Transcription
Singapore is one of the most digitally advanced societies in the world. When you use our government services, be it paying for your taxes, registering a company profile for your first startup, or even applying for grants, the process is surprisingly painless. If you are a Singaporean in the know, your mind will immediately think of Singpass. This humble little app is what we use to verify our identity on almost all the important transactions we take.
But what doesn't really get talked about is the army of public servants building and maintaining all of such services at the back end. I had the chance to speak to Sasha. He's the CTO of GovTech Singapore and the man in charge of developing our national tech stack. He happens to also be one of the leading practitioners of AI in government and building products. If you want to get a sense of what AI governance in Singapore actually looks like from the inside, from a practitioner perspective, this is the conversation for you.
The following conversation took place at a community event that SuperAI, Carta, Singapore Global Network, and myself jointly organized. I hope you enjoy this. I managed to speak to this gentleman called Merhan Go. I don't know if you've heard of him, uh but he he has this book called The New uh The New The New Map of Innovation, The New Geography of Innovation. And his central thesis was that if you look at the different countries all over the world, there are different spots that actually promote technological innovation in a very unique way. And Singapore was an example. And he said that, you know, in Singapore, the public service is leading in digital innovation. And that's actually not very common across the world. So, I wanted to ask you from your experience as someone at the front line, what makes it unique about Singapore that allows us to kind of deliver public service at scale through digital means efficiently and with excellence?
All right. Hi. Hi everyone. So, first of all, very high praise from Keith. I don't know whether I actually live up to it. It's actually teamwork. So, so I'm not really responsible for all the CDC vouchers you guys get. And every time you cannot log in the same pass, please don't don't call me. Of course not. So, but anyway, SingPass has 99.99% reliability. So, I think we are doing pretty well. Um anyway, to to answer your question, I think uh it could be a bit long. But maybe I'll start off a little bit about uh myself. I I joined GovTech about 3 years ago. But actually, I was with the journey or the tech journey with uh the predecessor of GovTech uh long, long time ago. So, uh I spoke to some gentlemen earlier on and uh actually I was in National Computer Board in '97. I think that they were uh not born yet at that point time. Uh but I I was very much earlier on in that organization. And believe it or not, our journey in doing digital transformation or what was called computerization at that point time started in 1981 uh by this agency called National Computer Board, which then evolved into the Infocomm Development Authority. And then about 10 years ago, it became GovTech Singapore, right? So, I was in uh NCB, National Computer Board, and uh I was part of the team that was doing computerization for the uh Singapore Sports Council. If you if you uh know what that organization is, it's actually embedded as part of the old National Stadium. Um so, so that was the authority that was promoting sports in Singapore. Anyway, I mentioned that is because um Singapore actually started very very early on. It wasn't a uh it wasn't a sort of like fluke that we suddenly became good at technology, right? So, Singapore started uh the computerization project in 1981, and that was a long time ago. And its journey actually continued then through different phases. Of course, there were different things that it did over time, but uh I think the very DNA of uh the Singapore government actually has technology built in. And I and I think that is one reasons why we do uh a little bit better than some of the other uh countries around the world.
I do interact with my peers uh in other governments around the world, and um I think when we when we talk, actually there are there are differences in the way that we're structured. As in uh many of the other countries around the world, the technology is provided by uh suppliers, vendors. Uh and it is often bought from an external organization. Whereas in government, we literally have a government agency in charge of technology. So, I In fact, we have more than one. We have three government agencies uh that are in charge of technology, and uh NCB was the oldest at the point time. Then we have DSTA, and then very recently we have HTX. Of course, for the healthcare cluster, we also have Synap. So, also actually it's sort of like uh we probably have four, right? Out of that. So, I I think we are kind of uh uh well well, I would say uh we we have a lot of technology, and we we are actually very serious about technology and how technologies can support our operations. So, I would say um it is wasn't like a very new thing that we we got into it, and uh yeah, I guess that's the reason why.
