AI for Bharat’s Health_ Addressing a Billion Clinical Realities

Thank you very much for having me here. at this very, very prestigious event. I just came in from Mumbai in the morning, and what I see over here is, I mean, I think it seems to be the microcosm of the globe, in fact. So thank you very much. Yes, I think, you know, I take all these compliments on behalf of Max, and I think it starts and ends with the trust which is sort of reposed by patients at our hospital system. Today, our occupancy is at least 10 % to 15 % better than the next best player in the hospital system. And, you know, one of the things that I want to point out is, you know, AI seems to be sort of the buzzword, of course, today.

But five or six years ago, when we started our journey, we started bringing digital technology at that point in time to the core. And what you see today, what you experience, and, you know, you mentioned better outcomes, and perhaps… patient services. But that is what you experience. What you don’t see is the technology behind it. And I think that is the true test of technology, and that will be the true test of AI as well. When you don’t interface with technology, but the experiences are improved. Having said that, I think, you know, like I said, we started this journey a few years ago. We started by creating a common -size data lake for all the patients which have been through our doors over the last 15 years, and which are doing so on a real -time basis today.

Having said that, you know, these were our attempts. We tried to sort of create a closed -loop system, like Google, so to say, for our doctors and our patients. But, you know, we, like many people, faced very early, very big setbacks because we didn’t have the technology. Because when we used to do search results, we used to get zero results. in the search engine because it wasn’t sort of native to the language, and that’s stuff that we’ve been playing with. But having said that, I think the early days of AI are going to impact tasks rather than, although one is moving towards institutional, adopting it from an ecosystem standpoint, from inculcating it within the institution, so it becomes an intrinsic part of the institution.

But I think today it’s affecting our tasks. It’s affecting tasks of efficiency. You know, we’ve already started doing predictive analysis of beds which are vacant and available and so on. It’s working on safety measures. I think though the early sort of wins that we have, especially with respect to patient satisfaction of the risks and so on, I think clinical support, you know, it’s data collection, a lot of… time by clinicians was being spent in the past to collect data. now a lot of that data is being collected through forms which are in our apps today. And you can speak to them. It kind of collates in a particular manner. So the clinician actually spends less time in perhaps gathering history than in providing a little more value of the value chain.

Great. And I think, Max, I mean, as you mentioned, there is this data lake that you have created is quite ahead in terms of digital adoption. And I’m sure when you would be starting, and this is a term that we use and see quite a lot, that adoption of digital in a hospital. So is that what is also happening on top of this large data lake and the EMR solutions that you have created on, for Max, is there like a AI adoption wave that is happening? And do you think, like, when the digital adoption happened, things such as NABH, ABDM, many of these things started coming up, talking about policy, talking about regulation. In AI adoption, are there any challenges, any things that you see that can help in this adoption to be much more faster?

Or you see that people are just going all crazy on getting AI adopted in the hospital settings?

I think, you know, there is a desire all across, you know. But having said that, desire does meet reality. I would say more than occasionally. And that comes in the form of failures, which are welcome, sort of. I mean, we quite welcome it, actually, because the more you try, the more you will fail, and the more you will sort of have better outcomes coming into the future as well. So we’ve had a lot of failures, I can tell you. You know, whether it is the longitudinal data of. Patient or. looking at, you know, ICD -11 norms, you know, tagging our data with respect to the WHO. I mean, we’ve been failing left, right, and center.

We’ve been reasonably successful as far as ICD -10 is concerned, but I think 11, you know, most sort of layers that are available in the market don’t work. The ones which work are very, very expensive. You know, so we’ve started, you know, we’ve been in -housing a lot of this. So we’re toying around with it. I have no doubt that the speed at which we are failing and the amount of failures that we have, shortly we will run out of all excuses and failures, and it will be like Edison, right? You would have found out every way to fail, and I think perhaps the only way to succeed will be in front of us. So, yes, there is a lot of enthusiasm, you know, towards adopting it.

We see this as a future. I think everybody does. There is, we, of course, have to be very, very careful. Because unlike, you know, let’s say something like education, where if you’re imparting. perhaps incorrect information, you know, it can be resolved. But this is healthcare. I think patient safety, data privacy, these things are right up there. We have very, very little, you know, standard deviation possible in what we do. And so it requires a large extent of supervision, I would say. And perhaps it will continue to for years to come. Although it makes life easy for most, but, you know, at least from a clinical prescription outcome, it will require a lot more supervision to come as well.

