SPEECH BY MR ONG YE KUNG, MINISTER FOR HEALTH AND COORDINATING MINISTER FOR SOCIAL POLICIES, AT NCS IMPACT 2026, 9 JULY 2026

Distinguished guests

Friends

Ladies and gentlemen

1.               Thank you for inviting me today. I have a very heavy speaking schedule in July. So when my office received the invitation from NCS to speak at this conference, I hesitated to accept even though I knew it was an important event.

2.               Fortunately, now with Artificial Intelligence (AI), it does not take very long to write a speech. But unfortunately, I do not use AI to write a speech. I still believe, and my staff still believe, we will still craft our speeches. We use AI sparingly in preparing speeches – mostly to fact-check, get extra facts, some copywriting. So this speech is very much like a Swiss watch or a song written and sung by a songwriter, or a thick broth of ramen cooked overnight – it is crafted with love, but with a bit of help from machines.

Impact of AI on Jobs

3.               My insistence on crafting my speeches is relevant to the current debate on the extent of AI's impact on jobs and the economy. We have been discussing the topic, in my view, in generalities, and I think it is time for commentaries on AI to be more specific, more detailed. What does it really mean?

4.               Yes, solopreneurs are now possible. We have heard those stories. But many of us are not solopreneurs, nor do we serve or compete with solopreneurs, so it does not quite affect us. Yes, AI can now read diagnostic scans at hospitals. But at the Ministry of Health (MOH), we still require the clinician to review the AI interpretation and provide a diagnosis, so clinicians remain essential. And you ask the clinician, is AI affecting your reading of X-rays and diagnostics? They said not really, but I heard the hype. Yes, AI can now generate a movie. But I think most of us still prefer a movie shot with real scenery, real actors and actresses, real Teochew language.

5.               The impact of AI therefore depends on the industry, how demand and market forces evolve, the behaviour of companies, consumer tastes, the exact nature of jobs, and also the action of the individual worker. So the first half of my speech today, I will try to go one level deeper into the details on the impact of AI on industries and jobs. And I’m glad today we have so many union friends with us today. So I hope I make sense to them.

6.               First, there will be industries that will grow regardless, where the job will grow regardless, because of powerful driving forces. For example, employment in many parts of the finance and banking industries continue to register growth, this is driven by Singapore as an important, international financial hub. That can’t be affected by AI yet. Climate change and the desire for greater energy security are likely to see expansion of jobs in the energy sector. AI-related sectors such as semiconductor manufacturing are also seeing job growth.

7.               A major area of employment growth is the health and social care sectors. It is driven by demographic changes. For these sectors, we are in fact hoping, and I see some demonstrations outside, we are hoping that AI and automation will help mitigate our manpower shortages.

8.               Second category, there will likely be, I think, a big group of industries where jobs will hold but there will be many opportunities to improve things, embrace productivity, and do things better. We do not have to look far to find these examples. They are all around us. They are in schools, where teachers are still needed but they are considering ways on how to use AI to educate our children better; in the building industry, where technological tools can help design structures, organise resources, order materials and stage project plans; in the service and hospitality sectors, which will continue to depend greatly on human interaction, that can be aided but not replaced by AI. Similarly, numerous small and medium-sized firms in manufacturing, in services, will likely continue to thrive on human relationships and networks.

9.               For these industries and companies, market demand matters. Consumer taste matters. If this audience no longer values a speech crafted by a public official or a Minister like me, there will basically be no more need for speechwriters in ministries, and next time you don’t have to invite me, I don’t have to turn up. But I hope that will not happen. Because that is not what you require. You want to hear the Minister speak. Similarly, if consumers prefer factory-assembled watches and eat microwaveable noodles, there will be no Swiss watches and no Japanese ramen.

10.           It is my firm belief that consumers will continue to value mastery and craft. Authentic human creation done by hands, by your brain, and this will limit the extent of AI diffusion. I was in Japan recently, I asked about anime. They said there is some use of AI in anime, but consumers really push back. They want anime to be drawn by artists.

