From Proof of Concept to Product — Chris Horn, Deriv

Are engineering leaders asking the wrong questions when deciding what to build?

In this episode of Futureproof, Xano CEO Prakash Chandran talks with Chris Horn, SVP of Operations at Deriv, about what it takes to build AI inside a regulated, global software environment. Chris explains the difference between prototypes and proofs-of-concept, why data architecture is the real unlock, and how Deriv used a Shark-Tank-like model to introduce AI into internal operations. Together, they explore the mindset shift required for AI-native development — and why the most important question isn’t “Can we build it?” but “Should we build it?”

Topics covered include:

• Prototype vs. PoC: Why technical feasibility matters less than solving a real problem.
• AI as product work: The critical role of discovery, KPIs, and iteration in AI projects.
• Data as the foundation: How Deriv built a medallion architecture to get ready for AI.
• Internal AI first: Why customer-facing AI wasn’t the starting point (and what worked instead).
• Upskilling at scale: Building an AI-native culture through curiosity, training, and incentives.

Episode ID: 18501541-from-proof-of-concept-to-product-chris-horn-deriv

Subscribe to Futureproof wherever you get your podcasts.

From Xano - The fastest way to create a production-ready backend for any app or agent. Xano unifies AI speed, code control, and visual clarity, so you never trade reliability for velocity. Sign up for free today.

Transcript

TI-99 and Early Curiosity

SPEAKER_01 0:00

There's proof of concept and there's prototypes. You know, could we build this? That's proof of concept. Prototype is more like should we build it? You make sure that the right solution is being built. And then build something that's working. And if it's not delivering results, kill it.

SPEAKER_00 0:27

Hi there. Welcome to Future Proof. My name is Prakash Chandran, the CEO of Xano.com. Today we're joined by Chris Horn, who's the head of engineering at Deriv, one of the world's leading online trading platform providers. Chris's tech journey spans three decades across telecom, messaging, and fintech. He got his start programming on a Texas instrument TI899 and was part of the original team that sent the world's first SMS message. Very cool. Over the years, he has led engineering and technology strategy through multiple high-profile acquisitions, for example, helping Mobeon merge into Ericsson and MoVirtue into BlackBerry. Now at Deriv, Chris is focused on building a resilient AI-first fintech platform that serves millions of traders worldwide. His leadership philosophy centers on practical innovation, leveraging AI, automation, and data to drive a tenfold leap in productivity without compromising stability, compliance, or trust. I think we're going to have an awesome conversation, Chris. Welcome.

SPEAKER_01 1:28

Thank you very much, Prakash. It's a pleasure to be here. It's an amazing introduction.

SPEAKER_00 1:32

Yes. Well, you did it all. I'm just happy to be spending time uh with you. And I think it's pretty incredible that you have this career that has spanned 30 years. You know, um, I seeing the Texas instrument name, it reminds me of the TI-82 that I used to play Snake on. That's the graphing calculator for those of you that don't know. However, you worked on the TI-99, which is not a graphing calculator. It is a home computer. So I wanted to talk maybe about your start there, you know, like basically your early beginnings around how you got into programming and some of the pivotal moments that shaped your path up to where you are at Deriv now.

SPEAKER_01 2:08

Yeah, uh it's it it's got it seems like a life lifetime ago now. Uh it literally is. Um, so yeah, it's it was uh, you know, uh a present that I got when I was 11 years old, uh this uh TI99 first 16-bit um computer, home computer. And um I think you know, all my some of my peers and everything were just playing games on the on the on these machines at the time, really. And just, you know, you you remember uh uh you had to load the software on these um cassette tapes and it would take forever, and then you'd be filling around with a with a tone and the volume button trying to trying to get this thing to load. But I was more interested in how these things worked, you know. I was just fascinated by them. So I I I I bought various books and things and just taught myself how to code at that early age. And um it's just that that buzz that you got by um you know, sort of you've got something in your in your mind about how you want something to work and what you know, and then just things coming alive in front of you on the screen and um really getting into into the you know the I was even at that time I was you know programming in Assembler and things and just you know just having that sort of mindset. It was just fascinating time and I just really enjoyed, enjoyed it, got that immediate um buzz for it, and I was hooked, you know, at that stage, and um and I knew that's what I wanted to do uh from that from that early age. And um so I've always you know um continued to to on that learning learning path really, on that learning experience. And um, yeah, so um that that was that was how things got started.