Sorry. So is it now the priority for hospitals? For example, when the digitization adoption happened, it used to be a priority for CEOs that, okay, you have to make sure that all the billings are online. You have to have all the JCIA. You have to have all the discussions and UHS. ID created and so on and so forth. Is this a priority today, adopting AI at hospitals and as a KRAs for your CEOs or operators? Is that what, has AI reached that level today at hospitals?

No, I think clearly, clearly it has. And it’s really out of, I think, two sort of drivers that I find at least, and you know, this is at a very, very, at the outset level. I think one is the way the world, the consumer behavior itself is changing. I mean, a lot of the searches used to happen on Google, a lot of the searches now are happening through different platforms altogether. And the way they sort of seek, whether it’s their thought about your website or, you know, when they’re looking at, if you simply ask which is the best cardiologist in Delhi, you know, there’s a different way of people reading into that early and there’s a different way of people.

So you have to, whether it’s your collateral, whether it’s your digital. assets and so on and so forth. You have to make those changes. Information, I mean, if I look at ESG, if I look at investor ESG, how do I improve my ESG score? I mean, it’s like an encyclopedia out there, right? You just ask it any question, it tells you to do so. How should I present my annual report? How should I present myself? I think, you know, pretty much, you know, it is intrinsic to now everything that we do from that standpoint. Second is how can I use it as a tool to improve efficiency? And these are low -hanging fruits. I’m talking about the low -hanging fruits before we even, I mean, kind of, you know, absorb the entire ecosystem or create that ecosystem or participate in that ecosystem.

Make it a part of institutional habits, right? I think even prior to that, when we’re looking at it as a task stage, you know, how is it that I can improve? Now, you know, if I have a particular waiting for my patients, okay, how can I do predictive analysis of room availability? When I’m looking at discharge, how can I sort of this thing? When it comes to patient summary, how do I get, how do I unlock the time that my doctor spent on patient sort of history and so on and so forth. That all improves efficiency all improves outcomes see eventually the lens we are looking at it from is efficiency is accessibility is safety, is clinical support and finally to the experience I mean it’s quite a bit of breadth that you’re looking at AI

but a little bit more on generic terms everybody says that technology is moving very fast or things are changing so fast AI is also changing so fast and we also keep doing that like we want our businesses our operations, our sales also run very fast is more your internal feeling how are you feeling? how are you feeling? how are you feeling? health institutions moving as fast as the technology is moving even people in the organization be it doctors be it nurse staff all of them are looking at it and a lot of this is about India AI Summit as well because government is looking for educating people on how fast things are changing and we should all be ready for it so what are your views on your institutional readiness people in your institute and in general on this whole AI moving fast

so I think first and foremost also depends on you know the institutional culture we are very clear about one thing that we have to be more sort of circumspect about it than anything else we must go up the learning curve as far as AI is concerned things are changing very very fast we have close to 43 3000 healthcare workers which provide healthcare. You know, that means there are thousands, if not hundreds and thousands of work processes. For us to adopt AI in any task means, you know, you have to change huge amount of attendant work processes, even if this layer sits on top. And having said that, okay, is there something else which is better out there?

Is there something which will disrupt this further and so on? Should we wait for something to be adopted by and large to see, sorry, to see what the efficacy of that is and see what the, you know, see once it’s sort of established before adopting it. To me, look, having the first mover advantage in this is not going to do anything. But getting it right is. I think because we can’t afford to get it wrong. These are human lives, these are people. So I think there’s a huge amount of learning within the organization which is happening and I meet you know phenomenal people people across the board, okay, for various aspects. I think since the morning of today, if I look at my sort of schedule, 30 % of my meetings would be people, you know, from a technology background, pitching various sort of applications where our lives can be improved and outcomes can be improved and efficiency can be improved and so on and so forth.

But, you know, at the very least, you have to be very, very circumspect about what you’re going to adopt and what you’re going to roll out.

No, and I think you touched a very important point that we learned at Eka. We were earlier did a travel startup, me and Deepak. What we realized that in health, there is obviously innovation that people are looking for, but trust is the most important thing. I can bring a cool idea or there could be a cool way of doing a diagnosis at a clinic, but I as a person would trust only the doctor that I have spoken to or I have spoken to. been talked about. So there is this and that’s the reality that we learned when we started doing health that yes, innovation is definitely important, but trust is key. And I think Max has been trusted over the years.

