Agentic AI’s first job in Australian healthcare is fixing the fragmented patient journey

Australian GPs rank health system fragmentation and communication issues among their leading concerns for the future, according to the RACGP’s 2025 Health of the Nation report. The practical solution is to make the systems general practices already run work together, with agentic AI as the orchestration layer between them, before any of those systems are replaced.

The urgency is capacity. The same report found one in three GPs intends to stop practising within five years and 63 per cent are considering reducing their hours, while consultations now run to almost 20 minutes and nine in 10 GPs regularly manage conditions that would once have gone to a non-GP specialist.

For a practice, that is a business issue as much as a clinical one. Every hour a GP or nurse spends chasing a referral or reconciling correspondence is unfunded time in a model where Medicare rebates have failed to keep pace with the cost of care. Nearly seven in 10 GPs report burnout, and most are dissatisfied with their administrative burden.

A decade of digitisation has given general practices a growing mix of clinical systems, referral platforms, pathology providers, patient engagement tools and, increasingly, AI-enabled applications. Yet clinicians and practice teams still spend significant time chasing referrals, managing follow-ups, reviewing correspondence and coordinating care across systems that operate in isolation.

The Productivity Commission put it plainly in its research paper on digital technology in healthcare. Despite major investment in My Health Record, patient data is still fragmented across systems maintained by individual providers, and information sharing between them still breaks down, including from hospital back to the GP clinic.

That coordination gap is where agentic AI is likely to deliver value first in primary care. What follows draws on work by two collaborating organizations: FPT, a global IT services company that builds and integrates AI systems for enterprise healthcare, and Foundry Health, a clinician-led healthcare strategy advisory firm with offices in Sydney and Melbourne.

Orchestration comes before replacement

In the short term, the immediate opportunity is to deploy agentic AI as an orchestration layer across the systems a practice already uses. Agents hold the state of a task, such as an open referral, and move it between clinical, referral, pathology and engagement platforms through whichever interface each system offers: FHIR or vendor APIs where they exist, secure messaging and document ingestion where they do not. Every platform stays in place.

From FPT’s technology perspective, orchestration is how heterogeneous and legacy environments connect with modern platforms, and it is where the initial value in existing systems can be found, well before any healthcare organisation becomes AI native. The Productivity Commission points to research showing up to 30 per cent of the tasks performed by the healthcare workforce could be automated, and coordination work of this kind is a natural first candidate.

Governance is what makes orchestration scalable. Each agent operates within a defined scope: it can check whether a referral has been acknowledged and draft a follow-up to the specialist’s rooms, while changing a diagnosis, altering a medication or closing a recall stays with the clinician. Every action is logged to the patient record with a timestamp and source system, so the practice can audit what was done, by which agent, on whose authority.

Those controls sit within existing obligations: the Australian Privacy Principles and the My Health Records Act on data handling, AHPRA’s guidance on practitioners’ use of AI, and the TGA’s rules on when software functions as a medical device. Get that right and the same capabilities extend across enterprise healthcare environments, because the governance model travels with the technology.

Closing the loop across the patient journey

Three use cases stand out over the longer term.

The first is closing the loop across the patient journey: tracking referrals, recalls, follow-ups and care gaps across systems and surfacing what remains outstanding.

Consider a 68-year-old patient with type 2 diabetes and declining kidney function. Her GP refers her to a nephrologist, orders repeat pathology in three months and sets a recall. Those three actions typically live in three places: the referral in a secure messaging platform, the pathology request with the laboratory, the recall in the clinical software. If the nephrologist’s rooms never acknowledge the referral, nothing flags it. If the specialist letter arrives as a scanned PDF, someone has to read it, file it and confirm the GP has seen it. If the patient skips the blood test, the recall fires regardless. An agent working across those systems would flag the unacknowledged referral at day 14, match the specialist letter to the open episode when it lands, and confirm the result had been received and reviewed before the recall is closed.

Clinical preparation and scribes offer a second opportunity. Ambient scribes have largely been discussed as transcription tools. The wider role is preparing and contextualising the information a GP needs ahead of an appointment, including the specialist letter and pathology from the example above, supporting documentation during the consultation, and turning what was agreed into follow-up actions afterwards.

Referral and document processing is the third. Specialist correspondence, referrals and pathology reports remain document-heavy workflows in general practice, with much of that material still arriving as unstructured PDFs. Agentic AI can read, classify and route it, with clinicians retaining oversight of every clinical decision.

The right use cases start with context

As with many applications of AI in healthcare, the harder work sits upstream of the technology.

Finding the right use cases takes time and resource. That upfront investment reduces the time and resource otherwise spent on use cases that are wrong for the setting, or that simply don’t apply to how care is delivered.

Healthcare context is highly specific. Practices are already under workforce pressure and run critical workflows with dependencies across many systems and many users. In Foundry Health’s experience of Australian primary care, adoption and change management are the central challenge in any deployment. The question is how to introduce agentic AI across all systems without disrupting the people who depend on them every day.

The disciplined answer is to derive use cases from a close understanding of that context, then build iteratively.

The sector is at the beginning of this horizon. The method for identifying viable use cases is still emerging, and FPT and Foundry Health are developing it with primary care providers as an iterative process: testing each use case in live practice environments and validating it against clinical and operational outcomes before it scales. The same method applies to other areas of healthcare.

The broader opportunity lies in making Australia’s healthcare technology ecosystem work together more intelligently. Australia has spent a decade building the systems. Connecting them around the patient is the work that comes next.

FPT and Foundry Health are validating this approach with Australian primary care providers, testing agentic AI use cases inside existing practice systems and sharing what the results show. Providers and clinical software vendors interested in contributing to that work can get in touch via https://fptsoftware.com/contact-us.