Deel's COO on the AI Agent Writing Its Own Code

0:00Dan Westgarth: You're listening to a DayOne.fm show. I would say that if you're an inexperienced coder, using Floyd is actually very inefficient and time-consuming. Floyd was burning a ton of tokens, my laptop fans were buzzing, and we were red hot. In addition to this, we've built something called DCA, which stands for Deal Code Agent. And the idea of DCA is you can feed it the ticket, and it will actually analyze, understand the ticket. It will then find the relevant areas of the code base, and it will write code. you will create a PR.

0:30Georgie Healy: Dan, that is your AI hack of the week. I tell you, DCA is so strong. That is incredible. That's awesome.

0:38Dan Westgarth: Navigator finds a bug, passes to the DCA, DCA solves it, and repeat, the platform is self-improving and getting better over time.

0:50Georgie Healy: Hello, welcome back to In the Blink of AI. I'm Georgie Healy, and we have an exciting guest. We have Deal. If you don't know who Deal are, frankly, what are you doing? Go back and listen to our previous episodes because they are our incredible sponsor this season. They have made this show possible. Dan, their COO who joins us today was a really great sport. I absolutely loved it. So, okay, a bit of context about Deal. They are one of Y Combinator's success stories. They were actually there just before Sam Altman left to go all in on OpenAI, fun fact. They're doing over a billion in annualized revenue.

1:26Georgie Healy: And so they've not only navigated AI, they've fully like nailed this time in AI. And Dan as COO helps us peek behind the scenes of their operations. And where you still need human beings. Remember human beings? Yeah. In sales, there is a specific part of the process that AI cannot emulate. And he broke down how Deal actually runs customer support. They have three tiers. It's based on confidence and context. Can't wait for you to hear that. Like every other episode, we spoke about things that are a bit off-piste in the headlines that we just want a really smart person that's in the weeds and delivering day after day to give their insight on.

2:04Georgie Healy: Thank you, Dan. Thank you, Deal. And let's dive in.

2:08Dan Westgarth: You're listening to a dayone.fm show.

2:14Georgie Healy: Founders scale faster on Deal. Set up payroll for any country in minutes, hire anyone anywhere, get visas handled fast, and get back to building. Visit deal.com slash day one. That's D-E-E-L dot com slash day one.

2:33Georgie Healy: Dan Wesgarth, you are the COO of Deal, and I feel like household name amongst the tech industry, the startup industry, but just in case people haven't heard of Deal, Tell us what you guys do, essentially.

2:49Dan Westgarth: Deal is a HR and payroll platform. Our clients are able to hire, manage and pay anyone, anywhere. We started out as a contractor invoicing platform. We now do all different types of HR services, everything from a HR information system through to immigration and mobility work, to IT leasing, to global payroll processing. we aim to meet all of our customers HR and payroll needs.

3:18Georgie Healy: Incredible. Huge platform, huge deal, huge role. Before we dive into that though I'm gonna hit you straight up with some kind of AI hot takes. Are you ready for those?

3:31Dan Westgarth: Sure.

3:33Georgie Healy: What role in companies in general do you think has really fundamentally changed since AI has become you know, proliferated?

3:45Dan Westgarth: I mean, I'm seeing a lot of change in finance, particularly accounts receivable. A deal we do more than a billion dollars in revenue on a annualized recurring basis. And reconciling all of those invoices used to take a lot of time. Clients will send us the wrong amount or in the wrong currency, or they won't put the right invoice reference in the payment field and reconciling them was taking a long time. Since developing our own AI workflow, we're able to do that much, much, much faster. So I think the role of account receivable specialist has changed significantly. And I think that's just kind of a general trend that we're seeing across the finance and operations.

4:31Georgie Healy: For me, this is brilliant news. I did one accounting subject at university. I did a masters in commerce and I was like I'm gonna be across all of this stuff and for whatever reason that topic was so boring I could not get it to retain in my brain like I just could not find myself excited about things like balancing like invoices like I can't even remember what it was called but the like so I see this as great news is there any roles where you're like oh kind of wish AI didn't disrupt that? Or are you a purist in any sense in terms of AI disruption?

5:13Dan Westgarth: I don't think it's role by role. I think that AI is able to help just about every role in our business. I think it's more the application of it. And the majority of use cases are still human in the loop use cases. For example, when people are using AI to draft emails, great for drafting them, But I don't think that AI always has something that's ready to send. Usually you need human in the loop. You need to put your own spin on it, make sure it's in your tone, in your voice. And at the end of the day, that's a communication coming from your address as a human going to another human. It should have human in the loop.

5:47Dan Westgarth: It should have human oversight and a human should be finishing the communication.