And to be very honest, we also don’t know how to balance that out. The trust that has been created for institutions, for doctors, and now these technologies that are coming in, where it asks questions to the patients and gives relevant next suggestions. This trust factor is kind of getting a little sort of changed. Any views that you have, especially when it comes to patients, people trusting doctors to AI to institutes, any change that you see, and even doctors looking for AI solution and whether they feel that this is right now not as good?

So, you know, I’ll give you one example. At most hospitals, at least once every couple of months, you will have a patient who will come in, okay, with a pain in the chest. You do the ECG, and the ECG to the doctor seems sort of normal. He speaks to the cardiologist, okay, and the cardiologist says, okay, I don’t see anything wrong with it. And the patient is sent home, and he has a heart attack, right? Because ECGs, although they’re extremely, extremely common, okay, can be very, very nuanced. So, an expert cardiologist, okay, may be able to catch a particular movement there, okay, while somebody else may not. But even the expert cardiologist on a good day may be able to catch a particular movement.

able to catch it on a bad day not right now I’m not saying AI in its sort of this thing is complete but when a patient comes to ER okay I think it’s absolutely necessary to use that tool okay because that tool says requires admission okay whether the patient doctor sees it on admit him okay look by the end of the day you may admit 150 instead of 100 actual patients but don’t let that one go I think that’s the important thing you’re able to if you’re able to use this as assistive tool to augment your capabilities okay and I think that is what is emerging today you know I think it’s little too far out to say whether it will replace the clinician or not okay but I think right now clearly that is a very very essential tool that you can use and let’s start with safety before we go to efficiency or anything else you know so I think a very simple example like this okay and it depends it starts with leadership moves to institutional sort of habits okay to be able to adopt something like that change your work processes because the umpteen amount of work processes which have to change okay doctor when a patient comes where do you move when this sort of ecg report you move him to the cath lab okay and which is a 13 minute sort of this thing but that’s also preparatory time right you’re doing it within the golden hours you sort of move him into uh you know the icu how do you sort of interact with the doctor you have to call the doctor let’s say it happens at three o ‘clock at night okay the doctor the cardiologist has to come from his home and so on and so forth so the entire dance starts right okay but you have to make sure that you know you you can use this tool to err on the side of caution but i think at the very least that’s what you need to

and i think you touched upon um like these this complex healthcare process and uh when we look at it from a technology perspective this is what ai can solve for these extremely complex process that today there’s a multiple human touch point very simple such as doing an emergency call to a specific doctor with giving all the respective which today can be optimized, which can save lives. So that’s the sort of things that we keep discussing about during our board meetings and discussions. But a lot of these, and maybe health and non -health as well, what’s your view how the next five years, next six years, yesterday there was this conversation with Sam where AGI will be there by 2028.

What is your view on how next three years to five years? Now we can’t even say a decade, right? It seems like we don’t know what all will happen in a decade. But how do you see next three to five years changing in your hospitals or in general health care?

I think dramatically. Adoption of, and it’s not because hospitals or health care providers desire it, I think it’s becoming… absolute necessity for the country. One of the things, and perhaps one of the major things that propels our country forward is the demographic dividend. You know, the average age is 29, 28, 29, whatever. But make no mistake, 15 years down the line, it will be very, very close to the European age. And that’s the time people will require medical intervention. There just isn’t enough infrastructure and doctors available in the country. Okay, barely, barely sort of, it’s actually not even enough for the population today. I can certainly tell you 15 years down the line, there isn’t enough infrastructure which can possibly be built.

There isn’t enough money over here. Or, I mean, we’re just behind the curve a little too much, right? And if we have to solve this equation as far as healthcare is concerned, you know, you have no choice. But you have to, it has to be about predictive health. It has to be about, you know, sort of, before even patient comes to the hospital, falls sick, to be able to predict that he’s going to fall sick. and make amends there. Reaching out to people, unclog the hospital infrastructure, home care and so on and so forth. Okay, be able to replicate capabilities, skill sets of doctors to be able to take them to patients and so on.

I think all of that is a necessity. Without that, we will fail the future generation. So there’s no question of us. I think, you know, this is here. I mean, the future is here today.

