The interview was conducted by Ula Rutkowska (Senior Researcher, Apolitical) and edited by Christina Obolenskaya (MSc in International History, LSE and Communications Intern, Apolitical). Sau Sheong Chang was named on Apolitical’s Government AI 100 2025.

Singapore is often cited as a global leader in digital government, but its journey to the forefront of GovTech has been decades in the making. A shift in the mid-2010s saw a renewed focus on building in-house expertise, strengthening engineering capabilities and accelerating digital transformation.

Apolitical spoke with Sau Sheong Chang, Chief Technology Officer at GovTech, about Singapore’s approach to AI, innovation and modernising government systems. He shares insights into how GovTech is fostering AI adoption across agencies, building central platforms to streamline public service delivery, and tackling the challenge of ageing digital infrastructure. We also discuss the future of digital transformation in Singapore and what milestones lie ahead in the next five years.

What role has GovTech played in driving Singapore’s digital transformation success?

GovTech, the Government Technology Agency of Singapore, has been around for quite a while—over 40 years. It began in 1981, though not as GovTech at the time. It started as the National Computer Board, which was set up to introduce policies and governance around how the government uses computers and computer systems.

Since then, it has gone through several iterations, following industry trends. From the 1990s to the early 2010s, and even into the early 2010s, there was a lot of outsourcing to vendors, which led to a loss of internal skill sets—the ability to be more technical, to understand technology, and to use it to drive strategy.

Around 2014–2015, there was a shift towards bringing technology expertise back into government, and in 2016 GovTech was formed. From that point, there was a major push to build engineering capabilities within the government. COVID-19 played a big role in accelerating digitalisation. QR codes have been around for a long time, but adoption was slow. Almost overnight, because of contact tracing, everyone had to scan QR codes before entering buildings, which made it a norm — helping accelerate our digitalisation journey.

From an outside perspective, it may look like the Singapore government is always at the forefront of innovation, but like any large organisation that has been around for a long time, there have been challenges along the way.

What innovations have you introduced during your time at GovTech Singapore?

As CTO, I joined two years ago—my two-year anniversary will actually be this February. Before this, I spent my career in the private sector. I was brought in to drive the adoption of more technology, and one of the key changes I introduced was separating efforts to build products from efforts to build capabilities and technology.

I set up two organisations within GovTech. One is the Government Digital Products unit, which develops central products used by multiple agencies and provides the infrastructure for additional products built within agencies. This is very product-focused and includes engineering products, platform products and agency-focused products.

The second organisation I set up is the Government Technology Office. In the commercial world, this would be similar to a CTO office. It brings together various practices focused on core technology disciplines like software engineering, product management, user experience design, AI and data engineering. There are other areas as well, but the goal is to continually strengthen our technology capabilities so that we don’t fall behind the commercial world.

How do you see AI being used most effectively in government operations and public services?

Our strategy for AI deployment is about encouraging widespread adoption across agencies—what you might call a “thousand flowers bloom” approach. We actively get agencies excited about AI and enable them to explore its use through workshops, events, training and hackathons. Over the past couple of years, this has led to a lot of interesting developments.

At the same time, we are also building central platforms to support common AI use cases. One of the most common use cases is generating text in response to certain inputs. For example, when a citizen writes in with a question—whether about a piece of legislation, a policy, or even general operational matters—it usually takes a public officer significant time to draft a response. If there’s a high volume of inquiries, response times can be slow.

To address this, we developed a platform that helps public officers by generating a template response based on the query. The officer can then tweak it as needed, but this eliminates a lot of the tedious work and greatly improves response times. We piloted this with a few agencies last year, and now it’s in full production in several systems.

There are two key aspects: first, education—through the Government Technology Office, we ensure agencies have the capabilities to use AI effectively. Second, we are building central AI platforms that multiple agencies can use so they don’t need to reinvent the wheel every time.

With AI capabilities, are these meant to be general AI skills that all government employees should have, or is it more about building AI into government services and products?

For GovTech it’s more about developing services that have AI built into them to improve government operations. That’s because a key part of GovTech’s mission is to be the software builder and operator for central platforms across the government.

We enable agencies to develop their own systems and products based on their needs, but many agencies lack the necessary expertise, manpower, or even fundamental knowledge. That’s why we also help other agencies build their own AI capabilities for the general public officer. We run workshops, events, and hackathons, covering a broad spectrum. Some sessions are designed for those who know nothing about AI—some even fear AI will take their jobs—so we work to demystify the technology. One of the popular events we run is called Prompt Royale, which is a competition designed to have a gameshow-like format to let public officers have some fun and put them at ease even as they learn new skills like prompt engineering.

At the other end of the spectrum, we run hackathons where public officers propose ideas, and we provide engineers to help them develop pilots or proofs of concept. Once a project reaches a certain stage, we hand it back to the agency. They can then choose to continue developing it in-house, bring in their own teams, or engage a vendor, with our support during the transition. Ultimately, we can’t build everything for everyone, so there needs to be a structured handover process. GovTech focuses on central platform products that serve multiple agencies.

What were some of the biggest challenges you ran into when taking AI projects from the pilot phase to full implementation?

Manpower was the biggest—there simply weren’t enough people with the right skills. In the early stages, there was also a lack of awareness and understanding of AI’s possibilities.

Another challenge was that many agencies came up with the same use cases. At hackathons, we’d often have 15 teams, 12 of whom would be working on document summarisation. Because of this, we developed a central platform called AI Bots, which allows agencies to upload documents and generate AI-driven summaries or answer questions about the content. This eliminated redundant efforts and allowed us to move on to more innovative applications.

What is your vision for the future of digital transformation in Singapore? What milestones do you hope to achieve in the next five years?

One of the main reasons I was brought in was to address a critical issue—technology modernisation. AI became a big focus along the way, but originally, the key concern was that many of our systems were getting old.

Some of these systems were built on mainframes decades ago, and while some have been modernised, they are now 10, 15, or even 20 years old. Many of these systems are highly vendor-dependent, making updates difficult and expensive. In some cases, the machines themselves are becoming obsolete, and if agencies go back to the vendor for updates, the costs can be sky-high.

This is an issue across industries, but it’s particularly acute in government because agencies often lack engineering expertise. If you have hundreds of agencies, each with dozens of systems, you’re dealing with thousands of ageing systems. Not all need to be modernised, but a significant number do.