There's another part of the secret sauce. And I was to some of my friends in government and my friends from overseas. One of the big things in Singapore I've come to realize is that you have a lot of talent, especially in the technology space, that end up working in the government. Someone like yourself could be an example. You don't see that actually to be a very common thing across the board. I spoke to some friends in the US. I said, "Hey, like, you know, like if they if they had a career options, right? They would say, 'I would never really want to like be a civil servant as a software engineer.'" Whereas in Singapore we see that in GovTech you're actually hiring really world-class I have some friends that I think they are world-class software engineers to be part of that team. Can you help me understand a little bit more about that process? How does one How does the government actually position itself to be a employer of choice for many of these talents in the marketplace?
Well, I I think we have a very good marketing team. We have a very good recruitment team. Uh man, this not of my effort. I think uh there has been a lot of like selling of Singapore and we do sell beyond our shores, right? So, so we do go outside of Singapore uh to look for Singaporeans who have traveled outside, worked outside, and we try to bring them back in order to help us with this effort. So, so I think we we do provide a uh a good means for such technologies to come back to Singapore uh or enter the public service in order to to do good. In fact, uh I mean, I left NCB in '99 to do my my first startup. I spent many years uh doing startups, being in big tech companies, and so on. Um and I came back to public service because I feel that uh you know, that's a uh I need to serve the nation, right? And I think that's that's something that uh we do often sell to uh candidates. Um and when our candidates see our successes, when um Singaporeans see our successes, they do want to be part of that. They want to be able to contribute to real impact. And I think that that is also the other secret sauce because um if you've been out doing a lot of commercial work, very often it translates to just one thing, right? Which is revenue. Right? You con- contributing to shareholder value. Um but it's not so so much in Singapore government. You're contributing to the value for the whole country. It's your home country. And I think that sort of uh uh strikes a chord in many people because that uh I mean, you live in Singapore, your family's in Singapore, this is your home, this is your country. Then then it makes a lot of sense for you to want to contribute. Um and we are kind of open for people to contribute in different ways. So, we have people joining from uh at all levels, but we also have people joining uh as Smart Nation Fellows. If they're pretty senior, we provide options for them to come in to help for a period of time. And then uh they they can help uh which areas they have expertise in and and so on. And we won't hold them for a long time. So, the Smart Nation Fellowship is something that we uh use as a tool to to encourage more participation. Um I myself was like uh I think the second Smart Nation Fellow uh when they started the program. And uh eventually, you know, I I did join GovTech and uh actually, uh many of the Smart Nation Fellows do turn to to be GovTechies after a while because they they enjoy, as as myself, I can speak from my experience, um I enjoy the experience of being able to contribute to something uh valuable, something meaningful. And uh I think that for especially a lot of builders, right? Uh people people like myself, uh we build stuff. And building stuff to earn more money for our shareholder, it has a very different meaning to building something that has value for the country and for your family. So, so I think that's that's the secret sauce.
Can you give me an example of like some of the things some of the projects that you are very proud of that you felt like, "Wow, this was really a feat." and come up obstacles that you had to overcome?