And especially the whole vision of making India a developed country, we have to leapfrog. And many of these technologies can help us in leapfrogging the way you were explaining. But thank you so very much, Abhay, for your deep insights. I think we all love Max and the kind of work that you are doing. And we see more and more AI coming together at Max to solve for doctors, patients, and all of us. Thank you very much.

That is entirely mine. Thank you. Thank you so much. Thank you. Thank you.

as Director IT at the National Health Authority where he leads the technical architecture and implementation of flagship national initiatives including the Ayushman Bharat Digital Machine and Ayushman Bharat PMG. We welcome you sir. We have with us Ms. Padmini Vishwanath, Researcher at the WHO SEARO, Southeast Asian Regional Office, bringing a regional lens to health equity, digital health policy and evidence -based transformation in low and middle income countries. We welcome you Ms. Padmini. And last but not the least, we have Mr. Jigarth Halani, Director, Enterprise Solutions Architecture and Engineering at NVIDIA South Asia, a 20 -year technology veteran driving innovation in supercomputing, big data and AI infrastructure and a trusted advisor to government and industry on AI strategy.

I now hand over to Duy. Deepak to lead the panel discussion. I think we are short of space, so I’ll manage to standing high.

Thank you very much. It was a great session, Vikalp and Abhay just left. We have a short of time, so I will try to leave maybe 5 -10 minutes at the end for everyone to have questions. I would like to start this session with Dr. Gupta. Dr. Gupta, we were talking last night. You were instrumental in defining the whole first white paper around ABDM, how did it all started. There is obviously a lot of progress from when you conceptualized way back in 2019 -2020 to today. What do you really think has really worked towards seeing the reality and what are the challenges? How do you see going forward? How do this whole documentation moving from? Between patient and interoperability will start impacting the clinical decision making for the physician.

going forward.

Thank you Deepak and thank you Vikal for this wonderful session and giving me the opportunity. So it started actually in 2014. It was in BJP’s manifesto where I wrote and then in National Health Policy in 2016 and eventually when I was advisor to Health Minister. So you know, firstly we should compliment the ABDM team. There is no precedent. There is no precedent to create records for a billion population. How do you go about doing it? But I think people like Vikal and you every time you know you take a bold step, there are nurses who will say Bijli nahi aati, aap kese karoge. Today we have 860 million ABHA IDs. So I think if I look at the reality today and I know I am sitting on the right of the Director IT.

We have created the digital infrastructure. Now we have to leverage that to empower the people who are going to use it. I see a future where we will not have people using multiple schemes. That was our biggest problem internally. I can tell you why this got you know created and there are more reasons to but eventually technology will allow us to optimally use resources to clinically be precise in treating people remove redundancies and also my boss who is still the unit health minister we agreed fundamentally it will be tough to send doctors to relay they study for 12 years to make their life better not to go of course we want them it will take time to have that infrastructure where they can stay in rural areas but we believe that digital health digital solutions will be able to leverage this backbone that we have created to serve people in the areas where they need them the most I think that golden hour to platinum minutes to I think finally what I believe will be the digital health standards that in a minute you could get to what you need at least for primary care so I am very optimistic and we call was right the decade is not we used to talk decade at 2013 -14 now we talk three years max few months is better to talk So I think it’s a time where we should be really optimistic of the vision that we were able to build thanks to people like who had implemented.

You know, we had COVID at our hand, you know, when we looked at 2 .2 billion people, you know, getting vaccinated, not calling up people, just going to the app and getting it done. So I think the creators are in the room. The implementers are in the room. So ideators don’t need to worry much. Thank you.

Thank you very much. That’s very nice. So moving to your left, Nikhil. Nikhil, you guys have done a phenomenal job in, you know, deploying adoption of ABDM in public sector today. But we see a lack in the private sector, definitely. What have you been learning? How do you think this whole learning from ABDM deploying at a large scale PSU where obviously there is obviously massive, massive load of patients walking in and very limited physician and staff to support. Digitizing appointment has done a great way. do you see it going forward moving into private sector how do you see it going forward getting into the workflows even deeper which will really help better outcomes

which can be developed by IKAK and other health startups. Where ABDM created the federated architecture, where the model can go there and they can be tried. Because the simple making algorithm doesn’t make a solution in the health sector. Just now as the max safety is very much important. So what is missing in the foreign models, which is not tried in the Indian data, especially we can’t neglect the population, the rural population, small hospitals where most of the people go there. Where ABDM created HMI solutions, where we have access to the longitude records of the patients, where our Indian models can be tried, and where we can get the success actually. The ground is fertile enough right now to pitch in for the Indian startups to come in and try.

in your models especially because of federal drugs you don’t need LLMs you just need SLMs and some random models to come and do it so that the model cannot be biased because the biasness I can’t see only from the technical angle here you have to keep both clinicians and technologies so that the context data is available from across the India and across the population where the subject is the billion clinical realities and where and the AI model should be not only transactional they should be conversational where the literacy rate is very low so now is a fertile ground for the Indian startups to come in and show the brand value of the Indian startups so where I can see it.

