Along with my data scientist colleagues, I’ve been leveraging advanced analytics to derive insights from data for Singapore clients for years. Back in 2013, I worked with the Singapore Government on its Safe City programme, using video and sensor analytics to support real-time decision-making, anomaly detection and predictive insights. And I’ve been involved in a whole series of cutting-edge projects in health and public service since then.  

During this time, moving from AI experimentation to implementation has yielded incredibly promising results here, helping to improve the day-to-day lives of citizens and make public service delivery more responsive and efficient. AI has truly arrived in Singapore, no question. But there’s much further left to go. 

Right now, there’s a widespread belief among government agencies that AI can only ever be 95% accurate. In other words, although it’ll get things right most of the time, one in 20 times it makes a mistake. That’s not good enough. After all, the consequences of AI misses remain significant, damaging public trust and limiting uptake of digital services.   

To break through this threshold and unleash the full potential of these technologies, we have to move the dial to 99% accuracy. This would support substantial changes in operations workflow, maximise manpower utilisation and transform the way citizen services are delivered.  

In a recent white paper, I explored the challenges agencies face in bridging this accuracy gap – and how to overcome them. The good news? The goal of 99% is well within reach. In fact, we’re seeing organisations achieve it in a whole range of areas. Take Google, for instance. It’s created an algorithm that can detect metastatic breast cancer with 99% accuracy. 

Of course, this level of accuracy is unlikely to happen at the start of an AI implementation. It’s in the nature of these technologies that they learn and get smarter over time, helped by improved datasets and thanks to human input that trains the machines better. 

But with the right strategic approach – built around human-machine collaboration – government agencies can position themselves to reach 99% accuracy. As I explain in the whitepaper, this means focusing on a few fundamentals. 

First, double down on building datasets that proxy realities. The more this happens, the better AI will get at generating accurate outcomes. This requires not just experimenting with the coolest algorithms, but concentrating instead on the more mundane tasks of data collation, preparation and labelling.  

Recognise that humans play a key role by identifying priority data and ways to improve it. So there has to be a focus on building user-friendly interfaces to promote human-human and human-machine feedback loops that assess the performance of algorithms and continuously feed those assessments back. 

Crucially too, scrutinise how algorithms are created. Building AI capabilities using open-source frameworks could create opportunities for learning and improvement, and better serve the need for transparency. 

There’s more detail and examples in the white paper. I urge you to take a look. There’s no doubt that government agencies which approach AI implementation as a journey of humans and machines – one where results happen over time – will be on track to harness the full power of AI. And that has to be good news for everyone in Singapore – agency workforces and citizens alike.  

Ng Wee Wei

Managing Director, Lead – Health & Public Service, ASEAN

Subscription Center
Subscribe to Accenture's Smart Nation Voices Blog Subscribe to Accenture's Smart Nation Voices Blog