Singapore

When we talk about digital government and AI adoption in the public sector, we don’t can’t get very far without somebody mentioning Singapore. Singapore has been leading the way in using AI in public services for a number of years, including via their flagship Smart Nation programme, which even features in a National museum with an interactive exhibition!

Earlier this year, some members of the i.AI team visited colleagues in Singapore’s GovTech to find out more about their approach, and what we can learn. In this blog we share a few insights that we found helpful to shape our thinking about how AI adoption can be done well. 

This is not an exhaustive list of everything happening in Singapore, and does not cover any of the formal relationships that may exist between the UK and Singapore Government; just some reflections from the i.AI team after an enjoyable and enlightening visit!

Singapore places huge value on technical capabilities in government

• Singapore government prioritises internal technical builds and technical capability very highly.
• Their equivalent of GDS (GovTech) numbers in the thousands and includes a specific division for data science and AI, and is perm sec led (with a second perm sec). This is a very large team for any government, but particularly for a country with a population 10x smaller than the UK.
• As part of this they have really well resourced data and cloud engineering capacity
• It is seen as a prestigious job and salaries reflect this. Performance related pay (up to 50%) is often used to incentivise innovation and delivery. 

Singapore’s AI adoption is driving impact through encouraging tech-civil servants to innovate

• There is a strong focus on innovation and encouraging the large community of technical developers to try out new ideas to adopt AI. This has resulted in a plethora of chatbots to support citizens in engaging with the government. 
• The community of people working on AI in government is actively nurtured, including through a regular ‘AI Wednesdays’ meet up with technical demos, and an AI Practice blog where people can share their projects. 

One example that really stood out to us is LaunchPad. This is an engaging and open way for teams across government to share ideas and develop applications of AI. In some ways, this is similar to the work that our sister team 10DS do with Evidence House and their Hackathons - we love to see it!

LaunchPad

The Launchpad operates within GovTech, fostering innovation through ideathons and engaging with various government agencies. Over 500 ideas have been collected, with four agencies currently engaged and more in the pipeline. Target audience includes transformation teams and public officers, with a six-month timeline comprising a launch, sprints, and a final release. A web page facilitates idea submissions, leveraging LLMs for detail expansion, community upvoting, and direct links to prototypes.

They are placing a high focus on interoperable solutions for government

• The team in Singapore are good at making scalable solutions to repeated pain points, which are widely adopted; and successfully drive interoperability.
• Central investment in pan-government authentication mechanisms has made the rollout of new AI and data platforms to all government departments rapid. 
• We heard about enTRUST, their data exchange which went live earlier this year, having been inspired by the work at HDR UK and others on trusted research environments for healthcare data in the UK. They “inner source” their work within Government so that other public servants can re-use code.  
• Part of the reason for this success could be the approach taken to data sharing agreements, which are more prominent both across public services, and with the private sector. This is interesting food for thought, in particular in the context of the Government’s plans for a National Data Library and other data sharing initiatives.

Where do we differ in approach?

Whilst there are some brilliant things that the UK stands to learn from countries like Singapore, there are also some areas where the UK’s approach has delivered strong results. 

Fundamentally, the approach taken to adopt AI in UK government is integrated into the policy and delivery system. This gives us two distinct advantages: 

1. We can pivot to work on government priorities quickly, because we are plugged into central and mission-oriented decisions in real time.
2. Our small size and agile approach means we can collaborate more naturally and work directly with departments to co-deliver products that are integrated into their ways of working. 

This has allowed us in the incubator to quickly stand up lots of projects which take recent developments in AI and target them to solve real delivery issues facing public services, be that triaging complex information, matching supply and demand, or providing better customer services (see our taxonomy blog for more information about how we think about this). 

We are actively building trust through a proactive public presence, operating under a banner of “radical transparency, engaging closely with industry, working as blended teams with Departments, and releasing information and codebase for all products where there are no national security or business risks. This enables us to work more closely with industry and drive wider adoption. The 10DS team pioneered this model, their open source data sharing solution, rAPId, is used in 14 departments, at least 3 large companies, and has over 200,000 public downloads.

So what?

Well, of course context matters, and the UK cannot simply adopt some of the things that Singapore are doing, but there is some excellent food for thought about how good ideas can turn into real projects, and what it takes to build a community who are committed to delivering change. 

It’s also encouraging to hear that we share many of the same views on what matters for government adoption of AI - in i.AI we are committed to sharing our work in the public interest, and open source our work wherever it’s appropriate to do so!