What Happened
InSight+, the online publication of the Medical Journal of Australia, published a practical guide on 2 August 2026 aimed at Australian GPs weighing up whether and how to bring AI tools into their practices.
The guide draws on commentary from Dr Brent Richards, Adjunct Professor of Research at Griffith University and board member of the Queensland AI Consortium, an initiative focused on advancing statewide AI adoption and capability. Dr Richards told InSight+ that GPs should treat their practice like any other business when thinking about where AI can help first.
"Running a general practice is just like running any other business aside from the clinical aspect," he said. "Look at your back office. What repetitive tasks could you automate? You can potentially find a tool that does it for you immediately. Start with your pain points and move on from there."
The guide notes that AI assistants built into existing office software, such as Microsoft Copilot and Google Gemini, can help compose emails and summarise threads. Large language models including Claude and ChatGPT can summarise documents and assist with report writing.
Why It Matters
GP burnout is a real and documented problem in Australia. The guide cites workforce pressures directly: "We're all aware of the rates of burnout amongst GPs, and the workforce pressures that exist." One study cited in the piece found that self-reported burnout rates fell from 52% to 39% among clinicians who used an ambient scribe for 30 days. That is a 13 percentage point drop from a single workflow change.
The RACGP poll finding that four in ten GPs already use an AI scribe means this is not a theoretical future question. Practices that have not thought through their AI governance are already behind the curve on risk management.
Key Details
The guide's central practical advice is to start with administrative tools rather than clinical ones. Dr Richards suggests uploading spreadsheets into a large language model to analyse practice finances: "You can also upload a spreadsheet into a large language model and ask it questions. You can get graphs to see where your profits and losses are; what you're making money on and what you're not making money on."
On clinical tools, the guide is direct about where the evidence sits. AI scribes are the most mature application: "Most notably, AI scribes do a very good job of turning a conversation between the GP and their patient into patient notes."
The guide is equally direct about risk. All AI tools can produce hallucinations, incorrect data interpretations, and errors. The recommended safeguard is consistent: "The important thing is that there is always a 'human in the loop' to pick up on these."
Data representativeness is flagged as a specific concern for clinical tools. "If the data is not representative of the GP's patient cohort, the tool that relies on it will not be optimal." A tool trained predominantly on data from a different demographic or geography may perform poorly for a rural Australian practice or one serving a culturally and linguistically diverse community.
Background and Context
AI in general practice is not new, but the pace of adoption has accelerated. Algorithms reading referral imaging have been in use for some time. What has changed is the availability of consumer-grade large language model tools and purpose-built clinical scribes that any practice can subscribe to without a large IT infrastructure investment.
The RACGP has been tracking adoption through member polls. The four-in-ten figure for AI scribe use reflects how quickly the technology has moved from early adopter territory into mainstream practice.
What Comes Next
The guide anticipates that diagnostic and decision-support tools will become a more significant part of GP clinical practice. "In the future, we foresee that diagnostic and decision-support tools will help GPs with clinical practice," it states. The caveat is that these tools will only be as good as the data they were trained on, and GPs will need to assess whether that data reflects their own patient populations.
For now, the practical starting point remains the back office: identify repetitive administrative tasks, find a tool that handles them, and build familiarity with AI workflows before moving into clinical territory.