featured

How can AI help healthcare?

Podcasts | 8 May 2023

Lamont Tang

Director of Industry Projects

Maureen Murphy

Professor of Practice

Download The AGSM Business of Leadership podcast today on your favourite podcast platform.

Apple Podcasts

Spotify

Google Podcasts

Omny FM

The Business of AI (episode 2): At a time when healthcare workers are stretched to near breaking point, what role does AI have in augmenting natural skills and capabilities?

Guests:

Dimitry Tran, AGSM MBA alumni and Co-founder and Board Director of Harrison AI
Maureen Murphy, UNSW Business School’s Professor of Practice and Commercialisation Facilitator
Lamont Tang, Director of Industry Projects, AGSM @ UNSW Business School (podcast host)

Find out more about Lamont Tang and his work, below:

https://www.linkedin.com/in/lamonttang/
Read now: Can machines invent things? AI reveals the answer is 'yes'
Read now: Microsoft’s Lee Hickin on digital resilience beyond cybersecurity
Read now: How to avoid the ethical pitfalls of artificial intelligence and machine learning
Read now: How leaders should weigh up the risks and rewards of AI
Business AI Lab: https://www.businessai.unsw.edu.au/
UNSW AI Institute: https://www.unsw.edu.au/unsw-ai
UNSW Founders: https://unswfounders.com/
Tyree Institute of Health Engineering: https://www.ihealthe.unsw.edu.au/
Centre for Big Data Research and Health: https://www.unsw.edu.au/research/cbdrh
Centre of Health and Brain Ageing: https://cheba.unsw.edu.au/

Want to follow us?

The best way to keep up-to-date with the latest research and news is to subscribe to BusinessThink and follow UNSW Business School and AGSM @ UNSW Business School on LinkedIn.

Transcript:

Narration: Imagine starting every workday with an inbox filled with 500 new emails. Before you can do anything else, you need to work out which are urgent, which need a follow-up, and which can wait till later.

Now imagine you are a radiologist and instead of 500 emails, you have 500 scans to review. And you can't just skim. Each one needs your full attention; each one needs to be actioned. It’s a near-impossible situation, and one faced in some form by many healthcare workers every day.

But increasingly, smart technologies including artificial intelligence are being created to be used in healthcare to work on this issue. Helping with everything from repetitive tasks to diagnosis and treatment, acting as a copilot that works to free healthcare professionals up to focus on providing better patient care.

Today on the AGSM @ UNSW Business School Business of Leadership podcast, we continue to look at the future of AI – today, specifically in the healthcare industry.

This episode is hosted by Lamont Tang, Director of Industry Projects and Entrepreneur-in-Residence at AGSM @ UNSW Business School.

He is joined by Dimitry Tran, AGSM MBA alumni and Co-founder and Board Director of Harrison AI, a ground-breaking healthcare technology company that combines human intelligence with artificial intelligence, and UNSW Business School’s Professor of Practice Maureen Murphy, Commercialisation Facilitator with i4.

Lamont Tang: Hi, my name is Lamont Tang. I'm the Director of Industry Projects and Entrepreneur in residence at AGSM at UNSW, and I'll be hosting the AGSM Business of Leadership Podcast for the series on The Business of AI.

Today we have a truly exciting episode for you. We have two guests joining us, both of them who are UNSW and AGSM MBA alumni, and experts in their respective fields. First, we have Dimitry Tran, the CEO and co-founder of harrison.ai, one of Australia's fastest-growing scale-ups, and a leader in the artificial intelligence industry in healthcare. Dimitry, and his co-founder and brother, Aengus, have a wealth of experience in the field. Aengus was medically trained at UNSW, they've been both at the forefront of innovation and development in AI, particularly in healthcare.

Second, we've got Maureen Murphy, she's one of Australia's experts in startups, and government investment of capital into the startup and innovation ecosystem. Maureen has a deep understanding of the intersection of technology and business and has served as CEO of several companies from startup to exit, as well as now leading the efforts at commercialising new technologies at i4. i4 is the official delivery partner for Australia's industries acceleration of commercialisation service. Together our two guests will provide us with a unique perspective on the current state and future of AI startups, and corporate innovation.

So without further ado, let's welcome our guests to the show. Hello, Dimitry, and hello, Maureen.

Maureen Murphy: Hello, Lamont.

Dimitry Tran: It's wonderful to be with you today, Lamont.

Lamont Tang: So why don't you each give a 30-second, one-minute blurb about what you do, or sort of highlight key aspects of your career that might be relevant to our discussion today?

Maureen Murphy: Sure. So, I do have a background in the corporate world, primarily as an intrapreneur during those years, but for the last decade I've been deeply involved in the Australian innovation ecosystem as an advisor, a mentor, and an investor. I see companies at a very early stage, and I'm quite privileged, I think, to be able to work with Australia's best and brightest entrepreneurs, and also researchers with novel IP on their road to commercialisation. Over the years, I've actually supported hundreds of new businesses with high growth potential, from a range of industry sectors, including robotics, advanced manufacturing, energy, ag tech, and also medical devices.

Lamont Tang: Thank you, Maureen. How about yourself, Dimitry? Tell us a little bit about yourself, and maybe a little bit about Harrison.ai.

Dimitry Tran: As a graduate from AGSM, where I attended the Executive MBA Programme between 2011 and '15, it's great to connect with the AGSM audience this time. I spent about 15 years in healthcare, first with Ramsay Healthcare, a global healthcare hospital operator in Australia, but also operate in Europe and Asia. In the last four years, has been involved with the startup that me and my brother that you mentioned, Dr. Aengus Tran, who's also a UNSW alumni, founded harrison.ai.

Our mission is to scale the capacity of healthcare globally through AI technology. For example, in one of our ventures Annalise.ai, we build AI that can read chest x-ray or CT brain, that help clinicians to diagnose disease much faster and much more accurate. We're living in a world post-COVID, where the pressure in healthcare system is immense, and we believe that the introduction of technologies like AI can really help to address one of the fundamental problems of our society today.

Lamont Tang: So Dimitry, for those that are new to harrison.ai, can you explain a little bit about your structure with annalise.ai and franklin.ai?

Dimitry Tran: Sure. The way to think about harrison.ai is we are a platform company that enable AI as make or revise to build very quickly, so we're leveraging a common technology platform to build. Our first venture called annalise.ai in the field of radiology, that is in partnership with I-MED, the largest radiology provider here in Australia. So, Annalise has been a wonderful product in chest x-ray, on CT brain, on CT chest to diagnose hundreds of medical conditions, and help patient to get a better care faster.

The second venture within our portfolio is called franklin.ai, that we just started about a year ago. Again, in partnership with a large healthcare provider in Australia called Sonic Healthcare. Together we are building AI in pathology, specifically in histopathology, where we're looking to diagnose cancers like prostate cancer, breast cancer a lot earlier as well. In summary, you can see that we have a quite interesting business model, where we leverage the speed and the technology capability of a startup, being Harrison, and partner up with leading healthcare companies like I-MED and Sonic, to build ventures that within only a few years has become market leader.

Lamont Tang: And so how would you describe what you do?

Dimitry Tran: I think, fundamentally what we do is we provide a co-pilot to clinicians. When a radiologist looking at a chest x-ray, for example, there would be many things that take a long time to train for a physician to be able to pick up. It took up to 16 years to train a radiologist. What we do is we provide a co-pilot that can detect findings alongside the doctors. For example, sign of pneumonia on a chest x-ray, or sign-up stroke on a CT brain, and that will help the clinician to make more accurate diagnosis on a timelier manner. Certainly, healthcare is one area where we certainly need more support, given the ageing population and the demand in healthcare that we're seeing today.