SPEAKER_00 3:38

That's pretty amazing. Um did your uh you know, your family understand that you had like an inclination towards this? It's kind of like one of these pivotal moments where they keep thinking that in the world.

SPEAKER_01 3:49

I think they knew they wondered where I was. I was locked away in the in the in my in my dad's study for hours on end, you know, and uh until two o'clock in the morning. And um, yeah, so uh that that was it. It was pretty obvious, I think, that that was what I wanted to do. And um just under it's just you know, it's just really just got that buzz for how on learning how things work and uh because that's always been a fascination, isn't it? You just trying to understand actually, you know, rather than just you know, everyone else is playing games, you'd sooner be writing the games yourself and understanding how how the how these things, what's going on under under the under the hood, you know?

SPEAKER_00 4:22

Yeah, and it's very clear that that curiosity has served you throughout your career, but I still want to go maybe a bit further in the past. Like, so how did you take that uh initial uh curiosity that you have and like what what did that manifest in in terms of your first role?

The First SMS & Scale Lessons

SPEAKER_01 4:37

Yeah, so um I I you know after leaving university, I then I did study uh software engineering, and then I was lucky enough to I didn't want you to to do software development and um I was uh I joined a um I had a couple of choices actually. It's funny how your life uh pans out because you know there's there's the choices that you make, isn't it? You know, you can go one way or the other because I had a couple of uh options. I could have gone into the into the defence industry, which would have taken me a completely different path, but but I didn't do that. And um when we joined uh a software house which um was called Semagroup, they're now they're now part of um Atos and um learned a lot. I think those early years I spent the first eight or nine years there and um uh joined the communications division and we started working on uh some of the GSM um mobile technologies, and we had a contract for Vodafone at that time. And uh so um got some fantastic training there, fantastic team. Um, as you mentioned in your intro, there sort of uh eventually worked on in the small team that delivered the first SMS message and um uh you know started learning all about how to build you know scalable systems and um you know we were using um at that time it was Vax VMS, unbelievably, and uh desire base we were using as a database and things. But it's uh it's still, I mean, you know, a lot of the problems and the things they still still stay the same. A lot of the problems about how to scale systems and what you know what you know where the bottlenecks are, things don't change too much. Even you know, sort of fast forward 10, 20 years, it's still some of the same problems are still there. So yeah, that was um that was uh that was that was a very important time in my career, I think. And it was you know, I got some very good training there and um some good um grounding in how to do how to build software, how to how to work in teams. Um and also it took me around the world because um we then uh I then was uh traveled traveled all around the world actually installing the systems um and uh rolling as we because at that time it was in the mid to late 90s and the digital mobile networks were spr springing up all over the place. Um so that included um some time in the US um with the PCS networks as they were starting to uh to uh um to roll out uh with um and the baby bells and so on. So that was all happening at that time. And uh yeah, it was uh it was a it was an interesting time, very interesting time.

SPEAKER_00 7:13

Um yeah, and and then from that, I mean you kind of alluded to this, but you know, you learned a lot of lessons through there, including kind of like um maybe lessons around building something global and that needed to be highly reliable and scalable. Uh like tell tell us a little bit about um like you know the key kind of bottlenecks that you learn to overcome, like the key lessons that you learned in your time kind of in telecom and things that you take in your role today that you kind of got back then.

Transition to Fintech

SPEAKER_01 7:46

Yeah, I I think it, you know, things start. I think what I've learned over the uh there's some things that just never change. I mean, understanding the customer needs, having a product mindset, you know, um sort of um, yeah, building scalable systems, you know, getting the um the database is always a key thing to get right. The communication, the way that the services communicate with each other, um it's the same bottlenecks that you get everywhere, you know, um uh um I.O. Um, you know, network, network bandwidth, all these problems are still the same. Nothing changes, as one mentioned. Things just get faster and more efficient. But a lot of these problems from the early in that time is still, yeah, we still see today, right? Um, so um, yeah, that's and then sort of so all those principles are still the same. Um, having the right KPIs and metrics in place uh to monitor systems and making sure that uh you know you understand what's going on, being able to troubleshoot, um, having the right logging and and uh observability in place, um, you know, setting up the right KPIs and things. So from a technical standpoint, a lot of these things are the same. And they're we can talk about this a bit later, but they they also transition into other industries like FinTech as well, of course. Um telecoms is highly, as you mentioned, sort of, you know, global business. Um and incredibly scalable. Um, you know, uh we later on actually in my career, we we I was responsible for um we got uh built some voicemail platforms and um you know we had to build platforms that was at that time was scaling from um we started we had one architecture we sort of had to scale up. It started at about a million, two million uh subscribers, but it went over time it went all the way up to 80 million, you know. Um so serving 80 million users across a distributed um network of 10 different regions and things. And this is still but going back on going back, you know, 15 years or 20 years or something. But again, you again it's the same, it's the same, uh, a lot of the same problems that you we you know as we saw then as we're still experiencing today, even with cloud. It's still the same problems a lot more than that.

