00:00You say that you have already $50 million in annual recurring revenue,
00:04millions of patient interactions each week.
00:07Explain to us what exactly Heidi does and where your next phase of growth could come from.
00:15Yeah, hi, thanks for having me.
00:17So yeah, Heidi works by transcribing the conversation.
00:20Most patients and doctors will experience Heidi in the background,
00:24turning that conversation into words.
00:26And then when the visit ends, the doctors will click stop and will generate the clinical notes,
00:30documentation, and then actually trigger Heidi to go do some of the tasks
00:34that would normally burden the clinician for the rest of their day.
00:37And in practice, that sort of starting to do some of the work and the tasks
00:42that usually consume a lot of doctors' time is the way that we see our growth going forward
00:47and also how we actually give capacity back to the healthcare system.
00:52But I understand the goal right now, especially with artificial intelligence,
00:56is moving beyond what you're already doing in documentation,
00:59not just being an AI scribe, right?
01:01Would you also be able to do more, say, prescribe patients, perhaps under supervision?
01:08What would that look like?
01:10And are there any risks?
01:13Yeah, I think, so when I imagine a healthcare AI future, it's still with the doctor at the center.
01:20So I was a surgical registrar.
01:22I know what it's like to have, you know, hundreds of patients in the waiting room.
01:25In the visit, there's a lot of work that needs to be done before you see the next patient.
01:29So that's where we started with the scribe.
01:31So generating the clinical notes, the documents,
01:35maybe drafting orders or drafting referrals for clinicians' review,
01:38but still the clinician has to be at the center approving, prescribing anything that has high clinical risk.
01:44It's really important that these AI tools are built specifically for healthcare.
01:48So that's where Heidi comes in.
01:50We have a whole clinical team.
01:51I think a third of our team is clinical, and we just spend time on that safety piece.
01:55As we go forward, I do see lots of opportunity for tools that we build and other companies
02:01to actually get closer to clinical care.
02:04So imagine things like triage or even before the visit, Heidi can actually reason for, you know,
02:1010, 20 minutes about the case, look up all the evidence.
02:13And as a doctor, I can walk into the room with having an AI having thought about my patient
02:18for quite a period of time.
02:19Where these go wrong is where you have teams who aren't just focused on healthcare.
02:24So the stakes are different.
02:25You know, if I get a recipe wrong, if I'm a consumer AI tool, that's a low stakes mistake.
02:30But if I get a medicine name wrong or a dose wrong, that's incredibly dangerous.
02:35And so you need to have safety guardrails and evaluation built into the tool from the
02:40ground up to make sure that it's appropriate for the use case you're going for.
02:44And so that's what we do.
02:45But we are very excited about these agents starting to do more work for our doctors.
02:50What would those guardrails look like?
02:53I mean, we had the likes of OpenAI Anthropic already coming out and talking about outside safety
02:57evaluators earlier.
02:58So what would that look like for clinical AI if we move here beyond just documentation
03:04into doing more with AI agents in the healthcare system?
03:10Yeah.
03:11So there's really well-established standards in healthcare.
03:14So you've got things like medical device regulation.
03:18And also each organization will usually have a really strong AI and governance sort of body.
03:24And so typically what it looks like is in the product development cycle on Heidi's side,
03:29it's really important that we do copious evaluations before we ever put a product in someone's hands.
03:36So a simple example so everyone can understand is let's say we're answering a clinical question
03:40for a doctor.
03:41They ask us a question.
03:42We look up the sources and we give the clinicians an answer.
03:45We would want to run millions of queries and we do demonstrating that we're factual.
03:50We're finding the right sources.
03:51We represent the sources safely and find the edge cases where it goes wrong and actually
03:55build tools and systems to make sure that Heidi can be relied upon.
03:59Then once we've built it, we actually have to do it in the real world.
04:01So we have alpha and beta cohorts where we run trials.
04:05And this is third-party organizations, not just us marking our own homework, where we test
04:10the product in situ and see what actually happens.
04:12And I think that's the key part for all AI companies.
04:16It's really important to do that peer evaluation and have organizations have an evaluation phase
04:21when you're in that early deployment.
04:23And then once it gets ticked off, it doesn't end there.
04:26You keep going once it's in deployment, monitoring, evaluating it, having safety reviews every month,
04:31seeing how Heidi's performing and letting clinicians escalate any problems or things they're seeing
04:36go wrong.
04:38Where are you currently operating and where do you see the bright spots in different regions?
04:43I mean, the health care system is just so complex depending on where you are and completely
04:48different from what happens in the West, especially in the U.S. with insurance companies, everything
04:53related to what's happening over there with the big hospitals.
04:57But then in Asia, you also have more national health care as well.
05:03Yeah, we're seeing incredible adoption across the board.
05:06You know, a lot of modern economies are spending somewhere from 5% to 10% of their GDP on
05:11health
05:11care.
05:12Our populations are getting older, people are getting more unwell, and there's no support
05:16shortage of demand on the health care system as we get better at treating more complicated
05:20conditions.
05:21So actually, what Heidi does, like the core nucleus of patient-doctor interaction, supporting
05:27the clinician making the best decisions, reducing the amount of time it takes to provide
05:31incredible care, that nuclear experience is actually the same around the world.
05:35Now, the system around it, how it's funded, what the incentives are, you know, who we sell
05:40to, that changes by country.
05:42But the core product of making clinical practice incredible is actually quite similar.
05:47So that's how we've managed to grow.
05:48You know, you might notice by Australian accident, we started in Australia.
05:51We have amazing teams now across Asia, Europe, we're in the UK, Canada, and the U.S.
05:56as well.
05:57So it's really sort of expanded globally from that core Australian base where we started.
06:02And I think this is a global opportunity.
06:04It's probably one of the biggest applications of AI that is clearly for good.
06:09It still has to be stewarded carefully, but I'm incredibly excited to be a part of leading
06:13that front.
06:14It's probably a little bit more than if we're going to be a part of it.
06:14Let's get started.
06:14Everything was too hard to say.