Uh I think there are there are a lot of projects um that I can talk about. You just mentioned some of them earlier on, right? So, like SingPass for example. SingPass is actually a pretty old uh product. I'm not sure whether you guys know about this, but SingPass been I think last year was the 20th anniversary for SingPass. It started off as a login application uh for people to have a common login to to government systems and then it evolved to where it is today and it has a lot of a bumpy road along the way, but um it has come to a point where, you know, it is something that a lot of Singaporeans are very proud of. And a lot of people come to me like, "What happened to SingPass?" And uh first time they introduce me, they say, "Hey, you know, this is somebody who is part of the SingPass team who has delivered this." And um and and I think a lot of uh Singaporeans abroad I also identify with SingPass too that there is a certain value to it. Um but besides that, if you notice um a lot of the agencies uh applications for any one of you um from parents who have kids in school. Um you you use Parents Gateway is uh is a MOE system, but it's actually developed by by GovTech, right? And if you notice there are a lot of such systems in government. They're actually an underlying part of it is actually developed by GovTech or some GovTech team that is embedded within the agency. Like the government paid leave system. So, some of these systems are actually developed by GovTech as well. So, all of these have real value because they do contribute to the country. Again, CDC vouchers that you you receive things like very recently um very recently there was a culture pass that was that if you actually use anything on culture pass, there is a system that was developed by by GovTech as well. So so many many such projects. Of course a lot of them have its own fair share of problems. It's not always a bed of roses. I mean, developing the products is one thing. Very often the the journey of getting the product rolled out if you're developer or you're builder somehow, you know it's never easy journey. And very often there are a lot of different things that we encounter along the way. But that's part and parcel of any any kind of job, any kind of project that you do. So I wouldn't really elaborate on that, but I I would say the satisfaction of putting something into the hands of citizens that they that makes real impact to people on an everyday. I think that's the one that is very valuable.
There's this essay that he wrote. It's a 37-minute long read on Medium. I recommend you read it. It's it exposes buy versus build dilemma that governments face. Do you simply buy software from other people or do you build it on your own? And he tied I think he tied it to this AI movement as AI especially the large language models became much more widely adopted. The cost of say building [snorts] a product or paying a product was kind of automated. I wanted to get you to kind of elaborate a little bit more on how GovTech for example has is developing an AI strategy with regards to using AI. How does it inform the way you build or buy digital products digital services for citizens?
Yeah, so so this is actually a very very complex thing to talk about and probably I have to to explain it it would take days and I probably would not be able to explain it well. Government overall has 2,000 plus systems. Um and GovTech alone we have within GovTech which are GovTech systems we probably have like 70 odd systems that we run. Most of them are platforms like SingPass that power many other systems as well. Like so so we do build a lot of things but it's impossible for us to build every single system for the whole government. Inevitably there will be many systems that bought either as commercial off-the-shelf cuts or they are bought from suppliers like system integrator that will then engage Accenture say hey, you know, we need this system build this for us we provide specification and so on and so forth. In fact for a large period of time within NCB and sort of GovTech the IDA days and this is actually pretty common in the industry. Um offshoring or outsourcing was the the norm. Like so you would actually outsource your projects and you get a vendor to come in to build the systems for you. Actually there has I mean this actually caused a number of serious problems over time cuz initially of course there were a lot of value a lot of benefits to do that. You reduce your manpower requirements um and of course you get a larger industry and then you actually promote the industry to to grow as well. So so there were plenty of value at that point in time. But over time actually what happened was that we started losing our own capabilities. As in within the government you lose the capability to own the systems. And of course if it's a peripheral non-important system that you could lose and then this and then you still continue. That's That's okay, but it becomes core to your operations, for example, it's the core of what it means to be the government agency, the services that you operate on behalf of the government, then it is it is terrible, right? Cuz that means that you lose the ability to to operate as a government. And unfortunately, a lot of the systems tend to do after a while become like that. And part of the the reason that we set up GovTech was to rebuild these capabilities such that we are able to capture back the the capabilities and we're able to continue from that. So, there is actually a huge um uh movement to do modernization within government driving towards that direction uh where we want to build capabilities. Now, how how how does this uh mesh with the buy versus build, right? Is build the the answer versus buy? Actually, the answer is no uh because again, as I said, we would never have enough manpower to build every single thing. So, one thing is that you need to have a strategy about what is the best thing that you want to what are the things that you want to actually build versus buy. You do want to um you do want to capture the ones that are more important for you, but how do you define what is more important for you and what is less important for you? Because over time, the operations change as well, right? So, so I think it is a deliberate um matter of how do you retain the capability and the retain the the control over the systems that has become that needs to be a strategy. So, secondly, uh we are trying to push out a platform strategy. Platform meaning that you control the core part of the systems that are most valuable and you get uh suppliers to build on top of those platform. And you want to make the uh the uh systems more commodity such that you can switch between one system to another system. You want to be able to have some level of standardization so that you you can get your suppliers to adopt us similar standards such that you can easily switch from one supplier to another. So there are a lot of such strategies that we need to employ to to drive towards that. Um so and and the answering the other part of the question about the AI. It is true that AI has uh made a lot of things a lot easier to build. But you know we we at this point in time at least, I don't know, maybe like one month later or half a year later, uh we're not going to rebuild SAP systems, right? We're not going to build rebuild Oracle systems. Uh they're so embedded into what we do today and there's so much functionality and learning experience in those systems itself. You There are certain systems that you don't want to rebuild. But there are many others that are just simply built on top of uh whatever that is at this point in time. And those are the systems that we want to change. We want to establish platforms such that we ourselves can build quickly on it, but at the same time we also want vendors to be able to build on top of it. So generally speaking that's that's the strategy.