5:51Georgie Healy: Human in the loop with AI doing the boring, repeatable stuff I love. Okay, here's a headline for you. There have been a lot of hardware releases recently from big tech companies. They're not letting these AI glasses die, Dan. Like every company seems to have their iteration, Meta, Google, Snap. Can you see yourself purchasing AI hardware in the form of glasses? Is this on your horizon? you know i was with a friend recently um and he was showing me the latest oakley glasses

6:31Dan Westgarth: and yes they were really cool we were uh we were actually wakeboarding and um we were using the glasses to make videos and i thought it was really cool would i be a customer would i buy them probably probably not it's not really for me i'm not really a content creator i don't like to film I'll photograph everything. However, I am really interested in ergonomic working. I really love to lock into deep work, no distractions. I like to be really comfortable. I've spent a lot of time setting up my home office. So I think if glasses are to move closer to tools for work, I would be interested. I was a customer of the Apple Vision Pro.

7:13And while that product isn't doing so well,

7:17Dan Westgarth: I do think there are some really cool learnings that will probably be applied to the glasses market.

7:21Georgie Healy: On that, we have the meta-Oculus, gathering dust at a very tall shelf in our house. Really cool for sports use cases, perhaps. Like my husband said that for a specific cricket shot, and I'm definitely showing our nationality here, it was really hard to practice that shot because if you missed it, you'd basically get hit in the balls or something. I can't remember what the shot was, but it was a really difficult one to practice. But with hardware, you can practice the angles of the movement to get to a good enough point that once a ball is coming at you, that is, you've got the runs on the board.

8:03Georgie Healy: Do you see sports as a use case for AI hardware? Like, I'm not sure if this is outside of wakeboarding.

8:11Dan Westgarth: Yeah, for sure. I mean, there'll be a ton of real world, real life utilizations. So, yeah.

8:18Georgie Healy: Amazing. Not helpful for me who's terrible at sport and has no interest, but hypothetically could be great. Okay, last hot take for you before we dive into the media part of the episode. You're a COO of a Series E company. You're probably quite busy, but have you been following the World Cup? And if so, any hot takes, who's going to win?

8:41Dan Westgarth: Well, yes, of course I'm following the World Cup. I would love for England to win. The time zone is not ideal being based in Europe to watch all of the games, but I would say that the highlights coverage is really good. And I've been watching a few of the games the day after. I would love for England to win, but I think they've got some tough competition.

8:59Georgie Healy: England's doing so well. It's very hard not to get pumped about certain players, right? Yeah. Yeah, been loving it. Okay, thank you for humoring me with the hot takes. Let's dive into what people probably have decided to listen to this episode for. Number one, there are a lot of HR platforms. So why is deals so successful? Why are you guys so good at what you do and have navigated like this AI landscape and have continued to be so successful year after year in one minute?

9:34Dan Westgarth: Yeah, I mean, the common denominator is people. We hire really good people in all of our functions and in all of our locations. But if I tell the story a little bit, we started a deal seven years ago. We made a bet on remote and international working pre-COVID. When the pandemic came along, every company wanted to hire people internationally, remotely, and we were able to support them. We built a lot of momentum, secured a lot of funding, brought a lot of smart, hardworking people together, and then the business has snowballed from that point.

10:04Georgie Healy: I've heard a lot of people say these businesses that built just before COVID or around COVID, where we assumed that we would all stay remote, like haven't been able to navigate past that. You guys obviously have. Why do you think that is?

10:20Dan Westgarth: Well, we quickly diversified. Like I said, at the start of the podcast, our initial product was a contractor invoicing product. So let's say an employee in their home country, wanted to move home to be with family during the pandemic, companies weren't able to pay them because they didn't have a legal entity in that country to run payroll. So we offered them a temporary solution whereby that individual would become a contractor, they would raise invoices for their work and those invoices would be paid. We quickly realized that that model was probably not sustainable for most of our customers and their particular use case.

10:55Dan Westgarth: So we diversified into employer of record, where we established the entity, we employ the person directly. And then again, we realized, once a company has a large number of employees in a country on this model, they're probably going to set up their own entity. So then we went to global payroll. And then after global payroll, we had clients that said, well, actually, can you help us get a laptop in the country? Can you help us do the person's visa? Can you help us provide benefits to the employees? And we decided to do all of that. So I would say that it's customer-led. I think the thing that makes us different from a lot of other companies in the space is the people.

11:28Dan Westgarth: We hire top talent and we're quite aggressive in our way of working.

11:32Georgie Healy: To you, what is great top talent? You know, I recently interviewed Eleven Labs and they basically said, we don't really coach. Like if you don't have that like internal drive and ability to level up, like we can't teach it. What do you look for when you're looking to hire?

