What is AI’s Equity Story?

Guillaume Duparc:
Thank you. Welcome for this really interesting topic. Of course, I think the team upstairs needs to switch to the other presentation that we had put together. Thank you. It's my pleasure on L.E.K.'s behalf to introduce this topic in this really fantastic panel we have here today. Saying just a few word of context since COVID, we've had a big interest in digital. There's a lot of providers in this room. We use digital as part of the operations, but for, I think a good few years now, we also have AI, whether predictive or generative as part of the toolkit and the technology has developed very significantly in terms of capability. We heard in the prior a panel that they are clearly concerned around hallucination, clinical quality and the usage of the technology from a clinical decision support standpoint can be very tricky. But there's a broader range of applications that also, I think are easier for private providers in particular.

But the capabilities is developing, the cost of compute is reducing. Clinicians widely use it, not necessarily integrated with their medical record system, but there's great level of interest and there are benefits and use cases across patient experience, patient quality, operational efficiency and otherwise. And our panel members will talk in more detail about this. So it's my pleasure to introduce first, Ola Spencer, who's both COO and CTO at Optegra, has been with the company for seven year and being key to the digital transformation of the business. Prior to working at Optegra, she advised Fortune 500 healthcare companies on digital transformation in particular and prior to that, she had experience with clinical research and leading cancer research programs. She'll talk, I think quite typically from what's possible with a mindset of private sector organizations. We'll then hear from Victor, chief innovation officer at Humanitas leading private provider from Italy, but with more than provision as part of its business.

There's also medical education and this very interesting AI center that Victor is leading and has launched about 50 digital health applications and products since the beginning of this center. Will talk about workforce efficiency but also about other benefits. And we also have the chance of having with us, Betul who's got a wide range of experience across digital and AI transformation. She's currently a board member and chair of the innovation and sustainability committee at Ypsomed and she's also at a DuPont Sustainable Solution as an advisable member focusing on digital and she has a wide range of prior experience across pharma lab and otherwise. And from L.E.K. we have Klaus, our global digital health lead who spend most of his life trying to implement digital solutions in hospital and other provider organization and will talk to us a bit about the topic as well. So thank you very much. We intend to talk reasonably shortly about these different use cases to have wide opportunity for the audience to come in and ask questions and topics. I think Klaus, you come next. Do you want...

Klaus Boehncke:
Perfect, so let's kick it off. We're of course talking about artificial intelligence and I think one of the most important aspects is really how fast this field is developing and everybody of course, has heard of Moore's law, traditional computing power doubling every 18 to 24 months. And it's important to see, as you can see on the slide on the right-hand side, that the trajectory has really increased. This is on a logarithmic scale and so we now see the capabilities of these AI models developing much faster, usually doubling capability every six to three months. So it's a field that is accelerating very, very quickly.

What you can see in the middle there is an organization that regularly does so-called intelligence quotient tests with AI models. And just by looking at the orange bubbles there, you can see how much open AI by way of one example has really advanced over the last sort of year in terms of IQ moving roughly from around 80 to 110. So we can see that these models are getting much, much smarter and we can also see based on the color that they're getting more multimodal. The black color is really talking about different modes, voice, vision, language, whereas they used to be very much language focused in green and vision based in blue.

If we look at the different areas of impact, we can structure this in many different forms. One way is to look at by the quadruple aim, looking at outcomes and quality, looking at efficiency and cost and looking at the experience from a patient and provider side. And of course, then looking at the continuum of care from general care to specialty to diagnostics and emergency. We typically see that the main application areas historically for AI have been in imaging. I believe there's now over 700 algorithms that are are FDA approved. We have the efficiency and support category where LLMs are really hitting in terms of supporting clinicians. We now, I think over 60% of clinicians that regularly use AI on nearly a daily basis and we're seeing impact in reimbursement and coding. So the administrative areas. And if we move on to the patient experience and provider experience, there, we really see an impact and we'll hear some great stories today from our panelists around, for example, call center and how great the AI is at the moment already handling 60 to 80% of calls in one of the examples.

So looking towards the future, right? We can see a lot of these quotes that you see there are very, very recent. Y Combinator is now saying their recent cohort of investments, 95% of code is generated by AI. So very large number we can see in healthcare, as I mentioned already, 60, 70% of clinicians use AI daily and we can see... What you see in that curve on the right-hand side, in red is a consensus estimate of when AI will have an impact on the scale of the industrial revolution. That timing has now come down to around 2033. So certainly, you should never think about the AI examples, even the great ones that you'll hear about today as being the state of the art. It is the state of the art today, but it's very, very rapidly changing and what's possible in one year from now is already going to be dramatically different from what you see today.