Lamont Tang: That's amazing. Maybe Maureen, you can tell us a little bit about how you define or think about artificial intelligence, and its application in healthcare. What maybe trends you've been noticing or what do you expect in the future?

Maureen Murphy: Yeah, sure. I think what we're seeing is the digitization of everything, and AI is all about data and being able to use that data to make predictions, and to sort through gargantuan amounts of data that's being produced these days. We can make things easier and faster for clinicians, and for users in every industry, it's not just in the health industry, so that we're augmenting the ability and the capabilities of professionals. In healthcare especially, as Dimitry mentioned, there are some really great examples where AI can be used to assist the clinicians to improve what they do, by augmenting their capabilities, those human capabilities. A couple of examples from Australian companies in this space, that I could give you would be a company called DetectedX, for example, uses AI to enhance images, x-rays and CAT scans, and so on, to help radiologists diagnose diseases a lot faster and more accurately, in particular breast cancer and lung cancer. It's assisting radiologists to be able to do that better and faster.

Another example would be where surgeons are able to practice a surgery, pre-plan a surgery before they start it. Again, an Australian company that's made huge advances in this space is a company called Vantari. They are using AI combined with virtual reality, in a way that enables the surgeons to take the MRI images, turn those into a virtual reality environment so surgeons can go in, and actually see what they're going to be doing when they make that first incision. So, they can see what's happening in the aorta, or they can see what's happening in the blood vessels. Obviously, if they're able to pre-plan the surgery, the outcomes for patients are going to be so much better, and the risks of surgery will be reduced, and that's great for healthcare.

Lamont Tang: Thank you, Maureen. There are so many things to unpack there. Maureen and Dimitry, how do you see AI helping our clinicians and doctors?

Dimitry Tran: I was recently talking with a clinician in Australia, a radiologist, and she told me that every day she starts her day staring down a list of 500 cases, of radiology cases that has been backlogging since yesterday and since overnight. She said she lose the will to work because she knows that she can try very hard, and bring it down to 200, but that tomorrow you start her day again with 500. I think each of those cases are someone mother, someone father, someone loved one that require the best of care, the best attention possible, yet our resources are so stretched that we are having someone who have to deal with hundreds of cases every day. We all have loved one who've been through healthcare, and the nervousness of waiting for a diagnosis for something to be done to them, and I think that is where AI plays such an important role.

One of the most important things that we do with our technology is our AI can process those 500 cases in a few second, and then we allow the clinician to sort their work list like an Excel worksheet, to sort and say, "Which cases?" Maybe Case 399 is a case that contained a critical finding, the stroke patient that need immediate care the next minute, and maybe the first 100 patients are all normal cases that the radiologist can leisurely work through the work list in the afternoon. I think this is something that AI can uniquely do, in addition to helping the clinicians when to diagnose the disease, when they open the case up on their workstation.

Another thing that is fantastic about AI, is that it provides a constant performance. I have another clinician in the UK, who told me recently that, "You don't want to be the patient that I diagnose after lunch, because my attention is at the lowest on the day. You want to be first patient when I have my coffee, when I have my donut, my performance is peak performance." So, even if the AI performance is slightly below human performance, whenever the human dip below the AI performance, that is ... those trough is where the AI can catch the human. I think this augmentation, to Maureen point, which is a key word here, it is about augmenting human with technology that provide constant performance, so that together, human plus AI, will deliver far, superior service. We all know that healthcare is such an important thing to all of our lives.

Maureen Murphy: AI eats data for lunch, and doesn't get tired. It has enormous value in also training medical professionals, so training doctors for instance, have to learn to do quite critical procedures like intubating people, which on the old traditional method, or still current method, is the see one, do one, and then teach one. But imagine if you are intubating a baby, for instance, it's incredibly stressful for not only the patient, but also for the trainee doctors. So being able to practice in a safe environment, where you're doing the same thing, but on, perhaps a virtual reality model, and you're getting results back telling you how successful you've been, you can do all of this before you actually do the real thing. That's the sort of thing that AI will enable, so we're getting better people coming through the system, and then augmenting the work of the clinicians once they're in those professional roles, like neurosurgery or radiology.

The stress on the healthcare system will be reduced, and we know it's only getting worse with an ageing population, with more chronic lifetime diseases, unless we do something about it now, we're going to face very difficult problems by the time we're in aged care.

Lamont Tang: Thank you, Maureen. I'd like to touch upon back one of the themes that Dimitry mentioned, which is this theme of augmentation. I think, in my mind at least, it looks like the popular media tends to veer on the narrative that AI is going to come for our jobs, it's going to destroy our livelihoods, but I think if you look at the history of innovation, it seems like every time we humans come up with some creative innovation, actually humans always find a way to create some type of new jobs for ourselves or livelihoods. I'm just curious how, Dimitry and Maureen, you think about, from a business point of view, for some of our leaders and executives, and or our alumni, how they navigate this space and how they can think about some frameworks about augmentation in their own careers.

Maureen Murphy: Perhaps I'll touch on some of the barriers to adoption. I think that's something that we know, and when we're in the industry we know that AI is going to change the way we do everything, not just in healthcare, but in almost every industry, but the challenges are often that there's no stakeholder buy-in. Sometimes the innovation comes, and quite often I see this in startup companies particularly, the innovation comes from the technologists, but it needs to be led by the senior managers in organisations.

There needs to be a strong level of executive support for any AI initiative to be successful in an organisation, and that executive support needs to include the team that supports the organisation, so the CIO, the head of HR, all of the users. In a hospital situation, the clinicians, the key opinion leaders, the administrative people, all have to buy into it, and that level of communication that has to occur needs to be organisation-wide. By showing the users that there is value in what's being delivered, and that their jobs will be made easier because they're going to have their activities augmented, that will help overcome employee resistance, but it takes time and it takes strong leadership.

Dimitry Tran: I think Maureen raising some very important point here, which is from a business framework, and coming back to my AGSM training, we're talking about disruptive innovation here, where we’re essentially changing the workflow of a clinician. There used to be healthcare in all clinicians in medical school, and my co-founder is a medical doctor, can speak to this best, is you are trained to become an independent operator. You are there, you are at the point of care, you make the decision, and you treat the patient. What you're changing is we're bringing a lot more data, we're bringing a co-pilot, AI, to the workflow. It's quite disruptive, and I think we need to recognise this. This is not sustaining innovation. This is not something that you incrementally change. This is something that you introduce into a workflow that already overtaxed with all the pressure on the healthcare system that's going through right now.

I think there's a humility that need to come through from anyone who tried to innovate in this space, that thing will take time, and big investment need to come into change management. We all have experienced change management in different way, but I think healthcare change management is incredibly difficult, and getting all the stakeholders aligned, like Maureen mentioned, but I think having the investors and the wider stakeholders' group within innovation to understand that as well. So, if a startup is looking to scale, and bringing the right investor who appreciate healthcare, I think that would be a crucial element so that they will have the right funding to go through this disruptive change period, and ultimately crossing the chasm to the other side. Otherwise, there's going to be a lot of good ideas being pitched to hospitals that they have no capacity to absorb, no matter how great is the idea that you present them with.

Lamont Tang: Thank you, Dimitry. I think there's so many things that we could discuss there, change management, strategic thinking, the technical understanding of leaders, can we delve a little bit more into your own journey, Dimitry? As you've gone from, I assume, in 2018 when you started a very small startup, just between you and your brother, basically much more energy than strategy, as you've grown and you've scaled up, how have you changed as a leader, your leadership style? How have you thought about change management in your own organisation?