11:53Dan Westgarth: Yeah, I don't think the fundamentals can really be taught. I think there are probably three buckets that I look for. I think curiosity and willingness to learn is number one. If someone is curious and constantly learning and constantly trying to figure out things, they're likely to do well. They need to be good working.

12:13Georgie Healy: So if I came in with my I refuse to read income statements, you'd be like, good to know.

12:18Dan Westgarth: yeah yeah i i don't think it's it's i don't think it's about kind of um uh certain topics or refusing to look at certain topics it's more being able to pull yourself into things so i love working with um self-taught people maybe they're self-taught coders or they're self-taught fan professionals or whatever and they're kind of being led into this area. And then they've done the qualifications, they've done the deep training, they've done the education, but to be self-led into it, I think is a big part of it.

12:56Georgie Healy: And in this time where the technology is moving so quickly, you have to continuously have that drive and curiosity, right? To continue to navigate like you guys have. before you joined Deal as COO, you were an early member at Revolut, another household name, a company that I'm sure many of the listeners are familiar with. But what we might not be familiar with is what it takes to leave one rocket ship for another rocket ship. How did you make that decision? What inspired you to do that?

13:33Dan Westgarth: I think it was opportunity, really. And I could be really, I could try and sound really smart, about this and say opportunity cost or a difference in opportunity but really the truth of the matter is I knew the founders of Deel I really liked them I thought they were really smart hard-working people they were curious they were eager to learn eager to build they had the grit and mentality I also got on really well with them they gave me the opportunity to come aboard as a COO a deal sorry at Revolut I was a general manager working on international expansion so So this opportunity was a step up and I really liked the founding team and I really liked what they were building.

14:14Dan Westgarth: It really resonated with me and I could see parallels between other very large companies and what Deal was building.

14:28Georgie Healy: Founders scale faster on Deal. Set up payroll for any country in minutes, hire anyone anywhere and get visas handled fast so you stay focused on scaling. Deal takes care of onboarding, HR, IT, EOR, benefits and compliance so your team can grow without borders. It's why more than 40 ,000 fast-growing companies trust Deal to move fast. Visit deal.com slash day one. That's D-E-E-L dot com slash day one.

15:01Georgie Healy: And just to give a little sneaky behind the scenes about those co-founders of Deal, They were part of the YC Winter 2019 batch just before Sam Altman announced he was going all in on OpenAI. I have to ask you, Dan, did they ever talk to you about that over a beer? Did they meet Sam Altman? Do you use any of the wisdom from that cohort to this day? Like any of that carried through?

15:31Dan Westgarth: I have no idea on the Sam Altman point, but I would say that the cohort companies tend to stick together and the friendship circles are formed and they also tend to stick together. While I wasn't a founder in the batch, I joined shortly after and a lot of those other founders and founding employees, we're all friends.

15:53Georgie Healy: I want to be on that WhatsApp group chat. Okay, let's talk about, you know, hitting your targets for seven years. Apart from people, what do you attribute that to? Like how do you set targets that you can stretch yourselves as a company but also achieve? Like what are the mechanics that go behind that?

16:15Dan Westgarth: I mean at this stage there's a lot of mathematics in it. The targets are not quickly or easily set. There's a lot of thinking and working that goes into those. But I think the important thing is momentum. It could be hard to measure momentum. It could be easy to talk about it and describe it. And I think it's really important. Once you have people working towards targets and hitting those targets, the culture of the company falls into place and you can keep growing and keep compounding.

16:47Georgie Healy: Tell me about your sales team. This is something that I find fascinating. As someone with no sales background, are your sales team using AI in a way that you find is beneficial? And what needs to remain human, do you believe?

17:08Dan Westgarth: I think that AI is helping a lot in understanding the customer, prospecting, building materials, proposals, comparison. slides, things like that. The bit that AI can't touch, which is especially fundamental to our business, given we're in peril in HR, is building relationships, building trust, and going the last mile on understanding the customer. AI can do so much of it. It can provide a great foundation, but you really need to understand what the customer wants. And in our business, it's not just about selling. It's about implementing. Implementation is a really important part of what we do, especially if we talk about enterprise customers, some of those implementations can take years.

17:52Dan Westgarth: For example, I was speaking with a major bank recently, they had over 200 ,000 employees, just imagine how long that implementation will take and how many stakeholders will involve. So, you know, actually, actually selling, selling the deal is one thing, but implementing it is a whole other part that is really important.