11.           And finally, there are jobs, or more like tasks, that can be substituted by AI. Routine and process-intensive work – data collation, report preparation, is clearly under threat. Drivers are worried because of the possibility of AI-driven autonomous vehicles. The concerns of those affected must be addressed. Governments, unions, employers, need to help affected workers make the adjustments.

12.           The Singapore Government has always emphasised the importance of an active labour market policy. What does that mean? It means we emphasised lifelong learning and also SkillsFuture. We now have a dedicated agency, the Skills and Workforce Development Agency (SWDA), to oversee this area of work.

13.           The SWDA will set out to do many things – they will gather labour market intelligence, strengthen training, career guidance and employment facilitation capabilities, raise HR standards, so on and so forth. But amidst all of its activities, at its core, it must remember – its mission is to bring opportunities, training, skills and people into close alignment. This means you make quality skills training programmes available and affordable, enrol workers who need training into the right courses, persuade employers to keep an open mind, and then facilitate placements and mitigate job uncertainty. You must get the workers, get the training, train the workers, place them into jobs. That’s the core of SWDA’s mission in the current economy that we are in.

14.           For certain jobs under threat, governments may need to decide whether regulations and guardrails are necessary. Autonomous vehicles illustrate this well. In San Francisco and other cities, I’ve not seen it, but I heard Waymo vehicles operate without a driver; and when the system encounters a glitch, the car stalls in the middle of the road, causing a big jam. Friends of mine living there say sometimes people will play a prank and jump into the path of a Waymo vehicle. The AI will make it stop, and they just want to test to make sure that it does that. This is not encouraged in Singapore. In China, however, I have ridden in autonomous vehicles many times. But each time I rode one in China, I realised the local regulations require a safety officer to sit behind the wheel, ready to intervene in an emergency. So there is no driver, but it is replaced by a safety officer who knows how to drive. So in fact, long before AI, aircrafts could already fly largely on autopilot, but international aviation regulations still require a full pilot crew.

15.           In safety-critical systems, human oversight remains essential. In sectors where human trust and empathy are paramount, work will still need to be primarily carried out by humans. These are the core considerations of using AI in healthcare, and I will cover this in the second half of my speech.

AI in Healthcare

16.           Good use of AI is not just about having the best AI tools. There need to be prerequisites. Otherwise, it is like having a state-of-the-art home appliance, say a very smart TV, but so what; you have no electrical sockets. There are no prerequisites to make it work. So what are these prerequisites? I think there are three – a strong digital operating environment, good quality data, and a sound policy and organising structure around the use of AI.

17.           First, the digital operating environment. After many years of investments, MOH is now in the final phases of replacing numerous legacy IT systems that do not talk to one another, with more capable and integrated systems covering the entire public healthcare sector. This has been a herculean task done over many years.

18.           This includes adopting a common Electronic Medical Record (EMR) system by 2028, which is the heart of our clinical operations, across all three public healthcare clusters. We are also adopting national IT systems for core functions such as billing, pharmacy operations, patient referrals across community settings and supply chain management. We will have an enhanced HealthHub application, which will enable residents to digitally make appointments, access selected medical records, pay bills online, across all three healthcare clusters. Now, you need one for each cluster. In future, HealthHub takes over everything.

19.           Second, as our operating environment becomes more digitally driven, we need to generate good quality data. We are used to the idea of a finance department, you have an human resource department, you look after money, look after people. Now, equally indispensable is a department looking after data, to consciously organise data used for research, analysis and operational improvement. So you will have a CFO, you have a CHRO, you also need a CDO – a data officer.

20.           One major initiative in data architectural development in healthcare is the National Electronic Health Record (NEHR) system. This allows patients' key health information to be shared securely across public and private healthcare providers. It facilitates a smooth transition across care providers and reduces unnecessary repeated tests. And I really thank our private hospitals for supporting this, even before the law came into effect.

