AI usage is inevitable, we need to prepare now

Rather than depriving radiologists of their job, AI is anticipated to take the load of scanning for abnormalities off them, allowing doctors to focus on interpreting the results instead. — IBM

I had the opportunity to represent the Malaysia Digital Economy Corporation (MDEC) at the 2025 Gitex Global in Dubai, the world’s largest tech expo.

The message was unmistakable: artificial intelligence (AI) is not due to replace humans – it is about augmenting us.

The “winners” will be the countries, firms and workers that harness it; the losers are those who are unable to adapt.

For Malaysia and the broader Asean region, this moment is an opportunity – not just to adopt AI, but to pivot whole sectors, including healthcare, onto a higher-productivity trajectory.

From novelty to infrastructure

Gitex showcased thousands of exhibitors, but the deeper narrative was about scale: national platforms, computing infrastructure, data pipelines, regulation – not simply chatbots.

Similarly, in Malaysia, the newly-launched National Artificial Intelligence Office (NAIO) is coordinating AI strategy, backed by a budget allocation and more than 50 public-sector pilots.

In Singapore, the AI Medical Imaging Platform is already live in public hospitals, offering an app store of imaging-AI models.

These developments underline a shift: AI is not just an innovation challenge, it is evolving into necessary infrastructure.

Across sectors, AI tends to automate tasks – especially repetitive ones – not necessarily entire jobs.

When done well, humans move to oversight, interpretation and higher-level decision-making.

As one speaker at Gitex put it: “AI won’t replace you, but someone using AI will.”

In manufacturing, for example, computer vision systems detect defects faster than the eye; in logistics, dynamic routing adapts in real time; in services, “copilots” summarise meetings and prefill documents.

The same pattern holds in healthcare: AI handles the routine, doctors handle the nuanced.

The (ideal) result?

Productivity goes up, but so do the demands on human judgement.

Fast-tracking diagnosis

In the health sector context, the stakes are especially high.

Radiology – once feared to be first in line for mass job cuts – is instead becoming a proving ground for AI-augmented workflows.

A recent review in Singapore shows that AI can optimise throughput, prioritise urgent cases and free radiologists for interpretation rather than rote reading.

In Malaysia, the Health Ministry is actively piloting AI diagnostic tools for cancer, tuberculosis and retinopathy, with the Health Technology Assessment Section evaluating safety and efficacy.

One pilot for diabetic retinopathy screening (known as DR.MATA) reportedly returns results in 30 seconds with accuracy above 90% in early tests.

Translated into practical terms: fewer scans languish unreviewed, more patients get earlier detection, and clinicians spend less time on routine checks and more time on care.

It’s not all roses

For all this promise, three critical caveats merit attention, especially in the Malaysian/Asean context.

> Data quality matters

AI is only as strong as its inputs (rubbish in, rubbish out).

An algorithm trained on homogeneous data may fail in real-world Malaysian hospitals with legacy scanners, diverse patient populations and inconsistent workflows.

If the pipeline is weak, the system will output rubbish.

Such factors are especially acute in South-East Asia where resources, equipment and data standards vary.

> Liability and accountability remain unresolved

In the clinic, when AI misses a tumour or flags a false positive, who is accountable?

The software developer? The hospital? The clinician who trusted the output?

Clear regulatory frameworks are still emerging.

Singapore and Malaysia are advancing guidelines, but full-blown law is still catching up.

Without clarity on liability, adoption may stall.

> Real-world deployment still demands care

Pilots are promising, but scaling is different.

Maintenance, model drift, integration with hospital workflows, training of staff, cybersecurity and ethics all matter.

A model good in one setting may underperform elsewhere.

The novelty of AI must be backed by disciplined and consistent implementation.

The right policies

Asean policymakers and hospital leaders must focus on three key priorities.

First, they should build the “plumbing”, not just focus on apps.

Malaysia’s sovereign data cloud collaboration marks a strong signal of intent, but health systems must also invest in interoperable data, shared standards and secure computing infrastructure.

Without these foundations, every promising AI pilot risks remaining an isolated experiment rather than becoming a system-wide solution.

Second, training the workforce is essential.

The idea of “every clinician with a copilot” should become a mantra across the region.

Radiologists must not only know how to interpret AI outputs, but also when to override them.

In Singapore, radiographers already anticipate working alongside AI, while in Malaysia, closing the AI talent gap has become a national priority.

Finally, regulators must strike a balance: ensuring safety without stifling innovation.

Malaysia’s voluntary AI Governance and Ethics (AIGE) guidance is a useful starting point, but its non-binding nature leaves gaps, particularly around automated decision-making.

Regulators and health ministries should create sandboxes for safe experimentation and demand strong clinical evidence before and after deployment to maintain public trust.

Radiology provides a clear playbook for Asean’s healthcare systems.

Hospitals should first target high-impact areas where speed and throughput save lives.

These include stroke cases, mammography, and screening for diabetic retinopathy and tuberculosis.

Workflows should then be redesigned so that algorithms handle the pre-reading and prioritisation, while clinicians retain the final interpretation and responsibility for communication.

Finally, results must be measured rigorously.

Turnaround times, diagnostic accuracy across patient subgroups, model drift, clinician workload and patient outcomes all need continuous monitoring.

Data-driven evaluation not only strengthens safety and trust, but also builds a solid business case for budget allocation and scaling.

Race against time

The risk for Malaysia and Asean is not that AI will take all jobs; it’s that those without copilots will be overtaken by those with them.

The digital economy isn’t just about carrying on as before with an app or digital interface.

It’s about rethinking how we serve, how we scale expertise and how we deliver value.

In healthcare, the stakes are high; it could mean better access, earlier diagnosis and less burnout.

The regional message is clear: if you build capable people, collect clean data and have trustworthy systems, you’ll reap the productivity gains.

If you skip the groundwork and chase hype, you risk disillusionment and wasted opportunity.

AI is not about just technology; it is a choice about how we work, what we prioritise and who we become.

Malaysia and its neighbours can lead this shift if we treat AI not as a threat, but as a catalyst.

The next decade in South-East Asia will not be defined by the most futuristic algorithm, but by the most adaptive systems, the most ethically-grounded organisations and the most resilient workforce.

In that race, the smart play isn’t to replace humans, it is to empower us.

Dr Helmy Haja Mydin is a consultant respiratory physician and a member of MDEC’s Board of Directors. For further information, email [e-mail]. The information provided is for educational and communication purposes only. The Star does not give any warranty on accuracy, completeness, functionality, usefulness or other assurances as to the content appearing in this column. The Star disclaims all responsibility for any losses, damage to property or personal injury suffered directly or indirectly from reliance on such information.