Dimitry Tran: It's a very interesting journey for me, because after AGSM in 2015, me and my brother start thinking about different ideas to start a business. Ultimately, in 2018 and '19, we formed a company, and like you said, initially there was just the two of us, and we are now a company of close to 400 people in a few countries, so it has been extraordinary amount of growth. A lot of the learning was by doing, so simply from AGSM training, and all the frameworks that you can apply, but simply as a leader, when you manage a much smaller team, it's all about the vision, it's all about the execution, and how do we survive, essentially, to the next milestone. Once we reached the scale that we are today, it's all about the people, and about creating environment where great work could be done, because we ourselves cannot do the great work anymore.

Most of my time these days involve in people matters, in terms of setting up our culture, recruitment and training. Especially, the tech market going through such a difficult time at the moment, with a lot of uncertainty, and it's all about how to retain the best people, take opportunity to bring on new talents as they become available, and keep an eye on the goal. I think it's been a very fast four years for us, for sure, and I'm sure there's still many lessons to come. We are very lucky that we have great mentors along the way. We have a great board with very experienced investors like Blackbird Venture, who led our Series A, and then Skip Capital, which is Kim Jackson, and Scott Farquhar from Atlassian. So having great investors along the way telling us what around the corner, telling us what kind of mistake we are about to make. I think just being humble about things that we don't know because it's such an exciting field, but the exciting thing about it is that not a lot of people have been through it yet.

Lamont Tang: Thank you, Dimitry. I think it's, Reid Hoffman of LinkedIn, and former PayPal Mafia, or it was Jeff Weiner whom he hired to scale LinkedIn, they said that as you move from inspiration to institution, which is sort of where you are along that journey, that it's not just about diversity, equity, and inclusion, DEI, but also diversity, equity and belonging. You've mentioned culture, and I think as you grow it's such a big part of a successful, sustainable, and thriving organisation. Where are you along this journey, and what challenges have you faced, or what lessons have you learned that you wish you would've learned sooner?

Dimitry Tran: I think Reid Hoffman had his comparison between being a pirate ship and being a Navy ship. Initially, you need to be a pirate and being very, I guess, entrepreneurial and hustling for things, but eventually you need to be a Navy if you want to build a scalable business. I think the thing different about healthcare, I assume Maureen would agree, is that you have to be a Navy day one, because what you're doing is not things that can be turned on and off again. It can't be something that is fail fast or you move fast and break things. These are things that's so important to people lives, and healthcare system are so conservative in their procurement that they won't procure an MVP. You don't want MVP drugs. You don't want to have an MVP medical device, and you definitely don't want an MVP AI that diagnosed your cancer.

Maureen Murphy: I totally agree, I think it's also one of the problems of scaling, not just an organisation, but of scaling AI in an organisation, the problems of getting beyond the pilot stage. Have you seen that, Dimitry? Organisations will get to the pilot stage, they'll get a nice little AI experience, but they have difficulty scaling it from pilot to an enterprise level AI improvement.

Dimitry Tran: Precisely, I think that's the big danger with AI, is that it's so easy to start. The amazing thing about AI algorithm is that with a thousand data point, you can build an AI that work about 80% of the time. The problem is the long tail distribution of data, is that to get the remaining 20%, you need exponentially more data. You need hundreds of thousands to conquer the last 20%, so big corporation, and sometimes because of the culture or the incentive, maybe misaligned, to try to have a great press release, to have the great announcement of something. That's what I used to call innovation theatre when I worked in corporate land, that you innovate to look good, and to have something to say to investors rather than true innovation that going to change the business, and set the business for success in the long-term.

So, I think that's the danger of AI, is it's so easy to start, but it's so hard to finish. That's why most startups that are much more focused or much more, I guess, survival-focused, that they have to solve the problem, will be getting there faster and solve the problem better than a large corporation. But what we have done at Harrison, that we uniquely do, is that we leverage the strength of large corporations with the speed of a startup. I think Australia, in particular, have great opportunities because we have wonderful healthcare company like ResMed, like Cochlear, like Sonic, like Ramsay, like I-MED, that operate at the global scale here on our backyard, and hence Harrison is set up to be that true partner delivering the focus on execution side.

Maureen Murphy: The co-pilot. Absolutely.

Lamont Tang: To build on that, Dimitry, I was curious, are there any strategic frameworks or ways you think about whether to build on your own intellectual property versus partnering with these corporates like I-MED and Sonic Healthcare? Do you have a decision tree? How do you guys think about when you decide partnerships, how do you approach these type of issues?

Dimitry Tran: You're testing my AGSM knowledge here, Lamont, but I remember that it used to be a view called resource-based view of strategy, where you think about what resource you have, and therefore that drive your strategy. Clearly, to view AI, to Maureen point, is AI eat data for our breakfast, so you need to have data. When a startup by itself wouldn't have any data, because by nature it takes a long time to accumulate data, and data is extremely valuable, but extremely difficult to manage as well, we saw a lot of examples of privacy issues of data not being kept properly, so I think there's a lot going into managing data, especially in healthcare. That's coming back to our earlier conversation between being a pirate versus a Navy ship, is that from day one you need to be a Navy ship, and everything needs to be locked down and secure, because you're dealing with something that is very consequential and very important here, which is healthcare data.

Lamont Tang: So, if you have these big tech companies that have the eyeballs, the talents, the pipelines, where does a startup play in this environment, and how do you make these decisions?

Dimitry Tran: I will limit my answer to healthcare because that's what I operate in, but I think healthcare is a very different field to general social media or web technology in general. The data used tend to be very private. The data tend to be very locked down. It is not something that you put online, it is something that you need to curate carefully, anonymize carefully. There's a lot of consent that need to go into processing of data. I think this is an area where actually large health system in the world is thinking long and hard about how to leverage their data, because they are the custodian of those data, but they need to use it for a better outcome, which is to build AI technologies that can scale capacity and improve quality of healthcare.

I look upon technologies like ChatGPT, of course with great interest, but I think we also need to appreciate the limitation of where AI is today. It seems to be quite human in interaction. Sometime if I don't want to draft a long email, I just type in a prompt into ChatGPT to get it to write email for me, and I edit from there, but I think the key element is editing, because sometime it could be very, wrong. I think in healthcare we cannot afford to be wrong. We try very, very hard to be as accurate as possible, given everything at stake.

The great thing about healthcare as well, is guardrail’s already in place. In Australia, we have the TGA, in the US, we have the FDA, in Europe, we have the CE, so all these organisations set a clear standard on is AI effective AI? It's no different than a blood test for COVID, or a sputum test for TB, all these testing have clear guideline on what make a good test. Even AI going to say cancer, no cancer is essentially a diagnosis test, and therefore it needs to pass the existing guardrail as well. I'm actually very optimistic about healthcare AI because we already have wonderful guardrail to make sure that only safe AI make its way to doctor's hand and impacting patient care.

Lamont Tang: How about you Maureen?

Maureen Murphy: Well, I'm thinking listening to that, the focus really is on the importance of the validity of the data sets. If they're open data sets, that's great, but the point is that AI relies on this massive amount of data to produce insights, and its functionality, its accuracy is all dependent on how good that data set is. AI, as a technology, can only handle the activities it's been trained to do, and so we have to be very careful about what data we are feeding into it, otherwise we can potentially have algorithmic bias, for example, and that's something that is of great concern when we look at a number of applications that are out there. As Dimitry says, in health, we can't afford to take risks, we can't afford to get it almost right, it has to be perfect. Addressing something like cognitive biases, we've seen in healthcare that there have been issues due to the insufficiency of the data sets that have been used.