Finally, we have an audience poll. We'd love for you to go through that. It'll take you a few minutes. You can do it in parallel while the panel is talking. This is the QR code. Feel free to load it up and it's a form so you have to basically answer all the questions and at the end, please hit submit. We'll share the results with you at the end of the panel discussion. So without further ado, let me hand on to Ola as the first presenter.

Aleksandra Spencer:
Thank you, Klaus. We're looking forward to your answers for the questions towards the end, but welcome everyone. Optegra, we are probably the fastest growing pan-European ophthalmic platform within Europe, five countries, 74 locations providing public as well as the private healthcare services in ophthalmology. However, all of those components that Klaus has mentioned are what enabled us that growth over the last years, is precisely very much digital enablement including AAI. We are strong believers of digital first approach to really drive the patient experience, streamlining the pathways and really driving the innovation in a healthcare sector. So back in 2023, we've decided to start the AI journey and that giving a rise to Iris, our virtual assistant. We have embedded Iris across multiple touch points of a patient journey. And that's very much starting from the initial booking management to then shifting to pre-operative assessments as well as the follow-ups.

So rather than as many companies also in the world has focused on a pure back office elements, we have actually put Iris at the front of our business, at the front of our patient's journey. So how does Iris work? Iris manages... It's a virtual conversational AI bot which provides the pre-operative assessment very much discussing all the medical challenges that the patient's having, medications they're taking, preparing them for the surgery. And that allows us to really very much release quite a lot of clinical time and effort out of clinics and really support and spend time with more complex patients. Similar for the follow-ups, really discussing post-surgery, their outcomes and ability to book them much fast into a second surgery. Manages all of that in six languages, but as we all know and those platforms, they've got capability of serving 130 different languages. So really supporting all of our patients and streamlining that patient experience.

So all of that. Now AI has been a part of a broader transformation of in Optegra that we've done the full digital end-to-end transformation across introducing one of the first clinics, introducing fully paperless environment for our patients and then underpinning with AI capabilities. So really building that infrastructure and a backbone to then enable to drive the value from AI. And what we've seen probably benefits coming through across four different pillars. One and foremost, patient experience. Many of you in the wrong world know typically in a healthcare industry, NPS score is around 60. We've got across the group, higher than 80% NPS score. But what is even more interesting, when we deployed Iris for our cataract patients, which are 60 plus elderly patients, we got a score 95% and that's continuously as we go at the moment. So far we've conducted 17,000 preoperative assessment and we continue and really giving that flexibility for the patients.

The second pillar in terms of the benefits is operational efficiency. We've been able to release more than 25% clinical capacity and really enabled to drive a higher volume of patients throughout the business and serve them much faster. The third pillar that we all love is the cost reduction. We've reduced our costs, administrative costs by 60%, driving something around 1 million of savings across the board. And probably the fourth pillar, the most important one, the clinical outcomes because why we all in a healthcare, that's what we focus on. We want the best outcomes for our patients. And as much as our business has been doubling the volumes of the patients we serve, we continuously have the best and the lowest complication rates across the industry and great visual outcomes for our patients. But all of those pillars, what's even more we've seen through the digital transformation is actually driving the revenue growth across our business and that very much looking across number of elements.

So if you look from ability and reducing non-shows for our patients, reducing cancellations, driving the conversion of the patients from the initial touch point with us to actually coming to the surgery. The second element to drive the growth was around recurrence. So really ability and gaining the competitive advantage across the landscape to be able to really drive that growth and new volumes of the patients. And probably the last one is time to treatment. I think across all the countries, the challenges of how quickly we can get the patients the surgery that they need and we've been able to accelerate that and provide the full surgery within few weeks rather than 12 and 13 weeks in some. So we've seen huge revenue growth and we continue that journey utilizing AI but also looking what's further and beyond. Thank you.

Guillaume Duparc:
Thank you.

Great, Victor, thanks very much.

Victor Savevski:
Hi. So maybe just a brief introduction. So whoever doesn't know Humanitas, it's one of the largest private healthcare providers in Italy, mostly based in the north. So we have hospitals, research center, university, ambulatory centers. I lead the innovation in the AI at a group level. What does that mean for us? It practically means understanding where technology can actually accompany us as management and our physicians to better improve clinical outcomes. Okay, efficiency, sustainability and so forth. So of course, we're industrial group, very much in tune with building companies, building products, building organizations, management-led organizations. So physicians are not part of the top management, but it's engineers, physicists, economists. Our main role in the last 10 years has been more or less digitalizing various parts of the process. So as you said, anything from patient, clinical files, all of this digitalization process is still ongoing but started in the last 10-year wave, I think with the large majority of the group here.