Thank you.

Thank you very much Nikhil insightful you touched upon cloud infrastructure and we have Jigar here. So Jigar cloud infrastructure has made AI scale. We all are using, everyone is using chat GPT today. Infrastructure, sovereignty and trust are hot things. We’ve been hearing about these words for last five days like I don’t know n number of times. It becomes super, super relevant for health as we heard in the last session, the trust. How do you think the models or the companies building in India bring that trust factor so that physicians and the operator like Abhay would trust the solutions and then start implementing which will really help, you know, people like Nikhil in building those models for the country.

So I think it’s a deep question. Trust has many aspects if you ask me honestly, right. Trust in my language could be the most accurate results and I’m happy because I’m a fast moving guy, IT professional, right. So we are known for it. The event gets over today and Monday we are going to do it. We are back to work and we know we are going to hog again for next five days to make sure that we are something better and bigger right trust for my mom could be a very different storyline right because for her everything on priority line is health nothing else essentially right for me plus or minus 1 % 3 % 5 % 10 % is also okay for her nothing is right and trust for a mother who has just a newborn in her hand it’s gonna be completely different right so it I personally feel it has many years but a fundamental layer if you ask me honestly model builders what they’re trying to do is trying to still accumulate the knowledge which is still available on the web right what we haven’t gone back and that’s where I would borrow if India is achieving these numbers which I was not aware I knew about pretty well I have it from myself as well although I have not used it yet but a but I am a registered user I won’t myself get enrolled into everything but I am a registered user I won’t myself get enrolled into everything scheme that government is coming of it just to make sure that to understand where all connectivities possible essentially right but I’m not even a willing one yeah but but but I’m saying that once we have this data how do we make use of this data better so that I bring not just the context of India which is so important and what a couple is trying to do just on the language side of the story which we all understand that language is so important to us but imagine the you know the the the the environmental changes that I have from you know place to place and basis which the changes that I have in my body structure and basis which what medicine helps me better and so on and so forth right it has its own subsequent you know chain of things that that is it how do I bring that data more into the ecosystem thereby I make those model more and more efficient better and in the lingo of what India understand not just in language but also in the lingo of health which is important acting that particular like for example I come from Gujarat but I stay in back Bangalore, I know for sure that environment is not suitable to me, right?

And I keep sneezing for the poll reason, of course, many of the moment I go back to Gujarat, I’m absolutely normal, right? Whether it is extreme cold, whether it is extreme hot, it’s raining, doesn’t matter at all. I never sneeze. I think things work in Gujarat. I go, I come daily, I don’t get sneezing at all. I’m just another example, right? So how do I bring that data more into the ecosystem, number one? And then number two, how do I train those models more efficiently and serve them back to the users? So that’s one aspect of it, right? The second aspect of it is I think unlike language, in the healthcare, we need a very large momentum of citizens to participate and help us to have a lot of feedback ecosystem in what they are pursuing from these models which they are inferring it.

Right? Thank you. like, for example, in your solution, which I’ve seen the demo because now it’s been a number of times in the demo booth I’ve seen this. If a patient is talking, right, and, you know, going through your recordings, let’s say, which he or she has just done, for him or her, it’s the most important thing. For the doctor, it’s like the next patient, right? How do I go to the next? But the patient will definitely go back and check the recording. Patient will definitely go back, as we all do, and for the rightful reasons. We check the second opinion with the doctor, right? But that information is only with me. What did the second doctor told me, right?

I check with you as a doctor, and then I say, all right, it’s a big operation. I should take a second opinion. And I go to her. I take another opinion, and then both they say the same thing. And I then still Google, right, and I take the opinion there. And I say, you know what? Looks to be that I need to get operated. But we’ll wait. And then five days. It’s free consulting. Four days later. will come to the doctor. There are four questions. So I think the user also need to put the feedback back in the ecosystem by using these models and then getting democratized. I think that’s how the trust layer is growing.