There are examples where, gender disparity has occurred in the treatment of ... the diagnosis, I mean, of skin cancer, for example, where the data sets are about primarily white patients, so the diagnosis for people who don't have white skin have not been trained into the algorithm. Similarly, and something close to my heart, is that in cardiology, for example, a heart attack is overwhelmingly misdiagnosed in women. Prediction models for cardiovascular disease that claim to predict heart attacks five years before they happen, are trained on predominantly male data sets. So, if cardiovascular disease has different patterns in women than men, then the algorithm that's been trained predominantly on data samples from men is useless in diagnosing cardiovascular disease in women. This is where, I think, larger data sets with more diversity is an essential component in the development of AI-enabled technology in healthcare.

Lamont Tang: Thank you, Maureen. This leads to one of the points that I'd like to maybe delve a little bit more, sort of the ethical considerations and bias. So Dimitry, are there any methodologies that you had to develop in-house to innovate in this manner that are different than what are typically taught at, say the Stanford Design School or the design thinking type of frameworks, to ensure the safety and trust in the AI algorithm?

Dimitry Tran: I think when it comes to clinical care, the first hurdle to pass is regulatory. The regulators are very sophisticated around AI, they have amazing data scientists, they have a panel of PhDs, so they ask very deep questions. For example, subpopulation analysis, like Maureen mentioned, you may have data set with 100 patients, but 99 of them are male, so one is female, for example. Then you can show a very high performance of your AI that doesn't work at all for female, but it's still going to show a very high-top line number. So, subpopulation analysis is the technique where you show your performance on different subpopulation and show that you have consistent performance in subpopulation. That's just one example where regulators are asking AI companies in healthcare to perform, and to show the data, I usually call that the end point. The starting point is you need to have a great amount of data that already are quite diverse in your training, and I think this is where Australia, again, have a lot of advantage.

We have a migrant country. We have people from a lot of countries coming here to Australia, therefore an AI trained in Australia will have advantage because this translates very well. When we take our chest x-ray AI solution to Malaysia, to Scotland, to Boston, all those places, the AI work out of the box because it has been trained on quite diverse population. Whereas an AI that have been trained uniquely on a certain demographic, or people of certain background will not translate so well. I think, to Maureen point, it's very, very important to start with the right data, and it's very hard to do that as a startup, because you just don't naturally have data advantage.

Recently, we actually see a lot of AI that even pass regulatory approval, and get into user hand and stop, and has we call performance drift, or data drift over time because the users use it in a different way than the population that the data was trained on. I think this is one of the things that we still coming to grasp with globally at the healthcare industry, is how to monitor AI performance in real-time. This is not a one-off check, you're safe, and from this point on you can sell to whoever and wherever you want. The AI technology needs to be monitored in real-time, and get the feedback from the user saying, "You are fantastic on every patient except for patient Asian female of a certain age group." That is the feedback that we need to go back and collect more to keep training AI for the better. This continuous learning, I think, is a key element of any AI business model.

Lamont Tang: What takeaways should leaders take when you have your AI, and you've got these market trends coming in, to what extent do they need to have an understanding of AI technology, the impact that it has, its potential applications? How do you think about informing them to help them make better decisions and utilise AI within their organisations? Any key takeaways as we start wrapping up?

Maureen Murphy: I think the main takeaway for me is that leaders in organisations don't necessarily have to understand how AI works, but they do need to understand the capabilities, and the power AI has when it's applied to add value to their organisations. Most clinical leaders don't get all that excited about data and technology, but when they start looking at how it can improve the bottom line, how it can improve productivity and ROI, then I think we start to see the value or they start to see the value.

Also, culture, I think we need to look at the cultures that embrace change, and that might require a different type of leader, leaders who are more visionary and reinventors, and they do have to have a particular skill, and that is, that they need to be able to keep business as usual going, but they also need to create quite disruptive change while that's happening. It's like they have to be able to keep flying the plane while also putting a new engine in, and it's not necessarily an easy thing, and I think Dimitry mentioned this earlier, change management capabilities need to be very strong, and bottom-line communication, good communication, and great data.

Lamont Tang: Great. How about yourself, Dimitry?

Dimitry Tran: If I think back to my time working in corporate, I think there are two things that I wish I know earlier. One, is the book that I can recommend, it's called Prediction Machine. It's a great AI book, it's nothing technical, but it's all about how the economics of AI will impact any given business, and it provide great framework to think through that. The thesis around Prediction Machine is that when prediction become cheaper, people use more of it, and if you can use more prediction, how is that going to impact your business? Just like electricity, we figure out a lot, we were able to use electricity because it's become so cheap, the same with computing. Now my microwave has a computer inside, because computing become so cheap, we figure out new way of using it. I think prediction used to be very expensive, it takes 16 years to train a radiologist, now you can predict a 100,000 x-ray in a few second. You can do it differently, but if you can predict so many things so fast these days, how is it going to impact everyone's business? I think that's the book I would recommend.

The second thing is to think about having an in-house AI expert at the C level or at the bot level, so the whole executive team do not need to become AI expert. We all can learn a little bit about AI, but I think having someone with deep enough knowledge on how to build AI at the decision table, I think, is important. Because there's so many gotcha in AI that we discussed today, like bias, like data uses ethics, like diversity, like deployment, like change management. I think AI has been around long enough that there are those people who have been through one or two deployments of AI, or development of AI, that can bring great insight to a discussion, because it's so easy to generalise AI into, "Oh, it's going to change everything. Oh, it's going to be so risky." I think having someone with that experience around the table, I think, would be very helpful to shape the conversation or strategy, or any decision that organisation going to make in this field.

Lamont Tang: thank you, Dimitry, thank you, Maureen, for joining us today. It's clear that AI is rapidly transforming the business landscape, and we're excited to see how your companies, and others continue to leverage this technology to drive growth and innovation.

I'll close by saying always be learning, and to Dimitry, and to our other innovators and leaders, and Maureen, stay hungry and stay foolish.

Maureen Murphy: Thank you.

Narration: Thank you for joining us for AGSM’s The Business of AI: How can AI help healthcare?

Want to learn more about our research and work in the area? Check out the show notes on our website to get more information on the research discussed on today’s podcast. Just search the-business-of-podcast online.

Or, drop us a line if you have feedback at brand@agsm.edu.au

New to the podcast? There’s a whole catalogue for you to explore. From mental health and AI, to the ethics of AI implementation, you can check them all out today.

Please share, rate and review and subscribe to AGSM’s leadership podcast on your favourite podcast platform and look out for future episodes.

In the meantime, follow AGSM at UNSW Business School on LinkedIn and Facebook for more industry insights for an accelerating world or find us at agsm.edu.au.

Business Strategy Corporate Governance Risk Ethics Artificial Intelligence Innovation Technology Analytics Data Health

Republish this article

Republish

You are free to republish this article both online and in print. We ask that you follow some simple guidelines.

Please do not edit the piece, ensure that you attribute the author, their institute, and mention that the article was originally published on Business Think.

By copying the HTML below, you will be adhering to all our guidelines.