18:11Georgie Healy: You made this clear earlier, the customers said, oh, can you do this as well? What about visas? What about this? And I'm sure the sales team want to just be like, yep, sure. And then like hand it off to someone else. Is there any tricks to knowing what they can say yes to? Or how do you message throughout the company? Yes, we want to do what the customer asks, but can we achieve it? Like I'm curious what happens as a overall logistics perspective.

18:42Dan Westgarth: Yeah. Yeah. I mean, it depends on the size of the customer. If it's a smaller customer, then we offer them what's off the shelf and it's prepackaged. It's mostly automated. The customer can set up on their own. If it's a large customer, and I'm talking about big international companies, banks, airlines, etc., then we'll do custom builds. Maybe they want immigration processing done in a way that's not off the shelf. Maybe they want the implementation done in a way that's not off the shelf. Maybe they want to completely white label the platform. Maybe they don't even want to see deals logo on it.

19:18Dan Westgarth: You know, all of this stuff is possible and that's what is required when you're dealing with companies of more than 100 ,000 people.

19:24Georgie Healy: Bringing it back to AI for a second, hot take for you. I've heard both sides for this. Do you let AI do your email triaging, answering customer emails, answering internal emails, or is that the most important touch point, Your emails with your customers should be human-led, should be hands-on. Don't remove that friction.

19:50Dan Westgarth: Yeah, I think the first point that I hope the listeners will like is we don't really have internal email at Yale. We know everything is done in real time on Slack. There's some volume of internal email, but it's really small. When it comes to customer email, it really depends on the topic. So when we're dealing with our customers, especially the medium and large size ones, we're dealing with many different stakeholders. So if the nature of the communication is to do with customer support or operational change, then I think AI can play a really good part, especially if it's a very easy transactional topic.

20:24Dan Westgarth: If it's a deeper change-related topic, then you can lead with AI, but a human tends to be in the loop. So what we found really helpful on our customer support program is we have an AI co-pilot, which alongside the communication, their co-pilot will build a context layer that will help the person respond to the customer. That's really helpful. The other thing that we've done is we've had AI generate a draft response. So that's there in the tool and they can use it if they want. and then for the really transactional basic things where we were pretty certain that we're going to be able to resolve it to a high standard and the customer is going to be happy well actually

21:05Georgie Healy: just respond to the customer directly this brings me to email ai meeting assistance uh not email just ai meeting assistance in general again such a contentious topic at a business at your scale uh pros and cons for using a meeting assistant when would you when wouldn't you Yeah, I think in general, you want to have AI having access to as many touchpoints and as much data as possible because that helps it make smart decisions.

21:42Dan Westgarth: It helps increase context. So a deal, we use Gemini because we're a G Suite customer. So we can have Gemini build a transcript, post a transcript. We can do summaries. We can do action points, all this stuff. But I think for AI to be able to build these self-reinforcing loops, it needs as much context as possible. And meetings are a really good context point.

22:11Georgie Healy: I heard a hot take yesterday, which was you need these meeting assistants to be in all your meetings. You can't just turn them on and off because if you do that, you're adding bias in. And, oh, this is an important meeting, so I'll have the meeting assistant. This is just a quick meeting. I won't have the meeting assistant. Do you agree with that?

22:33Dan Westgarth: Yeah, I do. I don't think it should have access to everything. I think there are certain confidential topics, for example, HR and people topics. You don't really want an assistant there and you don't really want all of that information adding into the internal wiki. However, if it's a product decision, if it's a customer issue, you do want that information transcribing and storing because that's going to help the AI next time it's solving a similar issue.

22:57Georgie Healy: Okay. And last one in that section is what makes an A-plus engineer in 2026, Dan?

23:08Dan Westgarth: Curiosity. Again, they need to be curious. I don't think we're ever going to spoon feed the best people, the best technologies. We have had quite a bit of success with AI and engineering around just over two-thirds of all code that is committed is actually coming from AI. So you can imagine the productivity gains we're having in engineering at the moment. Yeah, incredible.

23:37Georgie Healy: Do you code with AI now, even if you're not a coder?

23:41Dan Westgarth: Yeah, yeah. I use cloud code. I've got access to the different repos. I've made pull requests. there's nothing quite like the feeling of your PR being approved by an engineer and getting a little message being like okay good job that's really fun but this isn't this isn't my core work this is this is these are things I'm doing at weekends and in addition to this we've built something called DCA which stands for deal code agent and the idea of DCA is you can feed it the ticket and it will actually analyze, understand the ticket. It will then find the relevant areas of the code base and it will write code.