This is at a very high level. Policy level, things are going to be very, very different and I’m sure it’s a topic by itself. Some other day we’ll work on it and I’m

Thank you very much. This Google doctor has been very, very popular in clinics. When we meet a doctor, they hate it. I have seen a board many, many times outside the physician’s cabin. No Google doctor, please. Okay. So we discussed, so next question to tell me, we’ve been talking about private hospital infrastructure. There’s a mass of high quality infrastructure available in the country with really great physicians. On the other side, we have PSC, massive pressure, less number of physicians. How do you think, you know, builders, when they think of building solutions for both of these perspectives? Should they think of a single solution? So I think the answer is yes. So I think the answer is yes.

So I think the answer is yes. What do you think of two different solutions? What do you think how it’s going forward?

Yeah, so at PeoplePlus, which is an initiative of Aikstep, we do a lot of analysis on what are the adoption trends and for high need populations. So basically for people who are building in healthcare, education, agriculture, right? What’s the uptake? Who’s building what? Who’s not taking third -party solutions and trying to build internally? And there are a couple of points that have emerged in that thesis. I think the first is because AI is meant to be personalized and context -specific. It can deal with multilinguality and voice. There’s firstly a lot of opportunity to bridge some of these inequity gaps that exist. So the first thing is that today you can imagine as a builder solutions that are in some very, very regional, low -resource languages for the different beneficiaries.

And you can design them to be voice first, which in a way is inducing trust because now they are speaking to someone and they’re just not reading or… an answer and they don’t know who’s behind that solution. So the first aspect is that AI is meant to be personalized. So when you’re building solutions or, you know, I’m going to go a step further. I’m going to say it’s beyond a solution. It’s a transformation. You can create very customized transformations. That’s number one. I think the second thing over here is that when, you know, because it’s a very fragmented value chain, right? In the case of healthcare, like someone is paying, someone is using the technology, someone’s ultimately benefiting from it.

What we’ve realized is when you’re designing these transformations, a big part of a builder’s journey is not just making the tech stack, but spending time with people who will be adopting it, educating them at different levels to explain how this tech could get consumed or improve their life, right? So many times, and I mean, there are 700 plus healthcare startups in India who are doing all kinds of pilots and demos right now. And what we’ve realized… is that the demo phase goes really well. Like three months, six months. Because there is adopters who could be hospitals or other institutions sometimes play from a place of either fear or hype. Like I want to be aware of what’s going on and I’ll do the demo.

But after three months, this is just going to be a side window on my browser which I never go back to because it was never thought of as a solution that I would embed into my workflow. So you have to think from the start of this as a journey and not just a one -time switch. That I get that one -time contract or that one -time demo and imagine that that will convert into some kind of impact. Now to build that trust, it’s very different in a private hospital which is maybe much more urban, much more aware of what’s going on versus a PHC, right? And the people in the PHC. So I think a tech stack and the solution is one piece of it.

But when you’re designing the transformation which comes with education, awareness, trust -building activities, creating safety and maybe feedback and evals that may be a little bit more make sense for a PHC versus a hospital which might be very different. So when you’re thinking of the… Transformation stack, it has to be very different. And transformation is about much more than tech. And I think that’s where people should be spending a lot more time as builders. It’s just not about cracking that first pilot or that first deployment, but saying what will it take to go from pilot to population scale, right? What will that take? Because that is a very different journey. That’s a systems journey. That’s not a tech journey always.

No, that’s super helpful. Continuing the same discussion, Dr. Gupta, when you look at policymaking, do you look at these two segments very differently or you think when you look at policy like private sector, PSC, you think health is one single sector or do you start defining, saying, okay, how will it work in public, how will it work in private sectors?

So if you look at the national health policy, this is the first time where I actually wrote the line both for private and public sector. In 2002, it was mostly written and even implied that it was only meant for public sector. I think if you really want to deliver care, you have to break that. barrier between private and public that’s how you will deliver care when patients has a problem it doesn’t see whether is the private or public should I get it the first hospital they get it so I think that was the thinking behind it and that’s what the policy is like

oh that’s great learn something I move on to Padmini from your regional vantage point how should AI system be designed differently to reflect the diversity of context capabilities and care relative across countries

yeah thank you first of all thank you so much for having us today WH was very glad to be representing the work that we do and you know listening to my co -panelists it’s interesting to hear about you know the the importance of tailoring because interesting and a little how do I say anxiety in using for me because the you know the work that we do is on the other end of the spectrum which is how do we create norms right how do we create norms and normative guidance to ensure that AI is equitable and moving in the right direction. And so I’ll talk from the regional perspective. And, you know, of course, we work with eight countries across CRO, and all of these countries and systems are at very, very varying levels of digital maturity, right?