SPEAKER_00 10:02

Yeah, so you know, at some point you decided to make this move from telecom uh into obviously gradually the role that you have today. So maybe talk about that transition, how you started to make that transition outside of telecom and um and what eventually got you to derive.

What Deriv Is

SPEAKER_01 10:21

Yeah, I mean, uh so I I did spent a large part of my uh uh career in in telco and then um you know uh had an opportunity. I worked in a couple of startups and um started to to uh bridge out into into some other areas, more consumer-facing products and things. But then, you know, the opportunity uh came up uh at at Deriv, you know, I was approached and uh I thought this is this sounds great. I think you know, fintech's always been an area of um interest of mine and uh I wanted to uh you know to get into that. It was very attractive. And I thought I was actually fintech industry is is it it can be a bit of a closed shop, it's quite hard to break into, you know. It's it's it's quite hard to break into because you don't have the industry knowledge, and that's one thing I would say about Deriv is uh you know it's a it's a very sort of open company in that respect. Um I think it was almost it's almost encouraged to try and get people from different industries that come with different perspectives. Um there's a lot of lots there are there is some um very relevant experience, I think. Telco and fintech share a lot of similarities, they're both highly um transactional oriented businesses, you know, uh global as we talked about before, um highly scalable, both regulated businesses, you know, um latency, response times, all these things are are important in both both both uh both industries. Um so a lot of the the fundamentals that you need to have in place when it comes to building highly scalable systems and um highly you know um productive software teams is it's the same.

SPEAKER_00 12:01

Yeah. Um for those that don't know, can you just give a high-level overview of what Deriv is?

SPEAKER_01 12:07

Yeah, we're a broke we're a broker, an online broker. We we're um uh we we're a global business. Uh we uh you know we have we offer um uh we have a mix of in-house uh developed um and also third-party integrated uh uh offerings for our trade for our customers, trading platforms. Um we uh we uh we offer everything from uh uh options through to CFDs. Um and offer, you know, we have something called synthetics, which is our um uh uh which is our sort of uh sort of types of indices and uh options products that can be available 24-7. Um and uh everything from crypto to Forex included. So yeah, it's um, you know, it's quite a broad, broad offering. We try and cover as many bases as we can, really.

SPEAKER_00 13:00

Yeah, and I mean you kind of mentioned this earlier, but you know, FinTech has some of the same requirements as uh what you experienced during your time your telecom days, uh global scale, regulatory demands, 24-7 uptime. So I'm curious as to your the engineering environment there that you lead. Like how do you keep things agile and innovative uh without compromising the reliability that you need to meet?

SPEAKER_01 13:26

Yeah, I I'll I'll I'll be I'll be honest with you know Prakash, it's uh it's uh have have we got to the level that we want that I'd like to be at? No, I don't think I have. I think there's always you know a level of innovation and agility that we need to we go to. And I think that's that's uh why we're stuck, you know, we're we're partnering with with companies like yourselves to uh to see if we can improve on that. Um I think time to market is a challenge for us. We started our business 25 years ago, and um that's a long time to be in the industry, and we we've had a uh you know, we've we've had to evolve our system over that time, but it's uh it's got to the point now and we're we're looking to see how we can be more uh faster and more innovative. So I think you know, we s now we're seeing um we're seeing AI really being very you know important in the software development lifecycle and low-code tools as well. And I think this is an area where we're starting to really get um seriously.

Starting AI Internally

SPEAKER_00 14:31

Yeah, I think I think I what I'd be curious about is you know, we have a lot of application development leaders like yourself that will be listening to this. Uh, talk a little bit about the moment you felt like, okay, AI is going to be something meaningful that we need to start experimenting with and adopting. And how did you start doing that? Like what areas did you, given all of the this regulatory uh and all of the guardrails that you have to operate within, where did you feel comfortable to start experimenting?