24:23It will create a PR and an engineer will

24:27Dan Westgarth: review that PR. And then if it's accepted, it will be pushed. So you can kind of imagine if you find something on the platform, you find something on the internal tools that's wrong or broken, you can write the ticket yourself you can send it to dca and within a couple of hours dca is finished uh writing the code and engineer an engineer will be reviewing it damn that is your ai hack of the

24:51Georgie Healy: week i tell you dca is so strong that is incredible that's awesome i can take it a layer um a layer

24:59Dan Westgarth: further um so imagine the dca process that i've just described on the flip side we have another tool which proactively navigates the application as if it was a user, as if it was a customer. And during those navigation sessions, it will find bugs and it will find issues, and then it will create tickets for those bugs and issues. So then if you stack that above DCA, and you can imagine then you have this self-reinforcing, self-improving process, Navigator finds a bug, passes it to DCA, DCA solves it, and repeat, the platform's self-improving and getting better over time.

25:37Georgie Healy: It's magic. Sometimes AI feels like magic, right?

25:41Dan Westgarth: Yeah, I would say that it's not a perfect process. Do we have any bugs on the platform? Yes, we do. I'm sure there are some somewhere. Is what I just described solving the majority of the easy ones? Yeah, for sure. So it's freeing up our engineers to work on more complex and interesting problems.

25:59Georgie Healy: I had such a chip on my shoulder not picking software engineering as an undergrad. out. I picked chemical engineering, which super not really helpful day to day, but increasingly that chip on my shoulder is getting smaller based on what we're able to do, even as like, I guess, non-technical people, even though I hate that term. Tell me a bit about, you know, you have customers that are trying to break into new markets. You've obviously broken into new markets, where is the sweet spot where they're not scaling too quickly? But also, like, is it true that you'll never be ready? Like, will you ever feel ready to scale into other markets?

26:45Dan Westgarth: Yeah, I'm going to talk about B2B businesses because I think B2C is very different. I think with B2B businesses, adopting a global approach from the get-go can be really beneficial. We did that at DL. So we were able to pay anyone anywhere, but also have clients, customers, businesses from anywhere use DL. And, you know, a lot of naysayers would say, this isn't going to work. How can these companies pay you? How can they pay in dollars from Brazil? You don't have Spanish language. Why are you asking for this social security number or equivalent when someone is from this country? Like there's a lot of different reasons why going global is difficult.

27:31Dan Westgarth: But I found in the B2B space, the really good customers, like the customers you want, the fast growing, the fast moving, they will see the value in your solution and they will figure out a way to navigate it. They'll figure out a way to fund you in dollars. They'll figure out a way to complete the onboarding. And we did that a deal. and while it didn't necessarily contribute to so much of our growth in the early days because most of our growth came from Europe and the US and that's where we were localized, we did see some international customers and it created momentum. I remember our first country manager we hired, her name is Natty in Colombia.

28:06Dan Westgarth: Our first country manager was in Colombia. Can you imagine it's such a strange country to hire the first country manager from? She'd heard about us. Maybe she'd even been paid for a deal and she commented and said, we want to represent you. I want to represent you in Colombia. And we're like, okay, sure, let's do it. And then Colombia started to build momentum. We localized the Colombian onboarding flow and we supported pesos on the platform and Colombia is now flying. So I think going international from the get-go and opening it up does have its merits for a lot of B2B platforms.

28:39Georgie Healy: And B2C, what's the strategy there?

28:42Dan Westgarth: I think B2C is very, very, very different. I think the infrastructure that's required to offer consumer services is generally, it really has a much higher barrier to entry, particularly when I think about businesses and financial services, or I think about businesses in the ride hailing space, livery space, the infrastructure requirement is way higher. And even after you've navigated all of that infrastructure, you then need a really compelling go-to-market motion to make it worth your while. So I think on retail businesses, the country-by-country approach is better. I do think there are some cases where passporting works across regions.

29:25Dan Westgarth: What's passporting?

29:27Georgie Healy: Sorry.

29:27Dan Westgarth: Yeah, well, there's a few different types. Let's talk about passporting within the EU. This is quite easy. One EU country can typically sell into another EU country. They're not required to have another consumer license because the consumer license in their home country can be passported in. Yeah, you've seen some really good cases of this and then some really bad cases of this. I think about the fintech space. I think about number 26, German fintech bank, quite a lot of German customers, but they didn't really figure out passporting into other countries. So in their case, not all of the infrastructure, but a lot of the infrastructure could be passported into different countries.

30:10Dan Westgarth: I'm not sure why that didn't work out. I'm not sure if it was the rest of the infrastructure they needed to make a really good proposition, or it was the go-to-market motion that wasn't ready.

30:20Georgie Healy: I want to know what the horror stories are when you don't use a platform like Deal that have seen everything, They've seen every case. They've seen, you know, what happens when compliance goes wrong or local payroll goes wrong. And they're like, no, we can just do it ourselves. Like, you know, maybe they think, oh, well, with Claude, I can just ask it and maybe I don't need deal. What happens then?