Can you talk a little bit more about uh AI adoption within GovTech? Um what are some of the considerations that you or maybe the organization has towards adopting AI cuz I assume they'll be quite different from private sector. For example, it might put a much higher premium on say security or ownership as you put it that you really want to own the platform? So in the way you integrate AI within your workflows, how how do you see it within GovTech?
So so first of all uh actually it is not uh uh it's not a question because we we are I mean, AI is the the concept of AI and whatever AI can do is changing, I think almost on a daily basis, right? So, uh yesterday last night, uh Tropic just released or just announced their Metos and also announced manage agents. So, suddenly, you know, a lot of things that we thought, you know, we should be doing or we shouldn't be doing has has changed. Um, like when Metos was announced and said that um you could actually uh be able to to to find vulnerabilities in systems that are I I can't remember how many years old um and suddenly a lot of cybersecurity people start to wake up. So, so it's actually a very difficult thing to to talk about, because you don't know, like I talk about it here right now, maybe I sound authoritative, but tomorrow everything is gone, right? Because it could be not true anymore. Uh, so it's kind of hard to to So, I I give this disclaimer, because I I don't want to give people a false impression that whatever I say here is is authoritative or whatnot. But, actually in government, while there are there are certain uh idiosyncrasies of how we operate, but by and large, many of the things are quite similar. So, adoption of AI, I think there's a lot of push up for adoption of AI in government. I mean, it all comes all the way from the Prime Minister, right? So, everybody and it the pressure trust me, the pressure is really really strong, right? For an organization like GovTech, cuz cuz we are GovTech. So, we have to deliver the AI that's required by all agencies. So, there there are a lot of pressure for us to deliver, not only deliver, but deliver very quickly. Um, and we need to look at it at at many many different facets. But, of course, the strongest facet for me and for GovTech is how do we use AI in software development? I think that's a key thing that we we want to to do. Um, and that is something that we are trying to drive towards a lot. So, for development, we had been we have been having a program for developer productivity for the past 2 years that that provides tools AI tools for our developers to use. One of our earliest tools, or rather the the earliest tool that we provided was the GitHub Copilot. We also added things like Codium that eventually became Windsurf. We also had the GitLab Duo, which is a very popular tool for us because we are big GitLab users. But over the past year or so, there's a lot of a lot of switch in the tooling the field, right? A lot of people have switched over to tools like Claude Code. So, today Claude Code is the most dominant tool, I think, for software development. There are other tools like Cursor is also pretty popular. Open AI came out with Codex and so on. But I think Claude Code is probably the most popular and we have been ramping up the the use of Claude Code. In fact, yesterday was the day that we launched the use of Claude Code official use of Claude Code for classified systems. Before that, the unclassified systems, we do allow people to use the different kinds of tools, but for classified systems starting yesterday, we we announced and we gave credits to all our developers to use such that they can use up to a certain amount of tokens. And then afterwards, if they still need it, then they need to justify, but that that number is actually reasonable, reasonably high, right? Moderated level to what the management thinks is high and what the developer thinks is sufficient, right? Uh so so we are still monitoring the usage and we are monitoring the ramp up as well and it's ramping up pretty well. But first the next thing that we need to monitor is like with all this money pumped in, is it really doing as well? Um so if you have you have read one of my articles, I did interview with a number of my teams who have early adopters of Cloud Code and ask them how it has actually helped them and and all three of them tell me multiple multiples of productivity, right? So they're like five times, 10 times. Someone even told me 30 times in productivity. Not sure whether how true that is, but uh I but you know, they they have been able to do a lot more than what they could have done in in the past. So so I'm looking forward to more of such things. Uh so that's that's