End of year means prediction season. I’m generally sceptical of predictions—especially in technology, where hype outpaces reality. But I’m going to take a risk and share where I think healthcare AI is heading in 2026.

These aren’t wishes or hopes. They’re my genuine assessment of what’s likely to happen.

Prediction 1: Ambient Documentation Will Become Standard in Some Settings

AI-powered ambient documentation—systems that listen to clinical encounters and generate draft notes—will become routine in at least some Australian healthcare settings by the end of 2026.

The technology is mature enough. The value proposition (time savings for clinicians) is clear. The regulatory pathway (administrative support, not clinical decision-making) is relatively straightforward.

I expect adoption in:

• Private specialist practices (where clinician time is expensive)
• Telehealth services (where documentation from recorded calls is natural)
• Larger GP practices willing to invest in productivity tools

Hospital adoption will be slower due to complexity and change management challenges. But early adopters will move.

Prediction 2: Radiology AI Will Reach Routine Deployment in Major Networks

Multiple Australian radiology AI products are moving from pilot to production. By end of 2026, I expect radiology AI (particularly chest X-ray triage) to be routinely deployed across major public health networks and private radiology groups.

This won’t be universal—smaller practices and regional sites will lag. But AI-assisted radiology will shift from “innovative” to “standard practice” in metropolitan settings.

The driver is workforce pressure combined with maturing technology and clearer evidence.

Prediction 3: A Major AI Incident Will Make Headlines

At some point in 2026, I expect a significant AI-related clinical incident to receive public attention in Australia. An AI system will be implicated in a patient harm case, or a near-miss will prompt investigation.

This isn’t pessimism—it’s realism. As AI deployment expands, incidents become more likely. Not because AI is dangerous, but because any clinical system operating at scale will sometimes contribute to adverse events.

How the healthcare system and media respond to such an incident will shape public confidence in healthcare AI.

Prediction 4: TGA Will Clarify Continuous Learning AI Requirements

Current TGA guidance is less clear on continuously learning AI—systems that update based on new data—than on locked algorithms. This creates uncertainty for vendors and healthcare organisations.

I expect TGA to release clarified guidance in 2026 addressing:

• Requirements for predetermined change control plans
• Ongoing monitoring obligations
• Update notification and approval processes

This guidance will enable more sophisticated AI deployment while maintaining safety requirements.

Prediction 5: Investment in Healthcare AI Will Plateau

After several years of increasing investment, I expect healthcare AI investment (both venture capital and health service budgets) to plateau in 2026.

Why? The easy wins have been claimed. Demonstration projects are complete. Now comes the hard work of scaling, integration, and proving ROI. Investor enthusiasm cools when returns take longer than expected.

This doesn’t mean AI progress stops—it means focus shifts from new initiatives to making existing ones work.

Prediction 6: AI Governance Will Become a Board-Level Concern

Hospital and health service boards will pay more attention to AI governance in 2026. Drivers include:

• Incidents elsewhere that prompt governance questions
• Audit and risk committees asking about AI risk management
• Regulators signalling increased attention to AI governance
• Clinical governance standards evolving to address AI

Chief executives will need to demonstrate AI governance capability, not just AI deployment capability.

Prediction 7: Mental Health AI Will Face Pushback

The expansion of AI into mental health settings will face organised resistance from clinicians, consumers, and advocates concerned about:

• Quality and safety of AI therapy tools
• Privacy implications of mental health AI
• Appropriate role of AI in therapeutic relationships
• Risk of AI substituting for human care

I expect at least one high-profile campaign opposing specific mental health AI applications.

Prediction 8: Regional AI Disparities Will Widen

Despite good intentions, regional and rural healthcare will fall further behind metropolitan areas in AI adoption during 2026.

The causes:

• Infrastructure gaps that persist
• Workforce limitations that constrain implementation capability
• Volume economics that don’t work at regional scale
• Vendor focus on larger markets

This widening gap will receive policy attention but limited practical resolution within the year.

What Won’t Happen

A few things I don’t expect in 2026:

Autonomous clinical AI. AI systems that make clinical decisions without human oversight won’t be deployed at scale. Regulatory and professional barriers remain.

AI solving the workforce crisis. AI won’t meaningfully reduce healthcare workforce requirements. It might change some roles but won’t fill the fundamental gaps.

Universal AI adoption. Many healthcare organisations will still have minimal AI deployment by end of 2026. Adoption will remain uneven.

AI-driven revolution in primary care. Despite potential, primary care AI adoption will remain limited due to fragmentation, incentive structures, and change management challenges.

What I’m Less Certain About

Some areas where I genuinely don’t know what will happen:

MBS items for AI. Will MSAC approve any AI-specific MBS items in 2026? Possible, but depends on applications in the pipeline.

International AI expansion into Australia. Will major international healthcare AI companies establish significant Australian presence? Depends on commercial decisions I can’t predict.

Consumer AI healthcare use. Will consumers directly use AI for health purposes in significant numbers? The technology exists; the adoption dynamics are uncertain.

Why Predictions Matter

I share these predictions not because they’re certain—they’re not—but because thinking about the future shapes how we act in the present.

If you agree that ambient documentation is coming, should you be preparing now? If you think governance attention is increasing, is your governance framework ready? If regional disparities are widening, what should regional health services do differently?

Predictions are tools for planning, not prophecies.

I’ll revisit these at the end of 2026 and assess how wrong I was.

---

Dr. Rebecca Liu is a health informatics specialist and former Chief Clinical Information Officer. She advises healthcare organisations on clinical AI strategy and implementation.