Insights
The Agent will see you now
- Date 26 Jun 2026
- Filed under Insights
Note: This article was first published by Pulse+IT. Read the original here.
Australian healthcare has passed the point of no return
The transformation did not arrive with fanfare. It unfolded incrementally, then rather suddenly became normal.
Patients and clinicians began experimenting with ChatGPT, Claude and Gemini, initially as adjuncts to care, replacing Dr Google, then gradually replacing what came next. Within a remarkably short period, these tools progressed from idle curiosity to utility, from utility to preference.
By 2025, generative AI had already begun reshaping the clinical encounter, expanding access, enhancing patient agency and introducing AI as a routine participant in care interactions.
In practice, the “front door” of healthcare shifted. It no longer resided in the general practice waiting room, nor even in a telehealth queue, it resided in a conversational interface available at any hour, responsive within seconds and increasingly capable of meeting both clinical and emotional needs.
They offered no resistance
The prevailing assumption had been that clinicians would contest the encroachment of AI. Instead, they adapted pragmatically.
Empirical evidence from BMJ Digital Health & AI demonstrated that general practitioners anticipated tangible operational gains, with 59% citing improvements in documentation and 56% in patient information capture alongside significant expectations of efficiency gains.
Concurrently, rigorous evaluations of leading models showed diagnostic accuracy exceeding 90% in common clinical scenarios, positioning AI not merely as a support tool but as a credible participant in structured reasoning workflows.
What emerged was not replacement but reconfiguration. Clinicians increasingly delegated administrative and routine cognitive tasks to AI agents, reserving their own expertise for ambiguity, nuance and complexity. The AI interface became, in effect, the operating system of care delivery.
Patient choice
Patients, for their part, did not await regulatory nor government endorsement.
Studies of AI-enabled care platforms reported satisfaction rates exceeding 86% and near-universal agreement on time savings, alongside markedly improved perceptions of access.
Yet these metrics, while impressive, fail to capture the deeper shift: expectation. Patients began to regard immediacy, continuity and coherence not as aspirational, but as baseline.
In contrast to the traditional model… episodic, fragmented and administratively burdensome; AI agent mediated care offered a continuous, integrated experience, collapsing triage, advice, diagnostics, prescribing and follow-up into a single conversational journey. The care that patients had always hoped in vain to receive.
Empathy, unexpectedly, became a machine capability, a hand to hold.
Perhaps the most disquieting and consequential development lay in the domain of empathy. A landmark study in JAMA Internal Medicine found that clinician evaluators preferred AI-generated responses 79% of the time, rating them superior in both quality and empathy over human clinicians.
This finding was not isolated. A systematic review in the Journal of Medical Internet Research concluded that large language models consistently demonstrated recognition of emotional cues and provision of supportive, empathetic responses, in some cases outperforming human comparators.
Further research confirmed that GPT-4 outputs could match or exceed clinician-authored communications in tone and clarity.
In retrospect, this was a critical inflection point. Healthcare had somewhat inadvertently, rendered empathy computable, reproducible and infinitely scalable.
Compute is capacity, “cloud” has become quaint
As AI agents became the primary interface, the limiting factor in healthcare patient flow shifted decisively. No longer constrained principally by workforce or physical infrastructure, the system became dependent on new metrics: Compute capacity, Token throughput, Network latency, Security architectures and Interoperability frameworks.
Healthcare, in effect, became a compute-bound system rather than a bed-bound one.
Geography no longer dictates access to healthcare
Australia’s structural challenges: distance, workforce maldistribution and uneven service provision have long shaped health outcomes. AI did not eliminate these disparities, it altered their contours by enabling low-bandwidth access to clinically informed guidance, continuous engagement independent of proximity and care coordination across fragmented service layers.
If one can diagnose a rash with a photo, get a telehealth Doctor to prescribe a cream, get the QR code sent to the Pharmacy and have it home delivered in an afternoon, why drive for hours to the hospital to wait more hours in the Emergency Department with no end in sight?
AI systems reduced the primacy of geography. Access became less a function of location and more a function of connectivity, literacy, and trust. Regional and remote communities finally have similar access to their metro cousins and agents will further streamline that service.
Where we invest shapes the system
For decades, health system investment has centred on capital-intensive physical infrastructure: shiny new hospitals and greater bed capacity with disregard for whether there was a skilled workforce available. Yet AI has fundamentally altered the economics.
Hospitals are geographically fixed, scale incrementally and expensively, depend on constrained workforce supply. AI platforms scale elastically with compute, operate continuously, deliver marginal cost efficiencies at scale.
Clinicians and patients have tolerated the teething challenges of hallucinations. This raises a legitimate and deliberately uncomfortable policy question: Might Australian health departments achieve greater system resilience and capacity by investing in sovereign, health-grade, hyperscaler data centres than in additional hospital construction?
A national, sovereign AI infrastructure could underpin consistent, high-quality care delivery, integrate diagnostics, prescribing and logistics into unified pathways in order to deliver capacity unconstrained by physical location keeping patients healthy at home.
The sins of the fathers are visited upon the children
AI agent first healthcare risks creating a two-speed system, in which those with digital access, literacy and trust in technology benefit disproportionately, while priority populations particularly Aboriginal and Torres Strait Islander communities, Culturally and Linguistically Diverse, rural populations without satellite connectivity and those experiencing socioeconomic disadvantage are left further behind.
Equity in this context must be engineered deliberately, not assumed. Careful consideration must be given to:
- universal, low-friction access to power, connectivity and agentic AI interfaces
- culturally and linguistically safe, comprehensible and locally relevant model training
- subsidised connectivity and device access for those unable to pay
- transparent governance and bias mitigation frameworks designed to avoid past mis-steps
- Measurement of outputs not inputs
Medicare funding for AI agent health access
The seats in waiting rooms are paid for by tax dollars. Over centuries, society has normalised waiting long periods of time to receive healthcare. If AI agents now constitute the functional entry point to care, access to them becomes a matter of public policy.
It is therefore reasonable to ask, should Medicare evolve to fund access to AI agents, perhaps through token-based entitlements as part of health and social care?
Such a model would treat AI clinical agent interactions as legitimate care events, allocate usage proportionate to need, ensure equitable access across the population and anchor delivery within regulated, sovereign infrastructures. In this paradigm, tokens become not metaphorically, but operationally the new unit of care.
The speed of change will never be slower
The front door of healthcare has already shifted from clinic to screen to clinical agent. ChatGPT, Claude and Gemini have not replaced clinicians, they have replaced the interface through which care is accessed, organised and experienced. The strategic question now facing Australian policymakers is not whether this transition will occur it already has.
The question is whether investment will continue to privilege an outdated model defined by buildings or pivot toward one defined by compute, connectivity and equitable access to intelligent systems.
In an AI agent mediated future, the defining determinant of health system performance may not be the number of hospital beds, it may simply be who can access the AI agent.
References
- BMJ (2025) – How generative AI affects patient agency
- BMJ Digital Health & AI (2025) – Generative artificial intelligence in medicine: survey of UK GPs
- JAMA Internal Medicine (2023) – ChatGPT vs physicians in patient responses
- Journal of Medical Internet Research (2024) – LLMs and empathy: systematic review
- Digital Health Journal (2025) – GPT-4 vs physician-written communications
- JMIR Formative Research (2024) – AI telehealth usability and patient satisfaction
- JAMIA Open (2025) – Comparative diagnostic performance of LLMs
- JAMA Network Open (2026) – LLM clinical reasoning performance
- Asian Journal of Probability and Statistics (2025) – Virtual healthcare assistants systematic review