SPEAKER_01 15:02

Yeah, so uh I mean, as you know, this yeah, this AI's really taken off literally in the last, what is it, yeah, yeah, 18 months or a year or so, is really when it started to really pick up. We what we've done is um focused uh on we started focusing and on our internal operations. Our internal operations. We we wouldn't want to um you know start by offering, you know, exposing AI solutions to our clients and customers because that's that's potentially you know risky. So uh we we we focused on our internal operations. We want we want to um and still do want to really infuse AI across our whole operations, um, use it to turbocharge um uh turbocharge our operations teams. There's you know a lot that we can do there. There's there's lots of manual work, mundane work, routine work that we can easily um automate and replace with AI. Um and so that's been our primary focus at this point.

SPEAKER_00 16:06

So was that just to dive a little bit deeper there, was that like a recognition to say, look, we need to start with our internal processes, let's have a meeting or some sort of council to uh articulate some of the things where AI could help accelerate and then put together like little, you know, uh SWAT teams to go take care of them. How did you go about doing it?

SPEAKER_01 16:26

I was sort of smiling about that because we're we're we we're we're sort of um uh you know, Deriv is okay, we're you know, reasonable sized company, we're you know, we're over over a thousand employees, uh, but we tend to be quite agile in terms of the way we work. And um so it's more like you know, we roll our sleeves up and then we just start experimenting in different uh uh um projects and um across you know simultaneously and see what what we can do. So it was started off as you know, pockets of different ideas and products um that we started to build internally, and then yeah, things just start to snowball from there. So um, and then we went on a bit of a recruitment spree after that um to let's to to hire you know um skilled people in AI as well. So but it it can't it comes it comes from a passion from the top, you know. Um our entire um leadership team and our CEO is passionate and our co-founder of passionate on AI. Um and you know uh these things start from the top and then they they permeate down very quickly. So that's what that's the way things happened at um at the Riv and um and still do. You know, so um we really yeah, that's that's that I think that's that kind of support and sponsorship from from senior leaders is needed really to make this happen. Otherwise, otherwise things just become more like pet projects, don't they?

Building with a Product Mindset

SPEAKER_00 17:51

Um so I think that's yeah, there needs to be a culture that does start from the top of, you know, we are going to be kind of this AI native, AI-first uh organization. And then also um the the ability to, like you said, be agile and have the capacity to start experimenting with things. And so, you know, I want to maybe talk specifically about some of the things that were built. I know you said internal operations broadly, but can you maybe speak to some of like the concrete AI workflows or applications that you've built and some of the results that you've seen?