Press Ctrl-C to copy

<h1>How can AI help healthcare?</h1>

<figure><img src="?w=1320" /><figcaption></figcaption></figure>
 Guests:
<ul>
 <li>Dimitry Tran, AGSM MBA alumni and Co-founder and Board Director of Harrison AI</li>
 <li>Maureen Murphy, UNSW Business School’s Professor of Practice and Commercialisation Facilitator</li>
 <li>Lamont Tang, Director of Industry Projects, AGSM @ UNSW Business School (podcast host)</li>
</ul>
Find out more about Lamont Tang and his work, below:
<ul>
 <li><a href="https://www.linkedin.com/in/lamonttang/" data-new-window="true" target="_blank" rel="noopener noreferrer">https://www.linkedin.com/in/lamonttang/</a></li>
 <li>Read now: <a href="https://www.businessthink.unsw.edu.au/articles/machines-invent-things-ai" data-new-window="true" target="_blank" rel="noopener noreferrer">Can machines invent things? AI reveals the answer is 'yes'</a></li>
 <li>Read now: <a href="https://www.businessthink.unsw.edu.au/articles/microsoft-lee-hickin-digital-resilience-cybersecurity" data-new-window="true" target="_blank" rel="noopener noreferrer">Microsoft’s Lee Hickin on digital resilience beyond cybersecurity</a></li>
 <li>Read now: <a href="https://www.businessthink.unsw.edu.au/articles/ethical-pitfalls-artificial-intelligence-machine-learning" data-new-window="true" target="_blank" rel="noopener noreferrer">How to avoid the ethical pitfalls of artificial intelligence and machine learning</a></li>
 <li>Read now: <a href="https://www.businessthink.unsw.edu.au/articles/risks-rewards-ai-leaders" data-new-window="true" target="_blank" rel="noopener noreferrer">How leaders should weigh up the risks and rewards of AI</a></li>
 <li>Business AI Lab: <a href="https://www.businessai.unsw.edu.au/" data-new-window="true" target="_blank" rel="noopener noreferrer">https://www.businessai.unsw.edu.au/</a></li>
 <li>UNSW AI Institute: <a href="https://www.unsw.edu.au/unsw-ai" data-new-window="true" target="_blank" rel="noopener noreferrer">https://www.unsw.edu.au/unsw-ai</a></li>
 <li>UNSW Founders: <a href="https://unswfounders.com/" data-new-window="true" target="_blank" rel="noopener noreferrer">https://unswfounders.com/</a></li>
 <li>Tyree Institute of Health Engineering: <a href="https://www.ihealthe.unsw.edu.au/" data-new-window="true" target="_blank" rel="noopener noreferrer">https://www.ihealthe.unsw.edu.au/</a></li>
 <li>Centre for Big Data Research and Health: <a href="https://www.unsw.edu.au/research/cbdrh" data-new-window="true" target="_blank" rel="noopener noreferrer">https://www.unsw.edu.au/research/cbdrh</a></li>
 <li>Centre of Health and Brain Ageing: <a href="https://cheba.unsw.edu.au/" data-new-window="true" target="_blank" rel="noopener noreferrer">https://cheba.unsw.edu.au/</a></li>
</ul>
Want to follow us?
The best way to keep up-to-date with the latest research and news is to <a href="https://www.businessthink.unsw.edu.au/subscribe" data-new-window="true" target="_blank" rel="noopener noreferrer">subscribe to BusinessThink</a> and follow <a href="https://www.linkedin.com/school/3274515" data-new-window="true" target="_blank" rel="noopener noreferrer">UNSW Business School</a> and <a href="https://www.linkedin.com/school/15104530" data-new-window="true" target="_blank" rel="noopener noreferrer">AGSM @ UNSW Business School</a> on LinkedIn.
Transcript:
Narration: Imagine starting every workday with an inbox filled with 500 new emails. Before you can do anything else, you need to work out which are urgent, which need a follow-up, and which can wait till later.
Now imagine you are a radiologist and instead of 500 emails, you have 500 scans to review. And you can't just skim. Each one needs your full attention; each one needs to be actioned. It’s a near-impossible situation, and one faced in some form by many healthcare workers every day.
But increasingly, smart technologies including artificial intelligence are being created to be used in healthcare to work on this issue. Helping with everything from repetitive tasks to diagnosis and treatment, acting as a copilot that works to free healthcare professionals up to focus on providing better patient care.
Today on the AGSM @ UNSW Business School Business of Leadership podcast, we continue to look at the future of AI – today, specifically in the healthcare industry.
This episode is hosted by Lamont Tang, Director of Industry Projects and Entrepreneur-in-Residence at AGSM @ UNSW Business School.
He is joined by Dimitry Tran, AGSM MBA alumni and Co-founder and Board Director of Harrison AI, a ground-breaking healthcare technology company that combines human intelligence with artificial intelligence, and UNSW Business School’s Professor of Practice Maureen Murphy, Commercialisation Facilitator with i4.
Lamont Tang: Hi, my name is Lamont Tang. I'm the Director of Industry Projects and Entrepreneur in residence at AGSM at UNSW, and I'll be hosting the AGSM Business of Leadership Podcast for the series on The Business of AI.
Today we have a truly exciting episode for you. We have two guests joining us, both of them who are UNSW and AGSM MBA alumni, and experts in their respective fields. First, we have Dimitry Tran, the CEO and co-founder of harrison.ai, one of Australia's fastest-growing scale-ups, and a leader in the artificial intelligence industry in healthcare. Dimitry, and his co-founder and brother, Aengus, have a wealth of experience in the field. Aengus was medically trained at UNSW, they've been both at the forefront of innovation and development in AI, particularly in healthcare.
Second, we've got Maureen Murphy, she's one of Australia's experts in startups, and government investment of capital into the startup and innovation ecosystem. Maureen has a deep understanding of the intersection of technology and business and has served as CEO of several companies from startup to exit, as well as now leading the efforts at commercialising new technologies at i4. i4 is the official delivery partner for Australia's industries acceleration of commercialisation service. Together our two guests will provide us with a unique perspective on the current state and future of AI startups, and corporate innovation.
So without further ado, let's welcome our guests to the show. Hello, Dimitry, and hello, Maureen.
Maureen Murphy: Hello, Lamont.
Dimitry Tran: It's wonderful to be with you today, Lamont.
Lamont Tang: So why don't you each give a 30-second, one-minute blurb about what you do, or sort of highlight key aspects of your career that might be relevant to our discussion today?
Maureen Murphy: Sure. So, I do have a background in the corporate world, primarily as an intrapreneur during those years, but for the last decade I've been deeply involved in the Australian innovation ecosystem as an advisor, a mentor, and an investor. I see companies at a very early stage, and I'm quite privileged, I think, to be able to work with Australia's best and brightest entrepreneurs, and also researchers with novel IP on their road to commercialisation. Over the years, I've actually supported hundreds of new businesses with high growth potential, from a range of industry sectors, including robotics, advanced manufacturing, energy, ag tech, and also medical devices.
Lamont Tang: Thank you, Maureen. How about yourself, Dimitry? Tell us a little bit about yourself, and maybe a little bit about Harrison.ai.
Dimitry Tran: As a graduate from AGSM, where I attended the Executive MBA Programme between 2011 and '15, it's great to connect with the AGSM audience this time. I spent about 15 years in healthcare, first with Ramsay Healthcare, a global healthcare hospital operator in Australia, but also operate in Europe and Asia. In the last four years, has been involved with the startup that me and my brother that you mentioned, Dr. Aengus Tran, who's also a UNSW alumni, founded harrison.ai.
Our mission is to scale the capacity of healthcare globally through AI technology. For example, in one of our ventures Annalise.ai, we build AI that can read chest x-ray or CT brain, that help clinicians to diagnose disease much faster and much more accurate. We're living in a world post-COVID, where the pressure in healthcare system is immense, and we believe that the introduction of technologies like AI can really help to address one of the fundamental problems of our society today.
Lamont Tang: So Dimitry, for those that are new to harrison.ai, can you explain a little bit about your structure with annalise.ai and franklin.ai?
Dimitry Tran: Sure. The way to think about harrison.ai is we are a platform company that enable AI as make or revise to build very quickly, so we're leveraging a common technology platform to build. Our first venture called annalise.ai in the field of radiology, that is in partnership with I-MED, the largest radiology provider here in Australia. So, Annalise has been a wonderful product in chest x-ray, on CT brain, on CT chest to diagnose hundreds of medical conditions, and help patient to get a better care faster.
The second venture within our portfolio is called franklin.ai, that we just started about a year ago. Again, in partnership with a large healthcare provider in Australia called Sonic Healthcare. Together we are building AI in pathology, specifically in histopathology, where we're looking to diagnose cancers like prostate cancer, breast cancer a lot earlier as well. In summary, you can see that we have a quite interesting business model, where we leverage the speed and the technology capability of a startup, being Harrison, and partner up with leading healthcare companies like I-MED and Sonic, to build ventures that within only a few years has become market leader.
Lamont Tang: And so how would you describe what you do?
