I've spent most of my career at the point where medicine and technology meet. Sometimes, the technology made the doctor's job harder, not easier. Electronic records were meant to make care more productive and connected. Instead, many of us ended up facing a screen for half the consult and finishing our notes at the kitchen table after the kids had gone to bed.
That's not a small problem. The Royal Australian College of General Practitioners found 70% of GPs are worried about their admin workload, and it's now the leading reason GPs say they plan to leave their practice. When a GP leaves, patients lose more than an appointment slot. They lose someone who knows their history.
In medicine we talk about working at the top of your licence, doing the work only you are trained to do. I trained for six years at university, and after graduation many of my early clinical years involved documentation, filling out request forms, doing discharge summaries, and sometimes sitting by the fax machine waiting for correspondence. I enjoyed providing care, but the administration and work around the work was disengaging.
Where AI is actually helping
The first job we gave AI at Heidi was to simply listen to the consultation and draft the notes, so the clinician can look at the patient instead of the keyboard. The clinician still reads, edits, and signs off every note. That principle hasn't changed as the technology has improved, and it won't.
What has changed is how much of the work around each visit AI can now take on. Heidi drafts referral letters, discharge summaries, and patient instructions while the patient is still in the room. When a clinician wants to check a guideline or recent research, it can search trusted clinical sources using Amazon Bedrock and show where every answer came from, so the clinician can verify it before acting. After the visit, clinics can follow up with patients by phone or text through Amazon Connect.
In Australia, clinicians using Heidi tell us they get back around 59 minutes a day, and 87% of clinic staff say they're doing less overtime. Ninety-seven percent rate the documents it produces as good or excellent. Patients have been comfortable too, with 99% agreeing to its use in their consult.
Why we build our AI on AWS
Clinical AI depends on the infrastructure underneath it. A tool that drops out mid-visit, or an answer a clinician can't trace back to a guideline, does more harm than no tool at all. Heidi runs on Amazon Web Services (AWS), which gives us a secure, reliable place to store, process and host health information. That's why we trust AWS as our cloud and AI provider, and why we recently expanded that agreement for another three years.
None of that is visible in the consult, and it shouldn't be. Patients should notice their doctor's attention, not the technology behind it. We can do that across 190 countries, and still meet local data storage requirements, because we don't have to build that foundation ourselves.
Trust has to come before scale
Today, Heidi supports around three million patient visits a week across more than 190 countries and in 110 languages. Numbers like that only matter if every one of those visits is safe and private. In healthcare, a tool earns its place one consult at a time.
That's why we handle patient data in line with regional requirements, so Australian patient information is hosted in Australia. It's why we add extra checks on high-risk details like medications and allergies, and why we became the first AI scribe certified to ISO 42001, an international standard for managing AI responsibly.
What comes next
The next step for clinical AI is taking on more of the preparation and follow-through that surrounds care: a briefing on the patient before they walk in, hospital records coded automatically, referrals that don't get lost. The clinician decides what needs their approval, and the clinical decisions stay with them.
Australia has some of the best clinicians in the world. If we give them back more time, I know they'll spend it meaningfully with their patients.