2019, just pre-COVID era, the whole AI buzz and big data and how is this going to impact was definitely a big discussion point. We were very fortunate to have really support from our top management to embark on the journey to at least experiment and understand what AI is. And for us the real big questions that this part was looking at these main three areas. So hyper-synthesis of the last five years, we started practically doing research in AI, understanding how AI can help clinical pathways, clinical outcomes, very research-based, lab-based, let's call it experiments. Experiments that practically pollinated throughout the whole group. Having engagement over 100, 150 physicians at the time we won close to... In terms of consortium size to dates funding is close to 36 million together with all our partners. Meaning that this created almost a culture of why are you not doing AI?

Guillaume Duparc:
Academic.

Victor Savevski:
So the question was how much of this is academic, how much of this is research and how much is actually impacting the patients? Again, we're very fortunate, our physicians are very hands-on. We have a huge culture and quality and efficiency and what happened was the same physicians with us started asking how can we actually put use of all these things that we're building, that we're researching, that we're seeing that passed through our tables and how AI is really going to impact these three main areas. So I think diagnosis, they're bigger experts than I am and there was rush diagnostics here before. You're looking into standardized medical device, software as a medical device, CE-marked FDA-approved products and we've all seen that 80% is radiology, cardiology and imaging. So you need to have a clear strategy of how you're going to buy, how you're going to integrate, and how this is going to be adopted. With a very big question mark of sustainability, but I will not touch this for now.

And the main point here is really understanding the actual validity of the products out there because a lot of these products that are out there, you have over 2000 models today that are certified how they're going to impact patient care. The second aspect which is personalized care is actually in my personal opinion, a very local AI. So a lot of these models are built on huge, vast data sets, European global data sets. But when it comes to actually creating personalized care for a cancer patient, you are obviously struck by a bias of location, by a bias of country, by a bias of genes and genomics and clinical data. So if I see the five or six tools that are now coming into the clinic and that have come to the clinic for the past five years, we have support tools for our physicians in oncology.

So this is the main driver today, personalized care that can take hundreds of data points if not thousands of data points from various aspects of the patient pathway in the cancer center and basically create prediction models and prognostic models on patient cure. We do a lot of work in the clinic, but we also do a lot of publications simply to be able to confront ourselves and to liberally share what we do with our community. Because if you don't have a peer review, it's like if you didn't publish it, it never happened. But in a sense, it's also a way to get feedback from a community. So all of these AI models that we have in the hospital right now are also published publications and you can actually see clinical outcomes that are being improved in critical areas of lung, prostate, breast, hematology, and a series of other oncological diseases.

So this is a phase in between where sooner or later, we're going to see if you really consider what's happening, these are all the models that nobody in this room is ever probably going to come in contact with but are going to do the biggest impact in terms of clinical efficiency and clinical outcomes. And I always use this example, but in the 1950s, you had a hematologist looking at our blood samples through microscopes and people were sipping urine just to make our blood exams. There is no human being today that's going to prefer a physician looking at their blood exams and not a machine or algorithms looking at it. So these are all the things in the back end and all our MRIs today, PET scans, CT scans have hundreds of models inside that we are not even aware of and they're being deployed at an annually basis.

It's going to be common practice before we know it. So 2030 or 2050, I think it's so much sooner because we are so focused on what we can perceive by the media that at the end of the day, we are missing the real big picture of what's impacting healthcare today. The last part I think is where all the, let's say big discussions have happened and thanks to the work of many consulting companies, tech companies and so forth, we see a lot of emphasis on workflow. Workflow for me it's a matter of sustainability, efficiency, reducing costs, staff and so forth. So I don't believe in the replacing of physicians. I don't believe in the replacing of jobs. As I said before, it's not that hematologists don't exist before, they do a highly more sophisticated job today, back-office activity. Nobody goes to university to become a back-office manager or to add things on Excel.

So these are all roles that we can have more respect for the humans doing it and definitely give them more empowerment in the role. So anything in terms of call centers, back-office optimization, insurance companies. My objective is honestly that we implement all these models and all these projects in the company and nobody notices, in a sense that all we see is people moving to do higher level activities. And I think we have a very strong support from all our top management and honestly knock on wood, but we haven't had issues of adoption or resistance neither from our physicians, neither from our management. But speaking also to other big groups, it's really on how do we need to do it and how do we need to implement it rather than if we should implement it. And this is the full cycle of the activities in the areas that currently we're focused on.