But what we often find is the AI tools are developed for the most advanced, most connected tertiary institutions. And then adapted later for the most more remote settings, right? But we are finding that some of the countries, you know, pilots are looking at reversing this logic, which means that we start developing readiness frameworks for the most remote settings, understanding the frontline capabilities, device availabilities, all the factors that matter in AI readiness. You know, developing a framework, a framework for that level of remoteness. and then scaling it. And we do see that in contexts where we do that, there is higher provider trust, there is more equity. So I think that is, from our experience, we feel that maybe we need to slow down a little bit and look at how we can modernize existing legacy systems rather than kind of building on and adding new systems.

Yeah, I’ll stop there for now.

Please, Annie. So maybe starting from here.

I’ll go first. I think voice. It’s a common factor. I think it is horizontal, not vertical, but it’s very, very important for the country, right? If I understand what Tamil doctor is speaking with the patient and convert it into Hindi and have that deployed in Delhi and Gujarat, I think I’m home essentially, right? I’m solving many problems for years together that has been prolonging in the country essentially. I think it’s by itself is a reward to the country and we should be fully liberating it. One thing that I’m very happy about is the mindset change. You know, that’s going to be the biggest thing. It’s not a technology problem. It’s a mindset problem. And that’s what I’ve seen every single person, you know, talking to, they started more believing in the fact that, you know, the time has arrived.

I will tell two things. One thing is product. So I am very happy that a lot of discussion is going on about AI. So for any technology, anything to encourage the public, the thought process is very much important. That impact submitted created that impact to discuss things on AI basically. Now everyone will discuss on AI, like you beat a very rickshaw puller to the CEO, that discussion is happening. That is very much important to build systems, that thought process. Second thing, I visited few of the special start -ups. So very happy to see some start -ups are doing really great, like Eka scribe. So where the TVDM can use basically small clinics to reduce the burden of the clinicians from the non -clinical worker work.

So and there is one company and they are doing very great work on data anonymization. Because for many people, they have models to train after the advent of the technology. DPDP act so the data privacy and patient consent is very much important so they are working very really good in India so I’m very happy to such companies are there and they’re doing really wonderful so I’m very fed

I think for me it would be the emphasis on AI ready data systems because this is across sectors everyone is realizing that AI is only as good as the data for the model and application layer that it has access to and I really want to give you guys credit for that because you are pioneers in terms of putting out data and making it available to that MCP server that came out in fact we cite that as an example we are working very closely with Mosby right now and they want to make their statistical data sets available to the world they have put out their first MCP about a week ago but just the fact that you are you know institutions are just not extractive when it comes to data but they want to give it back so that others can build on top of it is very important so in health it’s crucial that happens because otherwise there is no personalization happening.

So I would say I think so far we have looked at a lot of quantitative measures of adopting AI in health, diagnosis, accuracy, number of patient visits, et cetera. But I think this time around we are seeing more discussions around the qualitative dimensions, right, empathy, dignity, care. And, you know, it’s interesting because in one of the pilots we are conducting on palliative care, we, you know, we didn’t even think about it, but a caregiver and a patient, palliative care patient visiting a nurse, you know, that’s their only source of human connection during that week. And so how does AI kind of change that dynamic of caregiving, right, in those little moments they spend together in the clinic?

So I think. The increased conversation around this. and just acknowledgement of not just the quantitative but also the qualitative dimensions is something that I’m personally really looking forward to.

Thank you. The objective of this question was not to get the promotion for Eka, just a disclaimer. I will… No, but thank you. This was super insightful. Audience, any question?

Sir, I just had a question. You said a voice language is not as new. So that mostly 90 % of them are on the cloud. So that needs to be on the edge only or on the cloud or hybrid?