The Data Layer

Upskilling and Culture

SPEAKER_01 18:27

Yeah, sure. I mean, so uh what I will say is uh as well, um Prakash, is it's this is about what we've learned. We've we've it hasn't all been smooth sailing, right? We've we've uh we've had to we've learned we've learned some lessons along the way here. And um I think what we've what we've learned is that you know um making successful AI um applications, let's call it, is is it's a lot like building products. You have to have a product mindset, you have to think like products. I'll we can touch on that a bit later, but some of the interesting things that we've done is um we have our um Amy uh customer support agent. Um this one now is handling um almost 70% of our uh initial um queries that come in through our customers, through our uh online chat. We um we could have taken some you know off-the-shelf solutions, they're very expensive, and we decided to build something ourselves which is a much more cost-effective way of doing things. Um and and that that project is is really delivering very important results, very key results. It's it's I mentioned before, it's it's handling 70%. Our CSAT score has gone up enormously, it was you know, hovering around 70% or something, and it's gone up now to the high 90s, and you know, our CSAT score. So it's improved our um customer satisfaction, it's improved our responsiveness. And it's also it means that when um uh you know our human. human operators can get involved in answering you know the the in the um interesting and important questions for our customers and they you know where they they can focus on the on the more you know the more um uh the sorts of queries that come through that uh that that need that human support and that that that human touch so um this has been a very important um you know product for us we continue to in to develop and enhance that further um we've also um uh we also uh uh we're very um big on security I would say in in Deriv you know we're we we we're we're paranoid you know about security we we have a I think I'm proud to say we've got pretty much a world-class security team and we've really uh who have embraced AI you know substantially uh we've got a number of uh uh successful applications around security one one I can just mention would it's a simple problem but uh imagine you've got um we've got over 300 uh systems that we use internally everything from you know um Slack that we use for our communications you know through you know we're using you know Google for you know various things um and um so lots of different third party systems one of the problems we've got is how to uh how to manage uh permissions and access and uh make sure that people have the right access to things and make sure that people don't have access to systems that they shouldn't have access to. So uh we have our um IAM team that that manage this we have our org chart and we have roles and we we actually we we this was an application for AI so we managed to build an AI uh application that can actually in real time analyze people's roles as they move people move between different roles in the company it it keeps up to date and it flags up when someone might have the be over provisioned or um under provisioned on certain roles it handles um onboarding and offboarding as well so we used um we used AI successfully for that we've used AI successfully um here for um our anti-fraud initiatives there's many uh because obviously if you're a trading company so you can imagine we've got fraudulent um clients that need to be flagged and handled so we use that successfully uh and uh we continue to to use AI in that area um so and then we we can you we we're using it internally in our compliance department um which is a big thing so many different projects actually we've got the last count over a hundred projects that we've got active at one time yeah um and um you know but but I mentioned before some of some they're not it's not all been plain sailing we've we've had to learn uh we've had to learn lessons along the way here yeah so let's let's I wanted to dive into that um you you talked about having a product mindset and you have to be very mindful when you're be you're building an AI application talk about like what that means and what people can learn from the mistakes that you've made so it so first of all what we did was we um we we we dropped this idea of proof of concept so we made the distinction between there's proof of concept and there's prototypes that the the there's there's a there's a subtle difference between the two proof of concept is typically you give to an engineer and they say oh yes you know I um I you know I and they can you make it work you know so they do something yes you know uh this works it's more like um could you know could we build this yes yes technically it's possible that's a proof of concept but a prototype is more like should we build it so you take you take it to the um to the you know your your customer your end user and you you build a prototype and you you you give something that's that's actually working. But it starts with um the product mindset which is understanding really the problem statement. So that requires a lot of questions you know and also it requires a consultant mindset because um the best the best AI engineers that we have internally are actually they act more like consultants. They're they're they're really answering all of the you know many questions to understand exactly what the um end user actually needs. Because if you ask somebody what you what they you know what they think they need and then and you take you say we're going to build an AI solution, they might make make a bunch of assumptions about what they think that solution should look like and what AI can and can't do. And they'll tell you something which isn't actually um you know really what they need. So it requires very much a lot of cons consultation back and forth and that sort of product mindset to actually make sure that the right solution is being built and then build something that's working as soon as possible and then be agile about it and iterate from the from there you know and if it's not delivering results uh quickly enough then be you know don't be afraid to kill it um in you know and instead of letting things drag on you know unnecessarily it's also very important to set the right um metrics KPIs in uh in place from the beginning so you know why are we doing this how many is it going to save a certain number of hours every month or you know how much is it going to move the needle by we we we we we set up a governance a small sort of governance um process in the UK we have this thing called Dragon's Den. I don't know if you have that it's I think it might be called shark tank in some other um yeah shark tank there yeah so it's it you know you come in and pitch your idea right and then you decide whether you're going to invest or not so we have our little um our little executive team I was part of that and uh you know we'd have our different um operations teams that would come in and pitch their ideas we we'd we'd make sure that we we we'd only pick pick the ideas that would uh have the potential to move the needle by sort of tenfold you know have a tenfold impact on on the business you know is it gonna save is it gonna is it really gonna save you know a huge amount number of amount of time and hours of people is it going to help us to be more efficient is it gonna you know somehow generate more revenue for our use you know for for for the company or what's it gonna do and so we um you know this is something that we uh that we really uh though I think was was helpful um and yeah so now now we're sort of looking at other ways of how we can scale up um AI across the the the rest of the business we we've got a number of projects as I mentioned but we um thinking now to you know put this more in the hands of the oper of our functional teams and um give them some more independence to drive projects.

SPEAKER_00 27:06

Yeah and then it sounds like there is an additional layer or filter that you have over it that you're saying look only 10x ideas is are the ideas that we want to consider. We don't want to necessarily waste our time um with anything else. So I really I think I appreciate that framework because there's so much hype right now everyone is trying to fumble and figure out what they should build. And it sounds like being very thoughtful saying at the end of the day we're trying to solve problems and there is a very methodical way that we should be approaching solving that problem. It just so happens that AI is a piece of leverage that we have in solving the customer's problem. Is that fair?