for my industry. But actually for many other different industries within government, different industries because the government takes care of the whole spectrum, right? There's momentum to drive adoption like Ministry of Education, for example, there's a significant adoption in tooling for teachers, for students, uh for general productivity, there's tools to help teachers to be able to mark a lot more uh papers, tools to help students to be able to learn better in terms of learning mother tongue. So so a lot of such tooling that's happening and other agencies, we are adopting different things as well to speed up in terms of the productivity as well as uh generally helping the general public use officers to be a lot more efficient in their work. So so quite a bit of stuff. Yeah, and we're reaching the the end of our time together. There are trade-offs to be made. You have a limited supply of workers. And there are so many ministries. So, how does GovTech kind of prioritize or look at deploying talent so that across the board, the entire government enjoys the benefit of um AI adoption in in their service delivery.
Yeah, so so there are different ways dimensions that we're looking at. Definitely one of the dimension is how many agencies are need the same thing, right? So, for example, one of the the most commonly used tool within the government agencies uh are case management. So, case management is actually a main use case for uh government because many agencies do processes do process uh uh case cases and uh there needs to be a way to uh manage uh cases. But, of course, different agencies, even within the same agency, there are different uh use cases for that. Um so, we want to build something that is both um there's a common platform, but at the same time it is um uh adaptable to different use cases. So, normally what we do is we build a a platform. And then uh we build some of them we will engage our uh suppliers to build uh applications that sit on top of it so that it can be adapted to the use cases. So, those that's one angle we look at. How many agencies and how many use cases there are, right? Secondly, which the uh second parameter we which is actually very important um which is not very obvious is the willingness for the agency for to for us to develop for them. Um it's it sounds ironic, but actually part of it is is uh is because sometimes the agencies want to um control their own fate. And so, they are very willing to pay money for uh developers to develop for them. But those in those cases is actually not very optimal because then you would have the same application replicated all over the place and it's not a good use of public money, right? So, what we need to do is then we need to engage the agency and get them to on board to what we do so that at least we can build once and it can use multiple times. Um in many many uh use cases. But it's not so easy because of course every agency would have their own way of thinking of how to do this. So, again, for me to develop something, I need the agency to collaborate and cooperate with us. So, the uh agency's willingness to work with us is also very important. Um thirdly is is the value of what the common thing can do. If the value of course the value is multiplied by the number of agencies using a number of uh use cases there are, but sometimes there will be just a few use cases, but the value is tremendous. Right? Because maybe there's just given one use case, but the value is so significant for the uh for that particular use case. It is very meaningful for us to then go in to develop. So, so generally these are the three parameters that we look at. Uh first is the the scale. Secondly is a willingness of the agency to work with us and third is the overall value of the system.
You have an example for the last point? Any examples for the last point where you thought like that that could be a useful
So, I have a system that is uh that actually processes value of a trillion dollars a year uh for Singapore and that's uh I would suppose it's a valuable system. >> Okay. And we want to participate in that. So, we want to help.
Okay, so uh I go on final question for you. You just said there are a lot of like younger folks here in this room. You're a tech leader yourself. You've been in public sector, you've been private sector. There's a a of AI anxiety now. I think I can't put a finger into it, but my suspicion hypothesis is that like knowledge work is being disrupted now and a lot of people just haven't experienced that kind of disruption before. From your point of view as someone who's leading a technology organization, uh how how do you recommend like younger folks think about using AI in your career?