30:47Dan Westgarth: Maybe I can answer in two different perspectives. I think for large companies, they don't do it themselves. What happens is there's a lot of duplication and fragmentation of tools and there's a lot of inefficiencies. So I think if a global business, let's say they employ a few thousand people across multiple different countries, they've probably got different payroll systems and different HR systems in every single country. Then they've hired people to manage those. They're building different currencies, different amounts. Nothing really talks to each other. There's very little interoperability and it just becomes really inefficient and painful.

31:24Dan Westgarth: And when you speak to those businesses and they're trying to increase bottom line margins, it's an area that they really care about because they can see how inefficient it is, how much duplication is happening and how much money they're spending. And DL solves that with one platform for all countries and all different services. if we go to the other end of the spectrum which is really kind of small companies maybe they've got less than 10 employees and um doing it yourself i think from a hr data perspective you can probably manage that pretty well off of a spreadsheet sounds crazy but i truly think you can um when it comes to calculating payroll and uh paying people and filing taxes with the government and filing it with the government and paying social security, child benefit and all of this type of stuff.

32:16Dan Westgarth: In most cases, it's not possible. In some cases, it will be possible. It will become a huge time suck, even for 10 people. There's a lot of fiddly things, government portals that need to be navigated. And it's really not worth the time. You can onboard the deal and for tiny companies, our fees are very low. You can get set up in five minutes and just run payroll immediately.

32:37Georgie Healy: It's like penny wise, pound foolish type stuff. Like, and if you get sued, I guess it's like, you saved not a lot of money for that.

32:48Dan Westgarth: Yeah, it goes deeper than that. I mean, you know, maybe an employee will sue you because you messed up their payroll. But imagine if you have the authorities coming after you and your business is a success, you have the social security administration auditing you for the next five years. And that's not a fun experience. And it just sucks more time.

33:05Georgie Healy: There's some good horror stories, Dan. Thank you. I appreciate that. AI agents. I went on a full rant about AI agents because I think people misuse the term. Like really autonomous, no human in the loop. Like that original definition of an AI agent. Let's call it workflows. Let's call it AI agents. But for you, what keeps you up at night with the idea of AI agents running the business?

33:34Dan Westgarth: Well, Akai keeps me up at night because I'm working 24-7. Akai is our internal tool to manage agents and build workflows. So I think we quickly realized at the start of this year that, yes, you can build powerful agents. You can get the proof of concept pretty quickly. But actually deploying and scaling them is a huge pain. handling things like two-factor authentication, dashboarding, being notified when the agent isn't working. Just things like that were a huge pain. So we decided to invest into an internal tool, Akai. We use it internally. It saved over 200 ,000 hours of manual operations time.

34:16Dan Westgarth: And we're now opening it up for our customers. The interesting stat that I can give you, and Akai is a contributor to this. We have best-in-class revenue growth at Deal. I can't say what the number is, but if you look at our annualized revenue growth, it's really strong. We haven't grown the operations team since March of 2025. So revenue's grown by a large amount, huge amount, but the operations team hasn't grown since March of 2025. So I'm really proud of that. AI is a contributing factor. alongside internal tooling and efficiencies of scale.

34:54Georgie Healy: And so what would you say to people that aren't seeing that as, well, this is incredible for the employees that are there? Like they are obviously, you know, doing incredibly well and the AI is doing the boring stuff. They get to do the more interesting strategic stuff. But to some listeners, they're listening and being like, oh my gosh, what does that mean for the workforce, Stan? Uh, what's your advice if people are like, I want to be one of those employees that is loving life?

35:26Dan Westgarth: Yeah, I think, I think the foundational, uh, roles are always still required. I mean, a deal with just over 7 ,000 people. And I think that number is made up a foundational role. You know, we have a certain number of payroll professionals we need. We have a certain number or finance professionals, you need, et cetera, et cetera. What I would say to those people is the bar is going up. So to be an amazing product manager or an amazing finance person, the bar has gotten higher. And it's emphasis on the curiosity and self-learning and being aware of all of these tools and being actually able to use them and apply them.

36:09Dan Westgarth: That's the reality. I don't think the number of jobs is necessarily going down. It's just that the bar is going up.

36:19Georgie Healy: You said you were coding on the weekend. Clearly, even COOs are navigating being more effective in their role. Did you expect to see yourself coding?

