every every bit of code before we ship it into prod. So I think you need to be prudent in your approach here and protect the downside. But but still like the productivity gains are are pretty obvious I think.

Where where else are you guys getting a lot of value out of AI outside of codegen? Is anything on the fraud side or increasing uh you know the efficiency of of lending? I'm just kind of spitballing. But uh where are you most excited?

It's pretty broad. It's pretty broad base. And so I'm not going to walk through the the entirety of the of the org, but I'll give some some good examples. I think you know uh customer support, customer service is a is a clear one where we have our own uh we have our own models that that we've built and we've trained and we've been able to automate like 75 to 80% of of chat inquiries ac across brands and that's without sacrificing uh customer satisfaction or or seesat. I think that's like a you know that's been that's been out there and I'm sure y'all are seeing what Sierra and Decagon and and folks are are doing. I think basically any deterministic task were able to offer a ton of leverage to the folks who are working at the company. So a lot of folks on the on the operational side are they're working cues. Um so you might have a cue to say like um hey does this person should this person pass identity verification um or should we file a certain um ticket related to related to this transaction um or hey anomaly detection let's look at this. um we have a human in the loop which I think is critical especially in a in a highly regulated uh area like uh like financial services but a lot of the pre-work and a lot of the context uh and a lot of the pattern matching we're able to do with with tooling and so it ends up that uh a given individual on that team can be massively massively more more productive. Um I I think the other thing is just we've built generalized tools that are that are quite helpful across the the org. So another code name for you is uh G2 which is I guess it's goose 2. I don't know it's goose's brother goose.

This is like a you can think of it as like an agentic operating system that's rolled out to the entire company and then we have MCPs into all of the tools that we use. So into into Snowflake, into um into Tableau, into Gmail, Calendar, Docs, like pick pick your favorite piece of software and then it's all one interface and then I can go in and create agents or create automations that do whatever I want. So um instead of spending an hour every morning looking through every single dashboard, seeing what's going on with with Square and Cash App and Afterpay, um I can just have a an agent that runs at night, does that, and then lets me know if anything is off. And you can kind of extrapolate from there all of the the different use cases where we're basically allowing everyone at the company to build automations on top of all of the different sources of of truth that we that we have internally. And then we're doing the same thing for our customers on the on the Square side and the and the Cash App side as well.

Very cool. Well, congratulations on the progress and thank you for coming to explain uh some tumultuous times, but uh it sounds like uh there there are green pastures ahead and lots of new things to build and we're excited to follow along. So, thanks for coming and breaking it down.

Yeah, great day.

Have a good rest of your day. We'll talk to you soon.

Thank you.

Goodbye.

Let me tell you about App 11. Profitable