SPEAKER_01 27:44

That's exactly correct and uh you know we've seen there's overlaps as well Prakash with uh AI and automation you know people get confused sometimes between the two things um so and we've we've successfully combined you know AI with with with um what I would call business process automation solutions as well so it's it's not just it's not you know it's not only just about pure AI is is usually you need some things around that you need traditional you know user interfaces and you know um and you know automation tools and other things as well the other thing I would say that um we that we've learned as well is that uh um data is key of course to any a any AI project it's it's only as good as as the data that uh that you feed it so we've invested in building a data layer um we're still you know continuing to do that and we're learning along the way as well with that because we've noticed and we realized that performance is key for the data layer and also having it structured in the right way as well is is very important.

SPEAKER_00 28:46

Can we at a high level talk about what that means because I've heard this before saying that data is the most important thing. Yeah are you yeah talk about what you mean by that because traditionally especially at our organizations that are as uh are are as large as deriv or otherwise have data siloed in all of these different buckets um and sometimes they feel like okay I can just install for example like a you know uh Microsoft uh and use some sort of SharePoint to like stitch all the data together and then I'm good to go. Talk about like how you've made your considerations around data, the data layer and things and lessons learned that we can take away from this.

Hiring AI-Native Talent

SPEAKER_01 29:26

Yeah what so what we've what we've learned along the way sort of the again is the hard way is we we started off with uh you know where traditional data warehouse and then building I'm going back now a couple of years you know before ai even building dashboards you know analytics using traditional tools like BigQuery and Looker Studio and things and uh you know serving up dashboards and reports for our for our teams um and you know uh a lot of the time that data is stale and out of date and things uh or it's you know it's potentially not bang up to date anyway. But so that was where we started and then um we we realized that we needed to get access to more real-time data for you know that's coming through uh with our trading your you know trading platforms and things um so we uh started to uh explore you know and build um build out from this um it's there's there's fundamentally two two parts one is a sort of real time access to real time data so having the right types of um the right types of architecture the right types of products that can stream stream things in real time and the other thing is curating the data properly so um having it processed you start off with we're actually building something yeah it's uh we're following a a model called a medallion architecture uh which is essentially made up of gold silver and bronze where your bronze layer is raw sort of you know unprocessed data that comes in from different platforms um silver layer which is semi-curated processed data which could be used by applications and then a gold layer which has been um properly um curated in a way that can be served up through APIs and used uh for applications such as AI. So this is what we've we've learned um but then what we're doing now is as well is we you know we've realized that some of the applications that we're building need even uh more um real-time access uh you know um access through instead of just um something which might be you know half an hour or an hour out of date it's actually wanting to to uh to take stuff straight out of memory so we're looking at uh we're looking at having data served you know um and curated through vector databases of course which are needed through for AI uh AI applications and things that and and actually serving some of this up in in in using you know inbuilt memory uh memory memory you know real-time memory services that you know so so things like you know we uh things like Redis or um mem0 memcache there's there's many services that can do that so this is the next the next step that we're that we're in um yeah but but it but fundamentally it's it's having having the data you know um accessible um and organized in the in the right way is very key to these things.

SPEAKER_00 32:39

For sure. I want to shift gears a little bit and talk about just the organization and organizational leadership, people management, et cetera um maybe a harder topic is you mentioned you know Amy for example, the support agent handling 70% of online chats. To me what that means is there's human beings on the other side of that that aren't handling that anymore. Talk a little bit about how you have had uh the decisions that you've had to make in introducing some of these efficiencies and generally culturally how you prepared the organization for changes like that.

SPEAKER_01 33:16

Yeah that's that's it's a very it's you know important question I think is you know it's it's this trend is unstoppable. It's there's this you know it's it's it's happening whether people like it or not. So I think what we're encouraging everybody to do in our organization is upskill you know everybody from people that's in HR you know through to customer service or whichever department they're in we're encouraging people to upskill and learn AI as much as they can. You know that means um we're you know we're we're using we're using the sort of you know the um I think we're quite interested in things like clawed skills which we're using internally as that looks really nice. You know uh ChatGPT Enterprise looks really nice. So we we we're using these tools internally as much as we can to across the teams um in all the different departments to um uh to encourage people to upskill and yeah we've already seen I mentioned with the CS uh the customer support um that uh you know some roles inevitably um will become uh you know replaced or will be um irrelevant as we know them um but that doesn't mean that there's uh you know the opportunity for people to upskill and transition into other roles as well of course so that's something that we're encouraging across all of our all of our teams but fundamentally as a business we we want we want to get more efficient um and you know we're very passionate about AI and we believe that um that it can allow us to do that and um and be more more cost efficient and and more you know uh be able to to serve up um uh provide a much better customer experience in the end of the day for our customers is what it's about.

