Reducing The Cost of Admin in Australian General Practice
“I get home an hour early every day.”
That line belongs to an Australian GP describing what an AI tool changed about her working day. Admin is usually costed in minutes – a few lost per consultation, tallied up at day’s end into an operational inconvenience. That accounting is comfortable, since it keeps the issue small enough to live with, but three studies published across 2025 and early 2026 – one from the American Medical Association, one a survey of general practitioners in the United Kingdom, and the Australian focus groups already quoted above – suggest the real figure sits differently.
Across all three, documentation is the task doctors nominate most often for automation, ahead of diagnostic support, referral drafting and patient communication combined.
Where the burden concentrates
The American Medical Association’s 2024 survey of nearly 1,200 physicians found that 57% named reducing administrative burden through automation as the single biggest opportunity AI offered them, ahead of diagnostic accuracy and personalised treatment.
Actual use nearly doubled over the same period – 66% of physicians reported using AI in 2024, up from 38% the year before, a rise of 78% in twelve months. Asked what they used it for, physicians pointed to documentation more than anything else too, with use of AI for billing codes, charts and visit notes rising from 13 to 21% of physicians in a single year, the fastest-growing application in the survey.
The same pattern holds in the United Kingdom, where Charlotte Blease and colleagues’ 2025 survey of 1,005 general practitioners, published in Digital Health, found that 25% now use generative AI in clinical practice, and that among those users, 35% applied it to documentation after patient appointments – the largest single use case, ahead of differential diagnosis (27%), treatment options (24%) and referrals (24%).
A whopping 71% of those users said that the AI tools reduced their overall workload.
“Physicians are increasingly intrigued by the assistive role of health AI and the potential of AI-enable tools to reduce administrative burdens, enhance diagnostic accuracy, and personalise treatments.”
The knowledge gap outranks the fear of error
Adoption remains fairly uneven, and Australian research by CSIRO and Avant Mutual shows why. Asked what stopped them from using AI, or what still concerned them as existing users, doctors in both Sydney and Melbourne ranked not knowing enough about how the technology worked or where their patients’ data went above every other issue, including the risk of the software producing an outright error. 93.3% of users and 94.4% of non-users raised it, ahead of AI error itself at 86.7% and 83.3% per cent respectively.
A doctor who had adopted the technology, asked the same question, admitted she had never been told the answer either. “I can’t say exactly how it works, because no one’s told me.”
The knowledge gap, on both sides of the divide, is not a training deficiency sitting at the margins of the job. It sits at the centre of whether a doctor can use the tool with a straight face.
“If patients ask me, is the data encrypted, where's it stored, is it in Australia or overseas, I don't have the confidence to say yes, this, this and this. And if I don't have that confidence, they'll look at me and say, well, can I trust your advice then.”
Nor is that deficit evenly distributed. In the UK survey, 95% of GenAI users had received no formal training from their employer, and 85% said their employer had not even encouraged them to use the tools they were already relying on. The Australian focus groups found the gap bites hardest for doctors in small or solo practices, who have to select and assess AI software themselves, against doctors in larger organisations with an IT department to do that research for them.
When vetting AI tools, who carries the risk?
Doctors with an IT department behind them are better placed than most, but they are not fully insulated from this cost either, since institutional support of that kind usually buys due diligence on one additional piece of software rather than the removal of the category of work altogether.
An AI tool bought as a separate product, however well built, still asks someone in the practice to vet it as a new piece of software before it can be trusted with a patient record. Software that reads or writes the note as a function already sitting inside the system a doctor uses, by contrast, removes that requirement at the point it would otherwise arise, since the practice assessed the record system’s governance once, when it adopted it.
None of this is particular to documentation, either. The same calculation applies to whichever category of AI a practice adopts next, whether that is filing incoming correspondence, summarising a patient’s history before an appointment, or something not yet built.
The object being processed changes but the underlying question does not. A capability bought from the vendor that already holds the record answers that question before it is asked. A capability bought from anyone else asks it again.
MediRecords’ two AI tools to date, Evolve Direct and Evolve Patient Summary, are built on that logic rather than around it.
Evolve Direct files incoming correspondence, referral letters and results into the right patient record automatically, without the documents ever leaving the practice’s own MediRecords environment, and is credited with saving up to two hours a day on that category of admin alone. Evolve Patient Summary works the other side of the same problem, generating a summary of a patient’s last three consultations, recent correspondence and investigations within seconds, so a doctor opening a file before an appointment is not the one doing the collating.
Neither asked a practice to onboard a separate vendor, since both run inside the record the practice had already assessed when it adopted MediRecords in the first place.
The pattern beyond documentation
Each new category of clinical AI will keep arriving promising time saved, and each time, what decides whether the promise holds is not what the software can do but where responsibility for vetting it sits once deployed. A practice that treats every new AI capability as a fresh procurement decision, weighed against a new vendor’s claims about security and data handling, pays that vetting cost again for each one. A practice that expects its record system to extend natively, function by function, pays it once. The saving the AMA and Blease surveys measured, and the hour the general practitioner quoted at the outset got back, hold only under the second model. Otherwise the burden changes shape, from typing to due diligence, without shrinking.
Rather than take the argument above on faith, the more useful test is a practical one. For the next piece of AI a practice is weighing up, ask whether it sits inside a system already vetted, or asks the practice to start that process again from a new vendor.
Nor is that deficit evenly distributed. In the UK survey, 95% of GenAI users had received no formal training from their employer, and 85% said their employer had not even encouraged them to use the tools they were already relying on. The CSIRO focus groups found the gap bites hardest for doctors in small or solo practices, who have to select and assess AI software themselves, against doctors in larger organisations with an IT department to do that research for them.
Sources
- American Medical Association. “2 in 3 physicians are using health AI - up 78% from 2023.” AMA, 26 Feb 2025. ama-assn.org/practice-management/digital-health/2-3-physicians-are-using-health-ai-78-2023
- Blease, C., Hagström, J., Garcia Sanchez, C., Kharko, A., McMillan, B., Gaab, J., Brulin, E., Locher, C., Hägglund, M., Riggare, S. & Mandl, K.D. “General practitioners' adoption of generative artificial intelligence in clinical practice in the UK: An updated online survey.” Digital Health, 2025;11. doi.org/10.1177/20552076251394287
- Irons, J.L., Duenser, A., Pickett, T., Haysom, G. & McGrath, M.J. “AI in medical practice: doctors' perspective on the benefits, challenges and facilitators of artificial intelligence scribe use.” Health and Technology, 2026;16:401–410. doi.org/10.1007/s12553-025-01042-x
Considering a move to MediRecords?
Choose your current system and enter your details into the form below. We’ll show your full migration breakdown, and if you’re ready to proceed, our team will be in touch with your timeframe.
Data we move across from
Expand a category to see where it lives in MediRecords →
Coverage reflects standard MediRecords migration scope. Some fields depend on how your current system exports, and in some cases data moves across but is stored in a different place in MediRecords.