36:32Dan Westgarth: Yeah, I mean, I think at the time that I was exploring that, it was a very hot, trendy topic and i just wanted to to dive in so um yeah i got i got set up and changed some of our internal tooling um i think in the end the engineers rewrote it anyway because what i built was very inefficient but just just being able to feel and understand how it how everything works and how quick and easy it is i wouldn't say it's actually that quick and easy i would say that if you're an inexperienced coder using using claude is actually very inefficient and time consuming i claude was burning a ton of tokens, my laptop, fans were buzzing and we were red hot.

37:12Dan Westgarth: And, you know, I was giving a prompt and I was having to return in five minutes, go downstairs, get a cup of tea, come back upstairs and the laptop's still burning away. So I don't think it's necessarily like super quick or efficient for someone that's not experienced, but it proved to me the capabilities and that certainly had a positive impact on the business.

37:30Georgie Healy: And I'm sure the engineers and you had a mutual respect for each other after that experience.

37:35Dan Westgarth: Absolutely.

37:37Georgie Healy: Okay. So we have made it to the rapid fire at the end of the episode. I want one minute answers, but if you're like, you know what, I've got more juice in the tank. I'll allow a little bit extra. How about that? The sasspocalypse. It is still an ongoing topic. Some people say it's a myth. Some people say it's just starting. What's your take, Dan?

38:01Dan Westgarth: I, again, think it's bar raising. I think that AI makes migrations easier. So customers that were tied to a certain SaaS platform are no longer tied as tightly. And that is going to increase competition. It's also going to create some consolidation. So I don't think a SaaS apocalypse is going to happen. I think if anything, the leading SaaS companies are going to get better. They're going to gain more market share and they can be better off as a result. And so are we as customers.

38:35Georgie Healy: And what would you hope that those companies that are navigating this, as you guys have clearly done over seven years of navigating different market cycles, what's the most important thing for them to be focusing on right now, do you think?

38:48Dan Westgarth: I think customer experience. I think there are so many SaaS platforms which are actually really bad and they can be way better.

38:55Georgie Healy: I don't know about you. I have seen some platforms and I'm like, how do you have even one customer? It's going to be really interesting to see, you know, people will have more autonomy to say no to bad experiences, maybe.

39:10Dan Westgarth: Yeah, yeah, I think so. I think, again, doubling down on that switching point and implementing, switching cost and implementation has been a huge time suck or focus suck on companies and on people. And if it's easier, quicker to switch and reimplement, then it's going to make the market a lot more competitive.

39:33Georgie Healy: We've seen it with LLMs. Like even with all the context that people put into Chack GPT, for example, and years of training the voice and all the rest of it, it's like if they want to switch to Claude, they will put up with those switching costs and they will figure out a way to make it work, right?

39:51Dan Westgarth: Yeah, agreed.

39:52Georgie Healy: Okay, so you guys created a physical deal card, right? Like a physical one. Be honest. Yes. Did you use it? Do you use it? would you ever use it?

40:04Dan Westgarth: Yeah, yeah. I was heavily involved in building it back in the day. It's quite an old product now. It's a few years old. It's designed for people in emerging markets that don't have access to really great fintech solutions like Robinhood or Revolut or whatever else it might be. Yeah, amazing product if you're in an emerging market. I would use it.

40:30Georgie Healy: I want to see what it looks like now. Did you create it like a neobank, like portrait mode? Seeing you were behind the scenes, like I want to hear about how you designed it.

40:42Dan Westgarth: I mean, the initial objective was to give workers in emerging markets quicker access to their money. So if you're living in the Caribbean, receiving your salary can take a long time. and there can be hidden fees. Your bank might actually charge you to get paid, which seems insane, but it's the truth of the matter. And sometimes it can take three to five days for that money to arrive. So we said, okay, how can we get money to people quicker? And this was kind of before the rise of stable coins. There were crypto solutions available, but they weren't stable. They were to pick the volatility. So we said, okay, let's do two things.

41:26Dan Westgarth: Let's build a deal card. And workers can get access to their earnings as soon as they're paid, like the same second. And then they can spend that money. And what we saw was we saw most people getting a portion of their salary paid onto the card and then the rest paid to the back. So then that moved me on to my second solution, which is crypto. And we have stablecoin offering. and then also the kind of traditional big crypto. So you can get paid in Bitcoin, you can get paid in Ethereum or USDC. And now we have both of those solutions on the table. We actually see like multiple things happening.

42:07Dan Westgarth: We kind of have some customers that are Bitcoin believers who get a portion of their paycheck deposited in BTC and assumably they hold on to that. And then some of the money goes on the card, some of the money goes on stablecoin and no money is going through the banking rail. So I think that's the real value that it delivers. Oof, they're living in 2035.

42:30Georgie Healy: That's cool. Listeners of the show will maybe be familiar with stablecoin. We have touched upon it before, but just a quick one-minute summary for the listeners. What is a stablecoin as opposed to like a volatile coin?