No, no, of course. I think… Do you use ChatGPT? Yes. One of the servers I hosted over in India. No, I’m just saying there’s cost factor is also there and they have data privacy also there, so… The moment you add cost, as long as it is in India, I think we are home. I don’t think so we could be… Ever cheaper.

So I was just… I’m asking a suggestion from you, so like what model should, like someone who’s creating such solution for voice and translation, multilingual, let’s just target 22 languages. So where should the MCV or the influencer or the activity server should be hosted? On the edge, on a gadget, like a mobile phone or something, audio recorder, or, you know, hybrid?

I would say, I would say it depends on the use case. If you have a very particular use case, very tiny one in a remote place, edge would be the solution. You don’t have a choice. You will lack behind the connectivity and few other aspects as well.

It will synchronize once a month or once a week or once a day?

No, voice is something you need to have connectivity in play.

Okay.

You can’t be having offline things. That’s my view at least. People are trying. I think Saboom had something on, on the device. 90 % we should go for.

Connecting with the cloud or the server?

Yes.

Even if it’s a local India hosted server?

That’s correct.

Hello everyone, we have seen a lot of stalls in the expo showing AI powered documentation and diagnosis. I am a dentist and I am currently pursuing MBA in analytics. So I am curious how far this AI, Indian based AI tools are relying on Indian data rather than global data sets.

Depends on what they are claiming that’s first. And the other side I also represent the Mayo Clinic strategy in India. So as Mayo Clinic platform we are opening in partnership with some of our data sets to look at but also collaborating with hospitals inside to leverage each other’s anonymized data sets. So I think important. Thank you. point to note is the culture of data is missing. I mean, we still have to get the culture of data to get to have those AI systems that are based on Indian population. I think this is still far away. I think with ABDM sitting next to we have 860 million of IDs, but if the number of records on ABDM, if you check, they are not what we want to be.

So I think we’re still not there in terms of if someone makes a claim, be careful. Thank you.

That’s great. We talked about what we really like, what we’ve seen the change fundamentally. But do you guys also think there are still few items we’re lacking behind as a country, as a health, where we should have been already seen? Or you think we are on the right path? And if we are on the right path, then what do you think would be the great outcomes in next year?

My answer is very frankly, even at the cost of obesity. See, the issue is not about the usefulness of technology or the use case. It’s about ethics and doing the right things. Most of the people are not using, not because the UX, UI, technology, outcomes, everyone knows that. How many doctors would actually want to tell what they charge for a prescription, how many prescriptions they may, and why they write three antibiotics for one case. So I think it is about regulating that unethical part, the way they were able to crack it, you’ll be seeing the mass adoption. The challenge lies in the medical practices and medical ethics, not on the solutions per se. Otherwise, we would be the most adopted nation in terms of digital technologies.

The great, last night, we were having this conversation where in China, I was surprised to hear this, that in the real time, when a physician, is writing a prescription there’s a data going back and if there are errors it’s coming back and the doctors are getting flagged if they continuously do this thing and then that’s the way one way of controlling what you really said and you know think of us we are still in metros literate but think of people in tier two tier three having three antibiotics at the same time i have seen in bombay a chemist saying as you throw my religion yeah we pop up so i think it’s an issue about practice medical practices good good pharmacy practices good prescription practices to follow i mean you could have given a cold and a cup syrup that would have made him money too

i want to give as if you had a point just want to add and she asked one question how many models are training on indian data so you said now that where we are lagging behind so we want to say like the behavioral change is very much important so we have solutions even And we gave very, with CDAC, we gave one e -shift flight, which is almost free to small hospitals. And all the government hospitals, including Ames, having the HMI solution where they can create the language records. But some of the docs are not ready to do, because they said that we are very much accustomed to writing on the paper only. So still they are doing, and we are accepting.

So where we are losing the context data from the major public hospitals. So where we need some tough stance, because I am now working in National Tadati, but before I was in Railways. Now Railways totally stopped physical prescription. Because they took a decision that no more physical prescription. They are doing only now online prescription, everything, even lab record, everything integrated. So they took one decision. They retested. So we need some tough decisions, and also we need some behavioral change, where we have to go for creating language records. then only we can give some context data to the Indian startups where our models can be deployed and trained then we can get rid of it.

Thank you very much you have been a great panel thank you very much for all your insight and I am sorry in the interest of time we will have to wrap up but before we close this session a sincere gratitude and thank you to all our panelists I request Deepak to just present a moment to from our behalf thank you Thank you. Thank you.