21.           We established TRUST, our national data governance framework and analytics platform. This gives public researchers and selected industry partners secure access to nearly 50 anonymised health and health-related datasets, with safeguards for data security and privacy.

22.           We also built HEALIX, a unified and secure cloud-based platform that brings together healthcare data, analytics tools, and AI capabilities. In other words, HEALIX is like an innovation factory, enabling public healthcare institutions to develop, train, test and deploy AI tools and solutions.

Adopting AI Soundly

23.           Hence, when LLMs arrived, heralding a new era of AI, we fortunately had the infrastructure largely in place. What we need is the last prerequisite, which is a sound approach to adopting AI.

24.           Healthcare, fortunately, is a well-regulated sector, so we are not at risk of being overwhelmed by AI technology. This gives us the time and space to consider carefully how best to proceed with AI. AI in healthcare should never be like the proverbial hammer looking for nails. A solution looking for problems. Instead, we take a use case approach. Ask ourselves, what are the problems we face, what are the areas of improvement, and how does AI technology help us overcome them?

25.           Most use cases are ground-up initiatives. When they are effective and impactful, MOH will try to scale them across the system. Today, we are pursuing more than ten such system-wide use cases. They range from automating medical note-taking and clinical coding, predicting outpatient attendance to facilitate hospital operations, assisting radiologists in reading diagnostic scans, and identifying individuals with a high risk of developing chronic diseases so that we can take preventive steps earlier.

26.           For each of these use cases, we consult, deliberate and decide how best to implement it. For example, in diagnostics, we know that AI is capable of detecting even the slightest anomaly. An anomaly that if you show your clinicians, sometimes they say, “I don’t know what to do with it. It’s such a small thing. Do you want a biopsy? Does it make sense?” So rather than employing it indiscriminately just because we can, we tune it to highlight only what we know, and where there are clear treatment pathways and follow-ups.

27.           AI diagnostics should also not fully replace radiologists. This is partly to maintain clinical skills and partly to ensure human judgement remains relevant when it comes to caring for another human being. Hence, for mammogram scans, we decided that AI can substitute one of two human readers.

28.           Another promising initiative is a significant project that originated ground up, and it is called SIMFONI - the Singapore Medical Foundation AI Model, it is a programme under the Consortium for Clinical Research and Innovation, Singapore (CRIS), which is part of the MOH family.

29.           SIMFONI will build AI models that could serve as trusted companions to our clinicians. They can suggest possible diagnoses, treatment pathways and next steps, supporting but not replacing clinical decision-making. Such models already exist, but they were trained using patient data and medical guidelines in other countries. In other words, they have not gone to our local medical school. SIMFONI would have gone to our local medical school. Their models will be developed and trained using Singapore's clinical practice guidelines and local clinical data, contextualised to our population and eventually integrated into our clinical workflows.

30.           We will start with a few focus areas – cardiometabolic diseases and eye diseases; in primary and specialist care settings. These are important use cases, and we will deploy them system-wide throughout the public healthcare system when ready.

Closing

31.           Finally, I thank NCS for the invitation today. NCS is more than a vendor or a technological provider. It is a long-term partner in national digital transformation, with deep roots in the Singapore healthcare IT ecosystem and is well-placed to support our current phase of development, from system integration and cybersecurity to AI deployment and operational resilience.

32.           We hope to grow and deepen this partnership with you – one that supports NCS's success while delivering better health outcomes in a more sustainable healthcare system.

33.           But at a deeper level, the public and private sectors must navigate this path of AI evolution together. We are at a technological inflection point, learning to work with and coexist with AI. We cannot charge ahead, driven solely by commercial considerations, even as recursive AI systems gain self-reinforcing intelligence, agency and influence. Otherwise, I can’t help but feel the machines just seem wiser than their makers.

34.           We must be wiser and more humanistic and practical. We must decide deliberately where to embrace AI, where to rein it in, and where human judgement and effort must prevail. We can only achieve this by Government working together with industry partners. I wish you a successful event. Thank you.