SPEAKER_00 35:04

Absolutely I think that a theme around upskilling is really important.

The Future: AI + Low-Code

SPEAKER_01 35:09

Can you talk just at a high level what what are some of the things that you at Deriv offer uh for upskilling I I know that people at least at other organizations that I've spoken to is like okay well use ChatGPT to get maybe some of your uh your job done more efficiently do you offer any sort of training or things that leaders can take away from it yeah yeah yeah we do we have we have we have we we offer the train yeah we have we have um you know uh training packages and things that we can that we provide as part of our uh benefits package but I think what's we we've also done things like um we've run internal competitions uh where you know where we we offer rewards to people so you know you could say it's not huge you know but you know uh whatever it is a few hundred or you know a thousand dollars or something uh you know drive some internal and uh internal competition for teams so they can come up with some nice innovative AI-based solution that can um move the needle for the business so we've we've done that successfully um a few times um and that that sort of encourages people to to upskill um and you know have some learning on the job because I don't know what you think Rakash but you can all the training courses and things in the world you can sit and read books and things but there's no the the best way to learn is on the job uh and if you've got something that in your mind that you want that you think you can build you know having that drive and going back to what I was talking about with the start of this podcast with the you know Texas Instruments TI99 it's just like you've got something in your in your brain that you want to build it's the best way is to learn on the job I feel and then of course forces you to do the research yourself and then you go off and you find the online courses and you read the books with a yeah I think what I have found is like this insatiable curiosity um that is required because things are changing so fast right you can't just learn a course or take a course learn something because the next day it's going to be different right so you have to be hungry you have to be curious you have to be that's what when people say like hey this person is AI native it means that they live and breathe in the space and it's just a requirement for kind of operating in today's environment.

SPEAKER_00 37:26

You mentioned curious so that's exactly the word that we've we use internally as well is is is um yeah people have to be have to stay curious and be curious and that's the best way to to learn yeah so this leads me to the other side of things you know we talked about you know the inevitability of when you drive efficiency you know sometimes uh some roles and positions will be made irrelevant and that doesn't that person can upskill into something else but that's one side of it. The other side of it is this new world we're entering into you talked about hiring AI engineers and AI native people. Like what does that mean? What does that profile look like to you?

SPEAKER_01 38:02

What do you look for that tells you like hey this is this is going to be a good fit in terms of leading our development uh in a new way for the future yeah I mean apart from in the end what would what really matters is is people that have a passion and and an adaptable mindset and the and the ability to learn quickly right so I think that's fundamentally what what we're looking for we're looking for you know very smart people that can pick things up quickly and learn of course they've got the AI you know background if they've done the the training and everything that I think that's that's key you know uh understanding the fundamentals um is key for as well but I was going back as well to this we we also um I was explaining about the the um product mindset and the sort of solution mindset so we're looking for people that can really have the ability to communicate well to ask the right questions be inquisitive like we said already be curious um and uh you know be adaptable so I think these things are just as important um as the technical capabilities I think this is you know this is this is key because if you um what you know we we've see people that perhaps you know there might be PhDs or you know um academics um from the university or something but they're not necessarily they don't always make the best sometimes they do but they don't always make the best you know um product builders uh ai solution consultants because they tend to be a bit theoretical sometimes I'm I'm being I shouldn't generalize too much but you know what I mean this uh so we we're looking you know we're looking for people that that uh that are that can um that are very well rounded um and that can interact with our with our operations teams and really trying to understand exactly what it is they need And have that consultative mindset that I was talking about. I think that's that's very important for the success of these of these projects.

SPEAKER_00 40:08

Yes, I I totally agree with you. And I think another piece of that is the just the versatility, like making sure that because things are moving so fast, there's new AI tools, uh, you kind of have to be able to be experiment. Because if you want to maintain the agility that we talked about earlier in the conversation, you have to be able to do that. And that's where, you know, AI, low-code, no code, even some of these other modalities that are coming out, like so critical to be a student of the space.

