The great normalisation. How AI is settling into the heart of global health.

---

The hype cycle around artificial intelligence has reached full saturation, and nowhere is that more evident than in global health. To the outside world, it can look like we are in the middle of a speculative gold rush: ballooning valuations, convoluted financing structures reminiscent of pre-GFC exuberance, and a relentless push to secure every last chip on the planet.

But if you look past the headlines about mega-rounds and trillion-dollar infrastructure buildouts, something more enduring is taking shape. AI is no longer the shiny new object. It is becoming part of the digital fabric, maturing into a genuine general-purpose technology; the steam engine of this century.

Technological revolutions typically follow a familiar pattern: early excitement, overheated expectations, correction, and finally, normalisation. Right now, we are edging out of exuberance and into the early stages of correction, but normalisation is clearly on the horizon.

Healthcare leading

Healthcare has a reputation for being slow to modernise, yet it has quietly surged ahead in AI adoption, deploying these tools at more than twice the rate of the wider economy. Frontline health organisations are leading this shift, with adoption rates now at 27 percent. For them, this isn't about chasing novelty - it’s about survival. Workforce shortages, clinician burnout, and mounting administrative pressure have created conditions where AI has become essential infrastructure.

The first wave of value is showing up in administrative and assistive use cases. Ambient clinical documentation - AI that sits in the consult room, listens, and drafts the clinical note - has become the standout category, pulling in roughly $600 million globally in 2025.

In New Zealand, clinicians at Tāmaki Health describe these systems as transformative, allowing them to focus again on the human side of medicine instead of being full-time typists. It is a pragmatic, human-in-the-loop model: AI handles the rote work, and clinicians remain fully accountable for the final output.

Bubble and the bedrock

We do need to be honest about the underlying economics. Much of the current AI expansion rests on fragile financial scaffolding, including significant borrowing by hyperscale providers to fund data centres and silicon. Some analysts see worrying parallels with the leverage patterns that preceded the 2008 crash. Add to that the fact that roughly 95 percent of generative AI pilots are failing to scale, and it’s understandable that many are calling this the "Trough of Disillusionment" after Gartner.

But history is instructive. The dot-com collapse wiped out a generation of companies but left behind the infrastructure that the modern internet runs on. The same dynamic is emerging now. The capital being poured into 2025’s AI boom is creating the compute backbone we will rely on for decades. We are in the midst of a silicon supercycle, where purpose-built hardware - Neural Processing Units and Digital In-Memory Compute - is taking over from general-purpose chips to push past energy and cost constraints. Valuations may correct, but the capabilities will remain.

The convergence

The real story is the convergence of AI with both digital workflows and the physical world. We are moving beyond conversational interfaces toward what many are calling Agentic AI: systems that can organise, plan, and action work with a high degree of autonomy. Nearly every development team on the planet is now experimenting with agents, and they are beginning to function less like tools and more like digital colleagues.

This convergence is extending into physical environments as well. Robotics infused with AI - what some call “Physical AI” - is now optimising logistics, supply chains, and hospital operations. Australian hospitals, for example, have automated up to 70 percent of consumable stock ordering. The line between digital capability and operational reality is dissolving.

Importance of governance

One of the clearest indicators that a technology is maturing is the arrival of meaningful regulation. The era of unfettered experimentation is closing. The EU’s AI Act has set a global benchmark by defining a risk-based regulatory framework, banning practices like social scoring and tightly controlling high-risk medical applications. California’s SB 53 has followed with requirements for safety frameworks governing frontier models.

Closer to home, the focus is shifting toward sovereign AI - ensuring data is held locally, governed locally, and reflects local cultural values, including Māori data sovereignty. Medical colleges are issuing practical guidance on the use of AI in care, emphasising transparency and informed consent. Rather than hindering innovation, these guardrails are creating the clarity needed to embed AI into enterprise-scale healthcare settings.

Quietly optimistic

Every major technology that reshaped society - the printing press, electricity, the internet - went through its own period of turbulence before becoming ordinary and indispensable. AI is tracking along that same trajectory. Over the next decade, it will increasingly fade into the background, becoming part of the plumbing of health systems.

By 2030, we may not talk about “AI in healthcare” at all. We will simply talk about care. Whether it is supporting older adults to live independently, enabling rural clinicians to access world-class diagnostics, or streamlining the complexity of modern health services, the convergence of AI with digital and physical systems promises a more humane, efficient, and equitable sector.

The bubble will wobble, as bubbles do. But the direction of travel is clear, and the progress is real. We will figure it out.