Dimitry Tran: I think, fundamentally what we do is we provide a co-pilot to clinicians. When a radiologist looking at a chest x-ray, for example, there would be many things that take a long time to train for a physician to be able to pick up. It took up to 16 years to train a radiologist. What we do is we provide a co-pilot that can detect findings alongside the doctors. For example, sign of pneumonia on a chest x-ray, or sign-up stroke on a CT brain, and that will help the clinician to make more accurate diagnosis on a timelier manner. Certainly, healthcare is one area where we certainly need more support, given the ageing population and the demand in healthcare that we're seeing today.
Lamont Tang: That's amazing. Maybe Maureen, you can tell us a little bit about how you define or think about artificial intelligence, and its application in healthcare. What maybe trends you've been noticing or what do you expect in the future?
Maureen Murphy: Yeah, sure. I think what we're seeing is the digitization of everything, and AI is all about data and being able to use that data to make predictions, and to sort through gargantuan amounts of data that's being produced these days. We can make things easier and faster for clinicians, and for users in every industry, it's not just in the health industry, so that we're augmenting the ability and the capabilities of professionals. In healthcare especially, as Dimitry mentioned, there are some really great examples where AI can be used to assist the clinicians to improve what they do, by augmenting their capabilities, those human capabilities. A couple of examples from Australian companies in this space, that I could give you would be a company called DetectedX, for example, uses AI to enhance images, x-rays and CAT scans, and so on, to help radiologists diagnose diseases a lot faster and more accurately, in particular breast cancer and lung cancer. It's assisting radiologists to be able to do that better and faster.
Another example would be where surgeons are able to practice a surgery, pre-plan a surgery before they start it. Again, an Australian company that's made huge advances in this space is a company called Vantari. They are using AI combined with virtual reality, in a way that enables the surgeons to take the MRI images, turn those into a virtual reality environment so surgeons can go in, and actually see what they're going to be doing when they make that first incision. So, they can see what's happening in the aorta, or they can see what's happening in the blood vessels. Obviously, if they're able to pre-plan the surgery, the outcomes for patients are going to be so much better, and the risks of surgery will be reduced, and that's great for healthcare.
Lamont Tang: Thank you, Maureen. There are so many things to unpack there. Maureen and Dimitry, how do you see AI helping our clinicians and doctors?
Dimitry Tran: I was recently talking with a clinician in Australia, a radiologist, and she told me that every day she starts her day staring down a list of 500 cases, of radiology cases that has been backlogging since yesterday and since overnight. She said she lose the will to work because she knows that she can try very hard, and bring it down to 200, but that tomorrow you start her day again with 500. I think each of those cases are someone mother, someone father, someone loved one that require the best of care, the best attention possible, yet our resources are so stretched that we are having someone who have to deal with hundreds of cases every day. We all have loved one who've been through healthcare, and the nervousness of waiting for a diagnosis for something to be done to them, and I think that is where AI plays such an important role.
One of the most important things that we do with our technology is our AI can process those 500 cases in a few second, and then we allow the clinician to sort their work list like an Excel worksheet, to sort and say, "Which cases?" Maybe Case 399 is a case that contained a critical finding, the stroke patient that need immediate care the next minute, and maybe the first 100 patients are all normal cases that the radiologist can leisurely work through the work list in the afternoon. I think this is something that AI can uniquely do, in addition to helping the clinicians when to diagnose the disease, when they open the case up on their workstation.
Another thing that is fantastic about AI, is that it provides a constant performance. I have another clinician in the UK, who told me recently that, "You don't want to be the patient that I diagnose after lunch, because my attention is at the lowest on the day. You want to be first patient when I have my coffee, when I have my donut, my performance is peak performance." So, even if the AI performance is slightly below human performance, whenever the human dip below the AI performance, that is ... those trough is where the AI can catch the human. I think this augmentation, to Maureen point, which is a key word here, it is about augmenting human with technology that provide constant performance, so that together, human plus AI, will deliver far, superior service. We all know that healthcare is such an important thing to all of our lives.
Maureen Murphy: AI eats data for lunch, and doesn't get tired. It has enormous value in also training medical professionals, so training doctors for instance, have to learn to do quite critical procedures like intubating people, which on the old traditional method, or still current method, is the see one, do one, and then teach one. But imagine if you are intubating a baby, for instance, it's incredibly stressful for not only the patient, but also for the trainee doctors. So being able to practice in a safe environment, where you're doing the same thing, but on, perhaps a virtual reality model, and you're getting results back telling you how successful you've been, you can do all of this before you actually do the real thing. That's the sort of thing that AI will enable, so we're getting better people coming through the system, and then augmenting the work of the clinicians once they're in those professional roles, like neurosurgery or radiology.
The stress on the healthcare system will be reduced, and we know it's only getting worse with an ageing population, with more chronic lifetime diseases, unless we do something about it now, we're going to face very difficult problems by the time we're in aged care.
Lamont Tang: Thank you, Maureen.  I'd like to touch upon back one of the themes that Dimitry mentioned, which is this theme of augmentation. I think, in my mind at least, it looks like the popular media tends to veer on the narrative that AI is going to come for our jobs, it's going to destroy our livelihoods, but I think if you look at the history of innovation, it seems like every time we humans come up with some creative innovation, actually humans always find a way to create some type of new jobs for ourselves or livelihoods. I'm just curious how, Dimitry and Maureen, you think about, from a business point of view, for some of our leaders and executives, and or our alumni, how they navigate this space and how they can think about some frameworks about augmentation in their own careers.
Maureen Murphy: Perhaps I'll touch on some of the barriers to adoption. I think that's something that we know, and when we're in the industry we know that AI is going to change the way we do everything, not just in healthcare, but in almost every industry, but the challenges are often that there's no stakeholder buy-in. Sometimes the innovation comes, and quite often I see this in startup companies particularly, the innovation comes from the technologists, but it needs to be led by the senior managers in organisations.
There needs to be a strong level of executive support for any AI initiative to be successful in an organisation, and that executive support needs to include the team that supports the organisation, so the CIO, the head of HR, all of the users. In a hospital situation, the clinicians, the key opinion leaders, the administrative people, all have to buy into it, and that level of communication that has to occur needs to be organisation-wide. By showing the users that there is value in what's being delivered, and that their jobs will be made easier because they're going to have their activities augmented, that will help overcome employee resistance, but it takes time and it takes strong leadership.
Dimitry Tran: I think Maureen raising some very important point here, which is from a business framework, and coming back to my AGSM training, we're talking about disruptive innovation here, where we’re essentially changing the workflow of a clinician. There used to be healthcare in all clinicians in medical school, and my co-founder is a medical doctor, can speak to this best, is you are trained to become an independent operator. You are there, you are at the point of care, you make the decision, and you treat the patient. What you're changing is we're bringing a lot more data, we're bringing a co-pilot, AI, to the workflow. It's quite disruptive, and I think we need to recognise this. This is not sustaining innovation. This is not something that you incrementally change. This is something that you introduce into a workflow that already overtaxed with all the pressure on the healthcare system that's going through right now.
I think there's a humility that need to come through from anyone who tried to innovate in this space, that thing will take time, and big investment need to come into change management. We all have experienced change management in different way, but I think healthcare change management is incredibly difficult, and getting all the stakeholders aligned, like Maureen mentioned, but I think having the investors and the wider stakeholders' group within innovation to understand that as well. So, if a startup is looking to scale, and bringing the right investor who appreciate healthcare, I think that would be a crucial element so that they will have the right funding to go through this disruptive change period, and ultimately crossing the chasm to the other side. Otherwise, there's going to be a lot of good ideas being pitched to hospitals that they have no capacity to absorb, no matter how great is the idea that you present them with.