Anything that is a medical device, a medical software today is outside of this process because it has a different kind of valuation. It requires a clinical valuation, sustainability valuation, outcomes valuation, and then of course how to integrate it within the whole ecosystem of tools, IT and technology. If I take the actual patient pathway, physician pathway or staff pathway, today with this strategy we can more or less comprehensively piece by piece, adapt technology that will in a way give us a pure advantage in the very near future but also in the long-term future. And I think the debate in most of these cases, but having discussions with a lot of you here and a lot of other hospital groups is what do you buy? What do you build? What's local? What can I actually share with other partners? And I think sooner or later, what we're all realizing is if we don't have a critical mass in the activities that we do, there's definitely not going to be advantages in terms of implementing this activity within a single hospital or single clinic.

And the more you go broad in a hospital, the higher the complexity of implementing these tools and these processes, more you go vertical into a single area, then definitely you have less visibility. But the problems to solve are more standardized in a sense. And today without getting into the projects, but we're looking to build our own intelligence. So a Humanitas intelligence that is based on small language models that are locally based into our hospital, practically trained on the last 25 years of data, whether it's call centers, infrastructure contracting, procurement documentation, and in order to provide literally models that are our own property and our own digital asset and therefore can give a huge advantage to various departments. We are starting from areas that come into contact with patients such as ambient, such as call centers, such as knowledge bases to move into higher complexity areas, moving into insurance, back office and so forth.

Into the clinical, today the largest strategy, I think, but this is for everyone, is imaging. Imaging is fundamental. There is a lack of radiologists, there is a lack of digital pathology, there is a lack of a series of medical professions. Again, I don't think if we need to do it, but we really need to understand that these areas are actually both clinical and tech companies because today an infrastructure... We had Siemens on stage the other day, these are highly deep tech decisions to be made that we're going to take with us in the next 10 to 20 years. And this is without exaggeration, raise of hands if I ask you, when was your DICOM last installed here?

Probably it's going to be in the 90s, 1995 or 2000. That's 25 years ago. So the decisions we're going to be making today, we're going to take with us for a pretty long time. So fundamental aspect of understanding that these new technologies are AI native, whether we like it or not. And the decisions here are high-tech decisions together with our clinical staff, with our top management and with our financial departments. So therefore the complexity is ever more interesting and intricate.

Guillaume Duparc:
Thank you.

Betul Unaran-Susamis:
Well, thank you Victor, Ola. And it's great to be here today together with all of you. And we see amazing progress with AI in healthcare is also shared by Klaus. And I would like to ask you, how many of you remember the Dr. Watson, the first AI tangible solution from IBM Watson? If you can raise hands. Quite some. Yeah. It has been more than a decade since we started this AI journey and it has exciting and we see a lot of developments. And from my perspective, when I reflect on the various roles I held over the last decade since we started with this journey, the impact of AI in healthcare is in dual ways. First of all, on one hand, we see AI is enabling us to improve efficiency across the healthcare value chain, bringing treatments and solutions to people's lives much faster and with lower cost.

On the other hand, we see AI is enabling us to personalize healthcare, improve effectiveness and improve people's lives. And Ola, Victor, you have shared amazing examples here. I will pick up some examples from my experience to make sure that we are really talking about real opportunities of AI as we stand today. As we all know, developing drugs takes many years and couple of billion dollars, it's very expensive. And when we look at what happened during COVID pandemic, we have seen companies like [inaudible 00:27:33] Medicine and Accentia bringing drug candidates in days. So we are really talking about moving from years to days, four to six days in fact in certain cases. And this is huge impact when we think about the research and development cycle in healthcare. Another area we have talked about in the previous session now is diagnostics. And this is very close to my heart.

Recent study that we have run at Unilabs with more than a hundred thousand people showed that AI-assisted radiologists were able to diagnose 29% more breast cancer cases and 24% more severe cancer cases by halving the workload. So 50% efficiency gaze instead of having two radiologists working on a case, we have one radiologist working on the case. And this is huge impact. And when we look at currently the FDA-approved AI solutions, there are 1000 of them and 70% focused on radiology. But another area as Victor alluded to in diagnostics, which is seeing great improvements and a number of solutions is pathology, especially after the approval of the whole slide MS scanners. And we see great examples like [inaudible 00:28:49] or Mindpeak delivering solutions that enable pathologies to identify cases much more accurately and in a more efficient way. And there are a number of examples, especially also when we think about the tangible cases in the administrative space like the note-taking or clinical documentation solutions like Nuance.

