Distinguished guests, colleagues and friends

1. A very good morning, and to all the friends and colleagues from overseas, a very warm welcome to Singapore. I'm very glad to see some good participation from many different countries, jurisdictions, all of us studying perhaps from a different place, slightly different background, but all here because of a common interest and passion, and that is really what is transforming lives through healthcare, leveraging technology.

2. We are entering a new phase of Artificial Intelligence (AI) in healthcare. The first wave of generative AI helped us generate. Draft a note. Summarise a record. Retrieve information. Answer a question. The next wave will increasingly help us act. Coordinate a workflow. Follow up with a patient. Monitor a care journey. Carry out tasks across different systems. We are moving from AI as assistant to AI as actor. And that changes the equation we need to ask.

3. For the last few years, we have asked: What can AI do for healthcare? Increasingly, we must ask a different question: What must healthcare become to take advantage of what AI can do? Because the next constraint on AI in healthcare is not what the technology can do. It is what our health systems are capable of doing with it.

From experimenting to scaling

4. We are seeing this challenge in Singapore. When generative AI emerged, we could have allowed individual hospitals and healthcare institutions to experiment independently. But then each institution would have to solve the same problems - infrastructure, security, access to models, testing and deployment, and many other similar challenges. So instead, our national HealthTech agency, Synapxe, developed Tandem: a common, secure GenAI platform across Singapore's public healthcare system. Healthcare professionals can explore GenAI, test ideas securely with domain-specific data, and, once validated, potentially deploy them across institutions.

5. Applications are already being deployed on this common platform. One example is Note Buddy, which can transcribe and summarise doctor-patient conversations, including across Singapore's four main languages. The benefit is simple: Less time taking notes. More attention on the patient. But the bigger lesson is what sits underneath. We are not simply deploying an AI application. We are building the capability for a health system to experiment, learn and scale. And that distinction matters.

6. We should not measure AI adoption by the number of tools we deploy. We should ask: What can our doctors, nurses and care teams now do better because AI is there? The value is not the AI itself. It is what people and organisations become capable of doing because of it.

From copilots to agents

7. And we are now preparing for the next step. AI is increasingly becoming agentic. Today, most AI systems wait for us to ask them something. But a new generation of AI agents can be given an objective and carry out a series of actions to achieve it. We are already exploring this in Singapore. Synapxe has developed AgentSea, a sector-wide agentic AI platform where our public healthcare professionals can build, deploy and manage AI agents. The response has been encouraging. Over 12,000 AI agents have been created by healthcare professionals as they explore and experiment with how agents can support their daily work.

8. Let me give you an example. One clinician built an AI agent to help him generate customised patient summaries ahead of consultations. The agent brings together the patient's background, clinical issues, care plans and outstanding follow-ups all in one place. The AI reduces the clinician's preparation time and helps him focus on the patient's needs. We are still in the early stages of exploring what AI agents can do for healthcare.

9. But imagine where the technology could eventually take us. Imagine an elderly patient with several chronic conditions who has just been discharged from hospital. Today, that patient may have medications to manage, appointments to make, tests to complete and several healthcare providers involved in their care. Now imagine an AI agent helping coordinate that journey. It checks that appointments have been made. It reminds the patient about medication and follow-up. It brings together relevant information. And it flags when an appointment, test or follow-up has been missed. Perhaps, one day, it detects signals that the patient's condition may be deteriorating. And when clinical judgement is required, it escalates to the care team.

10. The doctor remains responsible for the clinical decision. But the system around the doctor becomes much better at bringing the right information and actions together at the right time. This is not simply automating a task. This is redesigning a care journey. When AI can act, the unit of transformation shifts from individual tasks to entire workflows - and eventually to the entire care journey.

From recommending to acting

11. But this creates a new responsibility. When an AI system drafts a note, a healthcare professional can review it before anything happens. But when an AI in the system can act, the risk equation changes. We have to ask: What authority should the agent have? What should it never be allowed to do? What decisions must remain human? How do we know when something has gone wrong? Can we reconstruct what happened? And can we stop the system quickly when necessary? So as we move from AI that recommends to AI that acts, we must move from assuring outputs to assuring actions.

12. And this is something we have been thinking about in Singapore. The Ministry of Health recently updated our Artificial Intelligence in Healthcare Guidelines - AIHGle 2.0. It sets out responsibilities across the AI lifecycle - for developers to develop responsibly, healthcare institutions to deploy safely, and healthcare professionals to use AI wisely. That includes evaluating safety and performance, putting in place multidisciplinary governance, monitoring systems after deployment, maintaining incident-reporting processes, and ensuring that professional judgement remains with people.

13. Ladies and gentlemen, these principles become even more important as AI becomes agentic. The more agency we give to AI, the stronger the safeguards around those actions must become. This is not about choosing safety instead of innovation. Good assurance is what gives innovation potential to scale. Technological capability cannot be allowed to outpace our capability to govern it.

From adoption to outcomes

14. There is another transition we need to make. In the first phase of healthcare AI, we understandably focused on adoption. How many use cases? How many pilots? How many clinicians using AI? These measure activity. But they do not measure success. The question that matters is: Did it improve healthcare? Did it improve clinical outcomes? Did it give healthcare professionals more time for patients? Did it improve the patient experience? Did it allow care teams to intervene earlier? Did it make the health system more productive and more resilient? We therefore need to move from measuring AI adoption to measuring AI outcomes.