SPEAKER_01 40:37

Exactly, exactly. So yeah, I totally agree with you. You know, the other the other thing just that it's for me to mention quickly as well is about our software development. So we, you know, we we are, as you know, we're partnering with yourselves on uh on the software process side of things, and we're really looking to see how we can use AI um end-to-end in our process to speed things up, all the way through from requirements analysis through to, you know, um solutioning, you know, all the way through to obviously building and coding and things, but all the way through after that testing, um, and then when things are in the field, you know, monitoring. So the whole thing really is I think keep AI can really help in all of these areas and help us to um help us with that speed and agility because we're not there yet. We're not there yet. Um that we've still got a long way to go, actually. But um, I think this is important for us as well. This is a big thing for me as an engineering, you know, leader. I I need to really focus on that speed and agility and continue to focus on that. And I think that's what that's why AI is so exciting.

SPEAKER_00 41:41

What do you feel like the biggest bottleneck is for you in in getting there?

SPEAKER_01 41:46

Um I think uh it's a good question, Ashley. Uh it's you know, there's more there's more to building software than just writing code in the traditional sense. Um it's really about uh, you know, uh understanding cust the the requirements, really making sure we're building the right things and making sure that the quality's there. So these are some of the the the bottlenecks. I think there's a lot, there's sometimes there's there's there's a tendency to um jump into solutions and and develop things without necessarily making sure we're building the right things and making sure that there's also enough attention to detail that's played that's paid in certain areas. So um it's getting the right balance between spending the right time in the right areas and uh making sure that we spend uh yeah, that the attention to detail is there, and then using AI to speed up the uh uh yeah, speed up, speed up the process end to end as well. Yeah.

Final Lessons

SPEAKER_00 42:48

So as we start to close, I've really in enjoyed this conversation. We've we've touched on a lot. Um I guess one of the questions that I would ask is based on where Deriv is headed in the next 12 to 24 months, what what are you most excited about?

SPEAKER_01 43:05

Yeah, I think uh obviously it's a very, like we said before, it's a very fast-moving space. This um there's I'm very excited to see what uh what uh how the models are going to evolve now over the next 12 to uh 18 months or so. Um I don't think we're anywhere near reaching AGI yet, so there's but there's there's still hope for us all. Um but uh but uh I think for for us at Deriv it's it's about um more of the same in terms of more growth, more innovation, you know, more speed, um, and uh making and you know making sure that we we we you know we delight our customers. I think this is what we have. We have to be customer, and we are customer obsessed, and we we need to continue to do that because end of the day it's our customers that are paying the bills and we have to look after them.

SPEAKER_00 43:52

Absolutely. Um from an AI capability or tool standpoint for application development leaders that might be listening to this or or builders, where where do you see kind of the puck going? Like where is a space that you think more leaders and more builders should be paying attention to that is potentially a little non-consensus?

SPEAKER_01 44:14

Yeah, I I mean you probably like me for saying this, but I think you know, I mentioned before that uh low code is definitely an area that uh is a little bit perhaps um I think people are missing out on the comb combination of AI and low code. I think we're very excited about that. Um I think uh, you know, um the way that AI um can be uh works, it's based you know, based on sort of building blocks. And I think um that combination is can actually be quite powerful. So we're quite we're pretty excited about that. Um I think you know there's a danger with just traditional vibe coding that if you don't have the right guardrails in place, you could end up in a real mess. So um I think having having that combination is quite um important, and I think that's something that we that we'll continue to to focus on at Dariff.

SPEAKER_00 45:06

Nice. Um, I guess finally, if you had one piece of advice for an engineering uh leader like yourself that's listening to this, that you have learned over the past 18 months, uh, or let's let's shorten the time horizon, 12 months. A lot has a lot has changed. If there's one piece of advice that you could give, what would that be?

SPEAKER_01 45:26

Think, think, have the product mindset. Think like, think, you know, think about what the customer really needs and have that product mindset and build. Remember what I said before about building prototypes and not proof of concepts.

SPEAKER_00 45:40

That's awesome. Um, well, this has been a great conversation. Where can people go to learn more about Deriv, uh, about you and what you're all building?

SPEAKER_01 45:49

Yeah, uh, check out our we've got a tech blog um and uh check out our website, deriv.com, and you'll find there's uh various links and things on there. Um, access to training materials, learning paths, talk to Amy, ask it, ask it some of um questions around uh how to get started with trading.

SPEAKER_00 46:08

It's fantastic. Well, Chris, thank you so much for your time today. It was a real pleasure. I learned so much. I I always take notes during these podcasts, and I just have a sea of notes here on the left hand side. So I appreciate your time.

SPEAKER_01 46:20

It's a great pleasure. Thank you very much.