Lamont Tang: Thank you, Dimitry. I think there's so many things that we could discuss there, change management, strategic thinking, the technical understanding of leaders, can we delve a little bit more into your own journey, Dimitry? As you've gone from, I assume, in 2018 when you started a very small startup, just between you and your brother, basically much more energy than strategy, as you've grown and you've scaled up, how have you changed as a leader, your leadership style? How have you thought about change management in your own organisation?
Dimitry Tran: It's a very interesting journey for me, because after AGSM in 2015, me and my brother start thinking about different ideas to start a business. Ultimately, in 2018 and '19, we formed a company, and like you said, initially there was just the two of us, and we are now a company of close to 400 people in a few countries, so it has been extraordinary amount of growth. A lot of the learning was by doing, so simply from AGSM training, and all the frameworks that you can apply, but simply as a leader, when you manage a much smaller team, it's all about the vision, it's all about the execution, and how do we survive, essentially, to the next milestone. Once we reached the scale that we are today, it's all about the people, and about creating environment where great work could be done, because we ourselves cannot do the great work anymore.
Most of my time these days involve in people matters, in terms of setting up our culture, recruitment and training. Especially, the tech market going through such a difficult time at the moment, with a lot of uncertainty, and it's all about how to retain the best people, take opportunity to bring on new talents as they become available, and keep an eye on the goal. I think it's been a very fast four years for us, for sure, and I'm sure there's still many lessons to come. We are very lucky that we have great mentors along the way. We have a great board with very experienced investors like Blackbird Venture, who led our Series A, and then Skip Capital, which is Kim Jackson, and Scott Farquhar from Atlassian. So having great investors along the way telling us what around the corner, telling us what kind of mistake we are about to make. I think just being humble about things that we don't know because it's such an exciting field, but the exciting thing about it is that not a lot of people have been through it yet.
Lamont Tang: Thank you, Dimitry. I think it's, Reid Hoffman of LinkedIn, and former PayPal Mafia, or it was Jeff Weiner whom he hired to scale LinkedIn, they said that as you move from inspiration to institution, which is sort of where you are along that journey, that it's not just about diversity, equity, and inclusion, DEI, but also diversity, equity and belonging. You've mentioned culture, and I think as you grow it's such a big part of a successful, sustainable, and thriving organisation. Where are you along this journey, and what challenges have you faced, or what lessons have you learned that you wish you would've learned sooner?
Dimitry Tran: I think Reid Hoffman had his comparison between being a pirate ship and being a Navy ship. Initially, you need to be a pirate and being very, I guess, entrepreneurial and hustling for things, but eventually you need to be a Navy if you want to build a scalable business. I think the thing different about healthcare, I assume Maureen would agree, is that you have to be a Navy day one, because what you're doing is not things that can be turned on and off again. It can't be something that is fail fast or you move fast and break things. These are things that's so important to people lives, and healthcare system are so conservative in their procurement that they won't procure an MVP. You don't want MVP drugs. You don't want to have an MVP medical device, and you definitely don't want an MVP AI that diagnosed your cancer.
Maureen Murphy: I totally agree, I think it's also one of the problems of scaling, not just an organisation, but of scaling AI in an organisation, the problems of getting beyond the pilot stage. Have you seen that, Dimitry? Organisations will get to the pilot stage, they'll get a nice little AI experience, but they have difficulty scaling it from pilot to an enterprise level AI improvement.
Dimitry Tran: Precisely, I think that's the big danger with AI, is that it's so easy to start. The amazing thing about AI algorithm is that with a thousand data point, you can build an AI that work about 80% of the time. The problem is the long tail distribution of data, is that to get the remaining 20%, you need exponentially more data. You need hundreds of thousands to conquer the last 20%, so big corporation, and sometimes because of the culture or the incentive, maybe misaligned, to try to have a great press release, to have the great announcement of something. That's what I used to call innovation theatre when I worked in corporate land, that you innovate to look good, and to have something to say to investors rather than true innovation that going to change the business, and set the business for success in the long-term.
So, I think that's the danger of AI, is it's so easy to start, but it's so hard to finish. That's why most startups that are much more focused or much more, I guess, survival-focused, that they have to solve the problem, will be getting there faster and solve the problem better than a large corporation. But what we have done at Harrison, that we uniquely do, is that we leverage the strength of large corporations with the speed of a startup. I think Australia, in particular, have great opportunities because we have wonderful healthcare company like ResMed, like Cochlear, like Sonic, like Ramsay, like I-MED, that operate at the global scale here on our backyard, and hence Harrison is set up to be that true partner delivering the focus on execution side.
Maureen Murphy: The co-pilot. Absolutely.
Lamont Tang: To build on that, Dimitry, I was curious, are there any strategic frameworks or ways you think about whether to build on your own intellectual property versus partnering with these corporates like I-MED and Sonic Healthcare? Do you have a decision tree? How do you guys think about when you decide partnerships, how do you approach these type of issues?
Dimitry Tran: You're testing my AGSM knowledge here, Lamont, but I remember that it used to be a view called resource-based view of strategy, where you think about what resource you have, and therefore that drive your strategy. Clearly, to view AI, to Maureen point, is AI eat data for our breakfast, so you need to have data. When a startup by itself wouldn't have any data, because by nature it takes a long time to accumulate data, and data is extremely valuable, but extremely difficult to manage as well, we saw a lot of examples of privacy issues of data not being kept properly, so I think there's a lot going into managing data, especially in healthcare. That's coming back to our earlier conversation between being a pirate versus a Navy ship, is that from day one you need to be a Navy ship, and everything needs to be locked down and secure, because you're dealing with something that is very consequential and very important here, which is healthcare data.
Lamont Tang: So, if you have these big tech companies that have the eyeballs, the talents, the pipelines, where does a startup play in this environment, and how do you make these decisions?
Dimitry Tran: I will limit my answer to healthcare because that's what I operate in, but I think healthcare is a very different field to general social media or web technology in general. The data used tend to be very private. The data tend to be very locked down. It is not something that you put online, it is something that you need to curate carefully, anonymize carefully. There's a lot of consent that need to go into processing of data. I think this is an area where actually large health system in the world is thinking long and hard about how to leverage their data, because they are the custodian of those data, but they need to use it for a better outcome, which is to build AI technologies that can scale capacity and improve quality of healthcare.
I look upon technologies like ChatGPT, of course with great interest, but I think we also need to appreciate the limitation of where AI is today. It seems to be quite human in interaction. Sometime if I don't want to draft a long email, I just type in a prompt into ChatGPT to get it to write email for me, and I edit from there, but I think the key element is editing, because sometime it could be very, wrong. I think in healthcare we cannot afford to be wrong. We try very, very hard to be as accurate as possible, given everything at stake.
The great thing about healthcare as well, is guardrail’s already in place. In Australia, we have the TGA, in the US, we have the FDA, in Europe, we have the CE, so all these organisations set a clear standard on is AI effective AI? It's no different than a blood test for COVID, or a sputum test for TB, all these testing have clear guideline on what make a good test. Even AI going to say cancer, no cancer is essentially a diagnosis test, and therefore it needs to pass the existing guardrail as well. I'm actually very optimistic about healthcare AI because we already have wonderful guardrail to make sure that only safe AI make its way to doctor's hand and impacting patient care.
Lamont Tang: How about you Maureen?