We see that they are in fact helping HCPs, healthcare providers big time and increasing efficiency by 50%. And there are also very innovative solutions that we look at when it comes to optimizing the footprint and the location of certain facilities for healthcare providers, hospitals and diagnostics players with a number of labs and collection centers and it's complex. We're talking about thousands of tests in many different locations to make sure that can improve TAT quality while we optimize the operations. Well, on the other, hand we see that the personalization of healthcare precision medicine is taking off big time.

We have great examples and one example I really like is Tempus, a company that matches people, patients to cancer trials and enables personalization of cancer treatments, a huge impact in people's lives as well as in healthcare outcomes and overall efficiency of the system. And when we think about the personalization, a very mature example, it's very close to my heart again given my role at Ypsomed is bringing different pieces of solutions together including AI, Ypsomed being one of the leaders in injection device space, combining the devices with ChemDiv AI solution and CGM, continuous glucose monitor providers, Abbott and DexCom enabling personalized dosing and continuous monitoring and ultimately, closed-loop insulin delivery systems which is impacting the lives of people in time-in-range big time.

And also delivering business impact for all companies involved in this collaboration to personalize care for people living with diabetes. And maybe the last one I would emphasize because we have seen great examples, especially in the AI-assisted chatbots and experiences is the aging space in elderly care where we see AI implementation, looking at the biology of aging, making sure that we are talking about healthy aging for humankind. Now great space, we talked about great examples, how are we embracing these opportunities? I would like to share with you few numbers from a recent study, global study of McKinsey.

More than 90% of executives say that they are going to increase their AI investment in the coming three years. And 78% of them, say that they are already using AI in one of the business functions and this number in healthcare is 63%. But the less good news is when we look at the enterprise-wide level EBIT impact, more than 80% don't see the impact yet. But there is hope. Close to 50% of executives stated they're expecting to see more than 5% revenue impact coming from AI in the coming three years. So we are at an exciting stage and we can ask why aren't we addressing the full potential of AI? We are facing some challenges and I would like to highlight some here. When we think about over-usage of AI, we're talking about dehumanization of healthcare.

And the most impactful solutions we see in healthcare, they combine in fact AI and the human insight like in cancer care with the compassionate and emotional support of physicians or mental therapies with therapist interventions or even the Da Vinci surgical robots relying on the surgeon intuition. And if we want to have impactful solutions, we need to continue focusing on this marriage and integration. And another one is data privacy and data protection issues that we have been discussing. And thanks to our colleagues in L.E.K. Consulting, we see that there is tremendous amount of healthcare data and is projected to be at 10,800 exabytes in 2025 and growing exponentially. Well, when we talk about this much of data and when we think about healthcare as a trust business, it's very important that we focus on data privacy, cyber security solutions to make sure that we can create solutions sustainably and used in an ethical manner. And we need to stick to transparency and sincerity as people who run and drive these solutions to make sure we deliver real value in a transparent way to data owners, to people living with conditions and to people overall.

Lastly, accessibility. Only 10 to 15% of healthcare data is currently accessible. Most of the healthcare data sits in unstructured and different fragmented systems and that's why it's very important that we focus on solutions around better coding and integration to tap into the full potential of AI. Otherwise, we are really not doing justice, the data that we have around the world in healthcare. Last but not the least, it comes back to this room. When we think about AI's impact and how we leverage AI in our businesses, it's really not about the technology. There's amazing technology around the world and we see the acceleration in Klaus's light as well. We will have more and more and an accelerated speed in the coming years.

What is important is to focus on the change management aspect, transformation within our companies, making sure that we have a fully committed executive suite leadership who is bold and setting a big ambition and making sure that AI is integrated into the organization. And this is possible only if we start rewiring business processes and start training and rescaling the organization to ensure this integration of AI successfully in our companies. And only so we can address a long-term AI impact that can be up to $4.4 trillion based on the McKinsey study, which is bigger than the UK economy right now. So it's really exciting times ahead.

Guillaume Duparc:
Thank you.

Klaus Boehncke:
Thank you very much. Just a quick note, if you missed the QR code for the audience poll, let me just pull it back up while we have a discussion here.

Guillaume Duparc:
Excellent, thank you, Klaus. So I'm sure people will come in with questions, but maybe to get the topic going, you obviously talked about the technology is not the problem. I think nowadays we have clinicians that are pretty interested in the technology data and integration of that data is clearly an issue for a lot of the development of these opportunities associated with AI. Hence, from maybe more private sector organization and Ola, maybe you can comment here, use cases for you that are more tangible. Do they require deep data integration with your own patient data, looking up additional patient data if needs be or we in a world of there's a lot we can do with a limited data integration?