42:44Dan Westgarth: Yeah, yeah. Stablecoins are backed by something. If you think of USDC, which is a US dollar backed stable coin, for every$1 of a stable coin, there is a dollar sat in a safeguarded custodied account and there's no volatility. The price remains the same.

43:06Georgie Healy: Beautifully said. Would you switch LLMs based on performance or are you a clawed-pilled user like many of the people that I surround myself with? I'm kind of guilty with my laptop stickers myself. You are purist. Are you loyal? How do you read the room?

43:28Dan Westgarth: i think uh i've built up habits and i use different tools for different things like anything that's integrated into the g suite i use i use gemini so for meetings it's gemini for google sheets google docs it's gemini um for for emails and text-based things it's chat gpt for personal it's chat gpt i love to use uh chat to read menus in foreign languages or wine lists or help with uh construction of uh like personal projects um and then claude for anything which requires anything which requires kind of more than one run and he will just kind of get on with it like yeah there'll be some some prompts or authorization or prompts to uh to continue with uh with its plan but um yeah if something is deeper and requires multiple runs than Claude?

44:18Georgie Healy: I am the same about ChackGBT. I feel like the voice is more personable, more easy to communicate with. I had my mission this weekend, Dan, to add some skills into Claude, to train it again with my voice. I've got a weekly sub stack. I keep being like, why do you sound so formal? It still doesn't sound like me. But watch this space. I'm going trying i'm gonna try again and you know uh if you had to you know um build one project over the weekend is there anything is there any big project that you just need a lot of time with and an ai

45:06Dan Westgarth: tool that you would build yeah i'm gonna get like a very abstract answer that you probably wouldn't expect i i think um i think i think farming is a really interesting case for for ai i think that i think of all industries in the world farming is probably at the end or the back of the innovation curve and it's so important we all need farming to to eat and survive as humans and that's the area that I would look into no idea what I would do but I'm sure AI has some cool application in

45:44Georgie Healy: farming and when you say farming everything like have we got cattle have we got like produce is it a hobby farm like what are we talking my my husband in the background has been chuckling along to uh the the Top Gear guys farm show I forget what it's called he loves that show like what kind of farm are you having Dan?

46:07Dan Westgarth: I think all of the above I mean I don't personally have a farm but um yeah I think I think Clarkson's Farm on Amazon Prime is a great show. That's the one. It just shows you some of the struggles that um farmers go through and if you if you if you track the um uh the price points of these different food types whether it's beef or corn or barley they're typically all trending up there's a lot of energy production um I'm sorry a lot of energy that goes into producing these foods and fertilizers. And there's a ton of inefficiency. So, yeah, I don't know where I would start, but farming is definitely interesting.

46:41Georgie Healy: In Yorkshire, is there good farmland there? Exactly, yeah. Yeah, yeah, very good farmland. All right, that's the next career pivot, Dan. This has been an absolute pleasure. I feel like you are just a mine of incredible insights through what you've been able to scale and build at Deal and in your previous roles. How can people follow you? What would you like to shout out about Deal? And yeah, take it away for us.

47:13Dan Westgarth: Yeah, I mean, follow me on LinkedIn. That's probably my biggest platform. Putting out a little bit of content there on AI scaling operations and Deal itself. Deal has recently signed a sponsorship deal of Arsenal Football Club. So you'll see us on the sleeve from next season, which is really cool. Yeah, it's always interesting seeing some of the Arsenal players on the England squad. I know they're country squads, by the way. But yeah, follow along and we're here for any of your payroll and HR needs.

47:48Georgie Healy: Okay, you cannot go. Arsenal, is that your team? How did that happen?

47:55Dan Westgarth: Yeah, I mean, Arsenal wasn't my team growing up, but I think I'm converted to an Arsenal supporter by now. Yeah, I think Deal has an objective to increase brand recognition globally. And I think Arsenal was a really, really good partner for us. They have so much global following and different people from different backgrounds, working in different industries, doing different things. So I'm really glad we got the deal done and the Arsenal team would be fantastic to work with.

48:29Georgie Healy: Strong, mate. Very strong. Loved this episode. Thank you so much for joining. And, yeah, I'm looking forward to the next Arsenal game where I see Deal on their jerseys. That's awesome.

48:43Dan Westgarth: Thanks so much.

48:46Georgie Healy: Thank you so much for listening to In the Blink of AI. If you want to go deeper on anything we've spoken about today, I write a weekly substat called Attention is All I Need. Yes, it's hilarious. It's a pun. And essentially, I go into AI rants, tech news, events I'm going to, and more. It's bite-sized, and I hear it's awesome. The link is in the show notes below.