15. That is why I am especially pleased that today we will witness the MOU for the HIMSS AI Outcomes Framework initiative. This initiative brings together digitally mature healthcare institutions across Singapore, Korea and Taiwan to contribute real-world outcome data and expertise towards an evidence-based framework for measuring the impact and value of AI solutions. I commend the effort. This is an important step. Because beyond becoming better at deploying AI, we must become better at knowing what creates value. And once we can measure what works, we can learn from it and scale it.

The real bottleneck is capability

16. That brings me to the larger point. AI will continue to improve. Models will become more powerful. Agents will become more capable. And access to advanced AI technologies will increasingly become more widespread. But better AI does not automatically produce better health outcomes. Between an AI capability and a patient outcome, sits an entire health system: Our people. Our workflows. Our institutions. Our data and infrastructure. Our regulations. And importantly, the trust of our patients.

17. So let me also introduce another equation here this morning: Health Impact = AI × Health System Capability. AI creates possibilities. But it is the health-system capability that converts those possibilities into outcomes. And what does those capability require? First, integration. AI has to work inside real clinical and operational workflows, rather than sit outside them as another application. So that's integration. Second, redesign. If we simply add AI to today's processes, we automate parts of the old system. The bigger gains come when we ask how the work itself should change because new capabilities now exist. So we need to redesign our workflows and processes to make full use of the technology, and not just automate or introduce AI to the old ways of doing things.

18. Thirdly, learning. A capable health system must be able to experiment, measure what works, improve it and spread successful practices. We have to learn, scale, and keep improving.

19. And fourthly, assurance. As AI becomes more capable and autonomous, our ability to test, monitor and govern must grow alongside it.

20. That is what it means to build a health system capable of capturing the value of AI. For the past few years, much of our conversation has been about AI for healthcare. The next phase must be also about how to design healthcare for AI. Not designing healthcare around technology for its own sake. But redesigning our healthcare systems to harness AI and realise better health outcomes.

Make people more capable

21. And at the centre of that transformation must remain our people. AI will increasingly become better at automating execution. Documentation. Information retrieval. Data synthesis. Coordination. Many predictable parts of our work. But that should not mean investing less in people. As AI becomes more capable, we should invest more in the distinctly human capabilities. Judgement. Context. Empathy. Communication. Accountability.

22. If you are asking the question: How many healthcare jobs can AI replace? That is the wrong question. A much more ambitious question is: How much more capable can AI make every healthcare professional? Imagine a doctor who identifies deterioration early, before a patient becomes seriously unwell. A nurse who has the information she needs precisely when she needs it. A care team that spends less time navigating systems and more time exercising professional judgement. So our principle should be: Automate execution. Invest more in human judgement. Because ultimately, the measure of AI in healthcare should not be how intelligent our machines become. It should be whether they make healthcare professionals more capable in delivering care.

Beyond the hospital

23. And there is an even larger opportunity that we are exploring and focused on in Singapore. Most healthcare systems today are still fundamentally reactive. We wait until someone becomes sick. They enter the healthcare system. We diagnose. We treat. And hopefully, we help them recover. But AI gives us an opportunity to move progressively towards predictive, personalised and preventive health.

24. Singapore is already ramping up use in AI, in areas such as medical imaging to identifying patients at greater risk of chronic diseases and hospital readmission. Over time, increasingly capable AI could help us detect risk earlier, intervene sooner, support people between clinical encounters and coordinate care before health risks develop into serious illness. That could change not only how efficiently we deliver healthcare. It could change when healthcare begins. The greatest impact of AI may ultimately not be inside the hospital. It may be in helping more people avoid needing the hospital in the first place.

From capable health systems to capable societies

25. Healthcare is therefore a window into a much larger challenge. AI capability is advancing extraordinarily quickly. But technology alone will not determine which countries and societies benefit most. Our organisations must redesign themselves. Our workers must acquire new capabilities. Our institutions must govern new risks. And our societies must turn technological progress into better lives.

26. That is why I believe the next phase of the AI era will increasingly become a race of capabilities. Not simply who has the biggest and most frontier models. Not simply who has the most compute. But who can best absorb the new technologies, reorganise around them, govern them, and turn them into real-world outcomes. Healthcare, I believe, can be one of the most important places to demonstrate how.

Closing

27. So in conclusion, yes, we should continue pushing the frontier of AI in healthcare. But another crucial frontier has opened alongside it: The frontier of capability. Building health systems that can absorb rapidly advancing intelligence. Redesign care around it. Equip professionals to work differently. Learn what works and scale it. Preserve human judgement and accountability. And govern machines that increasingly act, not merely advise.

28. The next frontier of healthcare AI is therefore not simply more intelligent new machines. It is more capable health systems. Because: AI creates possibilities. Capability turns those possibilities into better health. And perhaps that is the larger lesson for the AI era. Societies that benefit most will not simply be those with the most powerful AI or technology. They will be those best able to turn technological possibility into human capability - and human capability into better lives.

29. So this is an important endeavour. I would like to thank the organising team for hosting this HIMSS26 APAC event with SingHealth. Thank you for all the hard work of bring everyone together, and HIMSS coming to Singapore- welcome back and co-hosting this important event with us

30. And all of you here, who took your time to be here with us today on this important topic. The possibilities are immense, and I'm sure you will discuss many of these topics today and tomorrow in this event. But I leave you with one final equation, it's Health Impact = AI × Health System Capability.

31. All of you are the most important stakeholders in building the technology. Thank you very much.