Aleksandra Spencer:
Yeah, I think that's excellent question. I think a lot of companies across the marketplace, you can see we all have challenges with legacy systems, siloed systems. And I think as much as AI has been exploding, what you see that a lot of that still happens very much in silos. And that is the challenge to really then unlock the value of the AI and the full potential for the patients and across the healthcare ecosystem. I think for us what one of the successes and other businesses that we've seen is that full integration end to end. So you're really creating healthcare ecosystem because I think as my colleagues mentioned today, none of us within the healthcare are working in isolation. We working in a healthcare ecosystem, both providing working with the private providers as well as the public and actually connecting the data and really driving value across that system and all the touch points either that from the patients or clinicians or the surgeons is really what brings the biggest value and a growth and upside for that from what we've seen.

Guillaume Duparc:
So data strategy is at its core, even if you have a mindset that's less clinical decision support oriented, but otherwise. And audience please comment in. But on another topic we are interested to gather your views and maybe, Betul you can come first on this is how do you think about AI and the competitive landscape? Or to frame the question at least we can think of our incumbents at an advantage, incumbents that try to adopt AI as part of their business model and probably change it versus new entrants would come in with new business models to start with. How do you think about that topic?

Betul Unaran-Susamis:
Thank you for the question. Well, we clearly laid out the disruptive power of AI in the competitive landscape. The question is not if it's about how the companies are embracing AI to make sure that we can really tap into its potential of improving or cost-based running or operations in a much more efficient way. At the same time, personalizing solutions and coming up with new business models to disrupt industries and to disrupt healthcare. So the AI is disrupting and is going to continue to disrupting competitive landscape. And when we think about the different types of players, incumbents, how definitely some advantages and when it comes to addressing big data sets, financial resources and having a brand reputation and trust base. But at the same time, as Ola alluded to, they are facing challenges in terms of legacy systems or the cultural challenges to adopting AI. When we think to smaller players, new entrants with their more agile and tech-first culture, they are facing less barriers when it comes to AI adoption.

But they are facing challenges in terms of scale, data access and also having the reputation and trust base to build their solutions to scale. So when we think about the most impactful examples and I have given one, we see it coming from the collaboration across, making sure that we can bring the best technology and use the data of big incumbents and bring it to people's lives in a scaled manner.

And Ypsomed has done this quite successfully with the closed-loop system that I mentioned, which is component of the device. Which is the core business of the company and also the solutions facing the patient and the physician, that's part of the suite that Ypsomed has and combining that with the AI solution, [inaudible 00:40:49] coming from Cambridge University Professor Roman Hovorka. Amazing AI solution for personalization of the dosing, combining it with the best-in-class CGM. So that when we think about this, this has a tremendous impact in the lives of people especially living with type one diabetes. But it also has the business impact for people, for companies being part of this collaboration. I certainly believe that we need to focus on the strengths of the different players and bring the pieces together to make sure we see bigger impact.

Guillaume Duparc:
So Ypsomed is a new business model in itself. So you have digital and AI I think at the onset of it, Humanitas with the AI center, you created something separate. Ola at Optegra, what does it mean? The AI team and you use third parties and you integrate your own thinking with third parties solutions, but what is your digital AI set up for your own organization?

Aleksandra Spencer:
Yeah, I think that's quite important. Quite often we still... What you see across industries, people building this enormous teams to really gain the capabilities, gained the expertise. But I think [inaudible 00:42:09] said that that's what we very much also done and seen across the industry more successful that you have to identify the strategic partners, the right partners who believe in the same culture, the same change, they are on the same journey to really complement you and help you to innovate much faster in more agile fashion. So a lot of things that naturally in a company for many years you don't have that change culture. I think the other element that fundamentally, I think being in a healthcare setting is the clinical leadership and the change element and that's in this digital era. You really need that because typically as I would say within the AI journey, 10% is your technical challenges.

Actually you can solve it, 20% we've spoken a little bit around data. You really have to have the ability to connect, but 70% is that change in clinical leadership and change management. And you really need the surgeons, the clinical team leading that digital transformation with you and be at the forefront. And also remembering that the processes that you're changing are not necessarily the same with AI and without the AI and that's fundamentally different. What you're creating, it needs to be different and it'll serve the patients and clinicians completely differently. So within the skill set, ensure that you've got the technical expertise from your partners and show that you've got data in the right place, but then ensure that you've got the right business and clinical leadership to actually make that change happen.