Yeah. Um so, this is a frequently asked question, right? A lot of people who who engage with me often ask this question. And uh the the real real ground truth is the real real reality is I don't know because all these things are new. But of course, I have sense of something that that's happening. Um because I I engage with this every day, right? Uh on on the field and and and daily we use this technology. It is real, right? The anxiety is real. You have all your right to be anxious because it is disruptive. The technology is very disruptive. It's not disruptive in in a way that you know, it it it destroys things, but in a way that it is so it helps you to be so much potentially helps you to be so much productive that the number of people that is required is significantly less, right? Uh I mainly talk about the software industry right now because I mean, I've been in in this industry for 31 years. But I see this happening in many many industries. Like I I have a good friend who was in the was a legal officer in in Telstra in in the public sector. And you know, the legal industry is totally disrupted except that probably they don't know so much about it at this point in time. But of course, all software engineers knows about this because they face this every day. So so, they have uh they understand this is total disruption. But how do you deal with it? I think as an industry, while it looks like oh, we're going to we're going to lose it, right? Because we're going to lose our jobs. But reality is I I've seen uh the industry responding in a way that's uh actually I mean gives me a lot of comfort because the industry itself is adapting to what is needed. And it's built building is a sometimes it goes down the wrong path uh but other times it is actually adapting to the situation, adapting to what is needed, and it's moving forward in the right path. At least that's that's what I see. Um and secondly, the uh actually affects the junior uh developers differently from the senior developers. And um if you look at it from the different perspective, the senior developers seem to have a lot of age, right? Because they know the the uh technology a lot better, they have a lot of experience, and using the tech the AI will amplify their capabilities tremendously so much more that the junior ones probably like they're going to be in trouble. Um while that is true that the the amplified, but at the same time a lot of the more senior engineers actually sometimes get very fixated with uh the processes they're used to for their entire career. Let's say you're so used to doing this this way for your whole life, and you only know that these are such roles and these are things you need to do. Sometimes you cannot think out of of that, right? Uh and that becomes a disadvantage because you are only able to to do things the same things faster. Which is great, right? You can do things maybe 10 times faster. Amazing. But 10 times productivity. But the junior person might be looking at this and say, "Hey, you know, how don't I just totally skip all of these things? Because anyway, I don't know any of these things." And then I'll do it a completely different way that turns out to be 30 times faster. All right. And I noticed that I mean this is this is really what I see, right? I spoke to the team uh who had done uh certain things and they tell me that productivity is 10 times faster. I was like, "Wow, that's amazing." Then I spoke to this other team, two junior data scientists, right? One was I think 25, the other one was 28. And they're in fact they're data scientists, they're not even software engineers. And they started using this tool, they told me, "Oh, I think you were what I estimate is like I'm probably 30 times more effective now." So, they totally skipped the a lot of the things that what the more senior people would assume to be the correct thing and then they they went ahead. So, so I would say the the juniors shouldn't despair, but the seniors, I think if they adapt well, they can be a lot maybe they were 50 times faster, right? I don't know. So, so I say in general the um pure industry could adapt along with the industry. A lot of things are happening, but I actually very comforted um as like a person in the industry for a long time I would be very sad if things turn out badly, but I'm very comforted to see that things are turning out much more better than I thought. Actually right now, I worry about the other industries like my my lawyer friend, right? It's like, how are they going to deal with it? Because if I'm a the technologists understand how this work and they are able to adapt and use the technology differently. But the lawyer, would they know? I'm not sure. I don't know, right? It's maybe I'm only worried, but maybe that's not the it's not going to be a problem after all, right? But I I'm not sure. Yeah.
All right, with that, thank you Sashon. Thanks. Thanks, everyone. Thank you for listening to today's episode. If you are watching this on YouTube, please consider subscribing and turning on notifications for whenever new episodes are out. If you're on Spotify or Apple, it would help us greatly if you could leave a five-star review on those platforms. Once again, thank you for tuning in to the Five Reasons podcast.