Meet the Patient
For RANZCO 2026 in Auckland, we built three conversational AI patients into two private consult rooms on our client’s stand — real conversations, not a script, that let surgeons rehearse a patient history before recommending a lens. One installation, one goal: turn something surgeons skip under pressure into something they wanted to do twice.
AI Patients built from real people.
What we did
Our client wanted surgeons walking onto their stand at RANZCO to practise something most of them skip under pressure: taking a proper patient history before recommending a lens. The brief was to build that rehearsal without it feeling like a rehearsal.
We built three conversational AI patients, each built from a real patient profile and voiced and filmed with an actor in a Melbourne studio: a retired farmer who drives unlit roads at night and keeps losing his readers, a retired English teacher juggling three pairs of glasses for her reading, quilting and theatre habit, and a financial planner who spends most of his day across multiple screens. None of them volunteers anything. Each patient opens with a greeting and nothing else, and only gives up what matters, their daily life, what they want from surgery, what they'd trade away, if the surgeon actually asks.
Two private consult rooms ran on the stand for three days. Each had a 55" portrait screen at seated eye level, a headset and boom mic, and a chair. Everything was started from a phone by the IOL Manager running the room, no one touched the equipment in the wall, and both rooms were monitored and adjustable remotely for the full three days without opening the stand.
The brief carried a strict privacy requirement that comes with any patient-facing content in a healthcare setting, so the system was built to that standard from the start: no transcript, no audio, no video, no names. All it keeps is the room, the patient chosen, the start time and the length, and every conversation is deleted from the platform the moment it ends. Meeting a bar that strict is usually where this kind of experience gets compromised, scripted down, made safe instead of real, to make the privacy problem go away. This one wasn't.
Why it worked
Across the three show days the two rooms logged 55 sessions. Forty-three ran a minute or longer, median length five minutes thirty-eight seconds, with fourteen running until the time limit cut them off, the longest a ten-minute-eight-second consult on the Monday. The other twelve were warm-ups and resets between surgeons.
Of the three patients, the retired English teacher, whose complaint was glasses for near and intermediate tasks rather than distance vision, was chosen nearly twice as often as either of the other two. That wasn’t surgeons picking at random, it was the IOL Managers running the rooms, and over three days and 43 real consults they settled on a favourite. That’s a real signal about which patient story felt most familiar to walk in with, not something you’d get from a survey.
On the floor, two things came up from surgeons more than once, unprompted:
“I thought this was real.”
“Can we use this for training in the hospital?”
That strict privacy standard is really the headline, met without the experience losing anything for it. With no transcript to fall back on and no video to review, the patients had to carry a real conversation on performance alone, and across 43 real consults they did. That’s the actual proof point: a hospital-grade privacy bar and a convincing experience aren’t a trade-off, if you design for both from the start rather than bolting privacy on afterwards.
Two production calls did a lot of the work. First, casting and filming real actors rather than generating a synthetic AI face. A patient who looks like someone you’d actually see in clinic reads completely differently to an obviously rendered avatar, and “I thought this was real” doesn’t happen with the latter. Second, the voice: a real actor’s own voice, recorded properly, with an actual Australian accent, not a text-to-speech voice defaulting to the sickly-sweet American tone most people now associate with AI. Small thing on paper, but it’s the difference between a patient who sounds like someone from down the road and one who sounds like a demo.