Guillaume Duparc:
[inaudible 00:43:39], if you wanted to add anything there and from looking... And I'm sure you're interested to talk briefly about the survey, hopefully you have enough answers, but what return you need to justify an AI investment? If I distinguish for instance, I'm making the numbers up a little bit, but call center operations, very high volume can go through the AI cost saving 80%. It's a very clear use case risk of hallucination or impact quite low. And human call center operators make quite a lot of mistakes. Physician productivity with co-pilots or ambient listening, I think we are more in the 20, 25%. And if you're already a bit digital, probably a bit lower, is that enough in your mind as an improvement, given the cost and time and efforts to deploy the technology and let everybody come in on this, but at what threshold you start to think that that's a real use case and whether it's operational cost or otherwise?

Victor Savevski:
Definitely return on investment is the basis, I think of all our discussions in all boardrooms. And my management and my CFO asked me what the return on investment is in this. But it's definitely a very short-term decision-making in a sense that really think about the conversations, 10, 20 years ago on certain technologies, if you didn't have them today, it would've been a matter of survival. Okay? So if we look at the logic from here onwards, of course we need to predict and make some estimates of your ROI in the first project, but the substantial tissue of what you're going to change, you're really making technological decisions that are going to revolutionize your processes slowly in time. It's a boiling frog theory, you're going to put the frog inside, increase the temperature, sooner or later the frog's not going to be there. So with the same logic here, I don't think it's very clear to a lot of people that this technology is actually changing your substantial tissue of what it means having a technology aspect of a business and becoming a technology company.

We're so focused on human personnel, which we all have physicians management staff, but we don't realize that a huge part of what's coming in and what's already here is actually that. So decisions today are going to be taking in the next years substantially. And I haven't heard anyone that speaks well of the IT department, but I personally support the IT department because decisions were made in a certain period of time for a certain reason. Now you need to understand what new decisions need to be made with new priorities for a future, let's say decision. So I think definitely, I would put a lot more emphasis in the actual technological choices that all of our companies are making. It is being really taking superficially a lot of the decisions of what tech you're using rather than a lot more focus is being put of course on implementation so forth.

But I think the fact that it's so highly complex, the decision is simply being outsourced to a third-party provider, to a consultancy, to an IT provider, which is definitely a very high-risk decision to be made. And understanding the substantial literally mechanics of this, you can't predict the future, but understanding the dynamics of this is evolving. Definitely you have higher probability of making the right choices with the right budgets and then naturally the ROI is going to come.

Betul Unaran-Susamis:
Is there time for a short question?

Guillaume Duparc:
Yeah. Please, please [inaudible 00:47:43]. I'm interested if people have... I have a threshold in mind, but-

Betul Unaran-Susamis:
Let me think about return-

Guillaume Duparc:
You wanted a number response-

Victor Savevski:
25%. There you go. I'll take that.

Guillaume Duparc:
It's interesting and Betul should comment because it's difficult to have a sense of how quickly the technology cost is coming down. And so there is maybe less so now than six to 12 months ago, but there is a question as to should I invest now or will it make more sense in six to 12 months? And whereas you can't quantify all the benefits and I think one for ambient listening is this culture of capturing information and being able to over time understand better the decision making. And that aspect, quantifying it is probably very difficult, but that's a non-quantified benefit. But I do think you kind of need to have a minimum threshold to feel like you're doing it at the right time in a bit the right way. That will come-

Betul Unaran-Susamis:
From my perspective, it's not about a number, it's about what is possible. So I like to think about pushing for the boundaries of what's possible with AI. And I fully agree with Victor's comments when we think about the ROI discussion, we tend to also think about the very tangible cases which focus more on productivity improvement. But we need to also realize that AI is evolving from being a productivity enhancer into a transformative superpower. That's why we need to think about our KPIs also differently. Also focusing on what is the impact we're generating for our customers. These are people, patients living with very serious conditions, which is going to drive the business impact ultimately. When we think about the different cases we discussed about, it's very important that we focus on KPIs, measure them, but we push for the imagination of possibility with AI.

Guillaume Duparc:
And sometimes it's very hard to measure the indirect benefits. So I think we have some survey results if [inaudible 00:49:48] can put them on on the website link. I think they might have heard me hopefully. Otherwise, Klaus sees them on these iPad.

Klaus Boehncke:
Yeah, otherwise I can just talk to-

Guillaume Duparc:
Tell us-

Victor Savevski:
On an AI panel, the tech doesn't work.

Guillaume Duparc:
And of course, they will be part of a presentation that will provide to-

Klaus Boehncke:
We'll circulate, make available the results. Just to quickly highlight, we had how is AI used? Are you using third parties? Are you using off the shelf or are you training your own systems? And most of you, the 36% off the shelf solutions. Some have said... Actually about 13%...