Maureen Murphy: Well, I'm thinking listening to that, the focus really is on the importance of the validity of the data sets. If they're open data sets, that's great, but the point is that AI relies on this massive amount of data to produce insights, and its functionality, its accuracy is all dependent on how good that data set is. AI, as a technology, can only handle the activities it's been trained to do, and so we have to be very careful about what data we are feeding into it, otherwise we can potentially have algorithmic bias, for example, and that's something that is of great concern when we look at a number of applications that are out there. As Dimitry says, in health, we can't afford to take risks, we can't afford to get it almost right, it has to be perfect. Addressing something like cognitive biases, we've seen in healthcare that there have been issues due to the insufficiency of the data sets that have been used.
There are examples where, gender disparity has occurred in the treatment of ... the diagnosis, I mean, of skin cancer, for example, where the data sets are about primarily white patients, so the diagnosis for people who don't have white skin have not been trained into the algorithm. Similarly, and something close to my heart, is that in cardiology, for example, a heart attack is overwhelmingly misdiagnosed in women. Prediction models for cardiovascular disease that claim to predict heart attacks five years before they happen, are trained on predominantly male data sets. So, if cardiovascular disease has different patterns in women than men, then the algorithm that's been trained predominantly on data samples from men is useless in diagnosing cardiovascular disease in women. This is where, I think, larger data sets with more diversity is an essential component in the development of AI-enabled technology in healthcare.
Lamont Tang: Thank you, Maureen. This leads to one of the points that I'd like to maybe delve a little bit more, sort of the ethical considerations and bias. So Dimitry, are there any methodologies that you had to develop in-house to innovate in this manner that are different than what are typically taught at, say the Stanford Design School or the design thinking type of frameworks, to ensure the safety and trust in the AI algorithm?
Dimitry Tran: I think when it comes to clinical care, the first hurdle to pass is regulatory. The regulators are very sophisticated around AI, they have amazing data scientists, they have a panel of PhDs, so they ask very deep questions. For example, subpopulation analysis, like Maureen mentioned, you may have data set with 100 patients, but 99 of them are male, so one is female, for example. Then you can show a very high performance of your AI that doesn't work at all for female, but it's still going to show a very high-top line number. So, subpopulation analysis is the technique where you show your performance on different subpopulation and show that you have consistent performance in subpopulation. That's just one example where regulators are asking AI companies in healthcare to perform, and to show the data, I usually call that the end point. The starting point is you need to have a great amount of data that already are quite diverse in your training, and I think this is where Australia, again, have a lot of advantage.
We have a migrant country. We have people from a lot of countries coming here to Australia, therefore an AI trained in Australia will have advantage because this translates very well. When we take our chest x-ray AI solution to Malaysia, to Scotland, to Boston, all those places, the AI work out of the box because it has been trained on quite diverse population. Whereas an AI that have been trained uniquely on a certain demographic, or people of certain background will not translate so well. I think, to Maureen point, it's very, very important to start with the right data, and it's very hard to do that as a startup, because you just don't naturally have data advantage.
Recently, we actually see a lot of AI that even pass regulatory approval, and get into user hand and stop, and has we call performance drift, or data drift over time because the users use it in a different way than the population that the data was trained on. I think this is one of the things that we still coming to grasp with globally at the healthcare industry, is how to monitor AI performance in real-time. This is not a one-off check, you're safe, and from this point on you can sell to whoever and wherever you want. The AI technology needs to be monitored in real-time, and get the feedback from the user saying, "You are fantastic on every patient except for patient Asian female of a certain age group." That is the feedback that we need to go back and collect more to keep training AI for the better. This continuous learning, I think, is a key element of any AI business model.
Lamont Tang: What takeaways should leaders take when you have your AI, and you've got these market trends coming in, to what extent do they need to have an understanding of AI technology, the impact that it has, its potential applications? How do you think about informing them to help them make better decisions and utilise AI within their organisations? Any key takeaways as we start wrapping up?
Maureen Murphy:  I think the main takeaway for me is that leaders in organisations don't necessarily have to understand how AI works, but they do need to understand the capabilities, and the power AI has when it's applied to add value to their organisations. Most clinical leaders don't get all that excited about data and technology, but when they start looking at how it can improve the bottom line, how it can improve productivity and ROI, then I think we start to see the value or they start to see the value.
Also, culture, I think we need to look at the cultures that embrace change, and that might require a different type of leader, leaders who are more visionary and reinventors, and they do have to have a particular skill, and that is, that they need to be able to keep business as usual going, but they also need to create quite disruptive change while that's happening. It's like they have to be able to keep flying the plane while also putting a new engine in, and it's not necessarily an easy thing, and I think Dimitry mentioned this earlier, change management capabilities need to be very strong, and bottom-line communication, good communication, and great data.
Lamont Tang: Great. How about yourself, Dimitry?
Dimitry Tran: If I think back to my time working in corporate, I think there are two things that I wish I know earlier. One, is the book that I can recommend, it's called Prediction Machine. It's a great AI book, it's nothing technical, but it's all about how the economics of AI will impact any given business, and it provide great framework to think through that. The thesis around Prediction Machine is that when prediction become cheaper, people use more of it, and if you can use more prediction, how is that going to impact your business? Just like electricity, we figure out a lot, we were able to use electricity because it's become so cheap, the same with computing. Now my microwave has a computer inside, because computing become so cheap, we figure out new way of using it. I think prediction used to be very expensive, it takes 16 years to train a radiologist, now you can predict a 100,000 x-ray in a few second. You can do it differently, but if you can predict so many things so fast these days, how is it going to impact everyone's business? I think that's the book I would recommend.
The second thing is to think about having an in-house AI expert at the C level or at the bot level, so the whole executive team do not need to become AI expert. We all can learn a little bit about AI, but I think having someone with deep enough knowledge on how to build AI at the decision table, I think, is important. Because there's so many gotcha in AI that we discussed today, like bias, like data uses ethics, like diversity, like deployment, like change management. I think AI has been around long enough that there are those people who have been through one or two deployments of AI, or development of AI, that can bring great insight to a discussion, because it's so easy to generalise AI into, "Oh, it's going to change everything. Oh, it's going to be so risky." I think having someone with that experience around the table, I think, would be very helpful to shape the conversation or strategy, or any decision that organisation going to make in this field.
Lamont Tang: thank you, Dimitry, thank you, Maureen, for joining us today. It's clear that AI is rapidly transforming the business landscape, and we're excited to see how your companies, and others continue to leverage this technology to drive growth and innovation.
I'll close by saying always be learning, and to Dimitry, and to our other innovators and leaders, and Maureen, stay hungry and stay foolish.
Maureen Murphy: Thank you.
Narration: Thank you for joining us for AGSM’s The Business of AI:  How can AI help healthcare?
Want to learn more about our research and work in the area? Check out the show notes on our website to get more information on the research discussed on today’s podcast. Just search the-business-of-podcast online.
Or, drop us a line if you have feedback at <a href="mailto:brand@agsm.edu.au">brand@agsm.edu.au</a>
New to the podcast? There’s a whole catalogue for you to explore. From mental health and AI, to the ethics of AI implementation, you can check them all out today.
Please share, rate and review and subscribe to AGSM’s leadership podcast on your favourite podcast platform and look out for future episodes.
In the meantime, follow AGSM at UNSW Business School on LinkedIn and Facebook for more industry insights for an accelerating world or find us at agsm.edu.au.

Comments

Podcasts

featured

How can AI help healthcare?

Download The AGSM Business of Leadership podcast today on your favourite podcast platform.

Republish

Related

Ethical marketing: how to navigate the AI moral minefield for marketers

The progress principle: how climate resilience is shaping jobs of the future

Risky business: adapting to climate change could ‘insure’ a better future

Find an expert

Connect