Do we have the... No, not yet.

Guillaume Duparc:
It's coming. I think.

The IT department.

Klaus Boehncke:
Fantastic.

Guillaume Duparc:
It's very automatic as you can see.

Klaus Boehncke:
So you can see the results there. So mostly everybody using the off-the-shelf solutions, but there's some training of your own AI solutions as we've also heard from some of our panelists here. And in already 20% of the cases you're really connecting the patient data and the medical records. So I think that's already quite interesting. By the way, there's around 17 responses. There's probably around 80 to 100 people here in the auditorium at the moment. So around 20% of you have answered. I assume that means 80% have not really had big impact or not leveraging AI. But that just by way of background. So if we can switch to the next slide please.

Victor Savevski:
It's impressive you have 20%, which is not the general public. Of course, here it's top-level companies and investors, so-

Betul Unaran-Susamis:
Yeah, exactly.

Guillaume Duparc:
Yeah. Let's go to number two.

Victor Savevski:
Impressive.

Klaus Boehncke:
So on the number two slide, I'll start talking... There we go. Yeah, as you can see, it's pretty evenly distributed where people are applying AI, right? Whether that's in the front office, in the back office for diagnostic algorithms, for patient engagement. So it's quite nicely distributed, which I also find quite a bit surprising. I would've thought there might be more applications in the back office area, but this is quite interesting.

Guillaume Duparc:
But the clinical decision support that's beyond imaging and lab, which I think is harder, is also where there's fewer use cases, which that's more expected, I think.

Number three.

Klaus Boehncke:

Number three, please.

Guillaume Duparc:
Thank you.

Klaus Boehncke:
This is the ROI. Victor, you said 25%. As we can see, most of our respondents also thought between 25 and 50. Then you have the rest, a big block, 30% that is not looking really for any specific ROI. Probably more looking at the tool, what you were saying. Can you-

Victor Savevski:
So I wonder if that 29% is the... 20% guys building their own product. So good for you.

Klaus Boehncke:
Exactly. So the next page, please. Maybe if you can click down below where it says all responses, please. Then we can see a bit of a list of what people put in. Yeah, so you can see a bit of outcomes, efficiencies, patient experience. So a lot of different input here by text.

Guillaume Duparc:
Can you remind people what EP-

Klaus Boehncke:
Yeah. E is the efficiency. O is outcome. And then we have the-

Guillaume Duparc:
P is patient.

Klaus Boehncke:
The patient experience and the clinician experience. And then finally, if we can go to the last page, please. And we can see here what Betul and others have already said, change management is really paramount. It's a culture game. Everybody has to be brought along that journey. Traditionally for the past many, many years I've seen it's often the IT organization against the clinical organization. This is not that kind of setup right here. You really have to collaborate, you have to get the support, the technical, but also the emotional support of everybody that's involved in the teams. Right?

Victor Savevski:
Raise your hand, who said fear of being too early because it's one person here.

Guillaume Duparc:
Raise a hand.

Klaus Boehncke:
We had this discussion a little bit earlier. What do you think? Is it the right time to invest, is now as we've seen with all of you? Or would you say No, no, it's equally fine if you wait for two years and then buy the best off-the-shelf solution?

Aleksandra Spencer:
Plus I think it's a really interesting question because what we've seen, as you've shown at the beginning of this session that we've seen a huge acceleration across digital technology, AI. We now probably something a generation leaps every 30 months. What we haven't spoken a little bit now is also looking into the future. And then you've got probably the word not many of us heard so far, but agentic AI, which is actually taking that to completely different level. So if you really wait for another two years, that's going to be a completely different world potentially with agents working completely conducting autonomous decisions. And if you look at the research, it's almost saying they're 18 months away, so not that far. So actually what we've seen across the industry, the companies that do really change, they survive. But the companies that... As Betul said as well, that the companies that truly transform and embrace AI innovative technologies, they really deliver the volume and the growth into the future. So I would say don't wait. Identify the biggest challenges that you have and what value you can deliver.

Guillaume Duparc:
Thank you.

Klaus Boehncke:
Thank you very much.

Guillaume Duparc:
Wonderful. Thank you. I think we might close.

Klaus Boehncke:
Yeah, thanks to the audience for sticking with us for 15 minutes longer and taking part of your break to stay here with us. Thank you so much to our fantastic panelists.

Betul Unaran-Susamis:
Thank you.

Klaus Boehncke:
Thank you.

Aleksandra Spencer:
Thank you.