 A very good afternoon ladies and gentlemen. I know it's going to be a little tough, but after that sumptuous lunch we focused on the stage and we are in fact discussing an interesting topic because what we have seen in the last 15-20 days, I mean the way our finance minister has been calling all the fintech founders recently and all the regulatory which is happening in this space. So it's very tough to keep innovation going at fintech companies when the regulatory environment is difficult. So that's why we thought to speak to these esteemed panelists today. We're going to talk about how difficult it is to look at innovation in the current macroeconomic situation and also talk about how technology plays a major role in a sector where of course money is the governing factor. So I'll just give 30 seconds to each one of you to just quickly introduce yourself and then we'll get into the discussion mode. Sure. I am Prajak Deval Singh. I am CTO at Tertlement. I've been an entrepreneur myself before coming to Tertlement. I've sort of worked in a variety of spaces from search and crawl to building developer platforms and locations before doing fintech. Tertlement essentially is a B2B2C platform for insurance distribution. We are a pan India, both general insurance and life insurance. We built a large agent network of more than 3 lakh agents who distribute insurance to their customers using our platform. So that's it. Thank you. Thank you for inviting us. My name is Veeradhar Yasa. We are a full-services platform for specialization. We have automation, distribution, critical lighting, post-services and collections in the platform. So what can we operate in each segment of the industry? Hi. Very good afternoon to all of you. So my name is Kaapik and I'm CTO at Karidbee. So what we do at Karidbee is we provide fintech services for all the customers all across India. It's more of a B2C platform. We have been operating in this space for almost 7 to 8 years. We have been a part of the digital journey of basically fintech when it started and we are still ongoing. I think we are having a fantastic journey till now. And going forward also I think we hope to basically ride the fintech wave that's going on. Thank you. Thank you. So first and foremost, as I was discussing earlier and that's the topic which you want to discuss today, I want to bring your focus on the current regulatory environment. And what is your take? I mean where does this hold this lead to the fintech sector overall? I want to understand your opinion. Let me start with you. Look for us in the insurance category. I think the regulatory changes have been a lot of positive regulatory changes in the last 5-6 years, frankly speaking. The regulator is very consumer focused so protecting consumer benefits is what their focus has been. But they have been also expanding in the private sector, private categories to expand the category to go deeper in India. Especially tier 2, tier 3, smaller towns, lots of innovations. They have been triggering lot of innovations. So in my opinion regulatory environment for treatment has been fairly good, very encouraging. There are I think the challenges from a technology perspective what happens is that it's such a evolving space and the changes sometimes could be significantly larger. So from a tech you need to respond to those fairly quickly to get real benefit out of the changes which are happening. Positive and negative both. And I think that's probably been the biggest challenge from an insurance regulator perspective. And then there are other sort of regulatory authorities who have also had significant changes in the way sort of streamlining the way the distribution is happening, the way sort of sales is happening. A lot of, as the digital sort of medium for distribution is increasing, sort of how do you bring transparency in that, how do you keep data, how do you maintain data with things like DPDP coming in now. All of those have sort of brings in lot of, the pace of change is significantly higher for tech. And I think that's probably been our biggest challenge in terms of how do you keep up with that and how do you be ahead of others in this category. Keeping the tech employees at the toes. I think we have a very progressive regulator right across insurance, across banking and capital markets. One of the important things with regulation is that at the end of the day it's trying to protect the customer and ecosystem. And avoid any large kind of imbalances in the ecosystem and also protect individual customers. It covers this entire spectrum and we have to respond to that. And typically if you are principle based and there are three or four of those, one is that you don't enter in normally. Second, you are very transparent. And third is that whatever you are doing, you have to be upfront about it. And this covers insurance, this covers banking, this covers digital funds and stocks as well. If you stick to these principles, then as an organization, anyone who is sort of trying to play outside these principles tends to be in trouble. So I think that's the way I look at it. Yeah, I think pretty much I would be echoing some of the thoughts what has been already mentioned. See, as we see as and as was mentioned, so the regulator here is pretty strong as in terms of basically in terms of the regulations that are there. I remember a paper which I used to read in my university days. It was called the more you don't syndrome. So what do you do basically when you have a gap in the divider and you don't have any directions what to do that. If you look at, for example, in India, people who do all sorts of things, but let's say take a country, for example, like Singapore, they would only follow what are the instructions that have been posted at the divider. If nothing is that, they wouldn't take a U-turn, they would just keep going straight. So similar to that aspect as in here, at least in the fintech space, there are certain regulations which are there in terms of basically the space that we are working on. So we have to basically, the RBI is doing all the regulations and with basically good intentions of the customer in mind. So we always have to keep in mind that customer is the number one priority. If you are able to keep it in mind, I think the rest of the things will fall into place. If you try to stretch the rules basically beyond normal reasoning and then English is a fickle language. So you can interpret one word or a lack of a comma or things like that in multiple different ways. You would definitely have a backlash effect later down the road. Even the regulator basically takes a consensus of that fact and the fact is that they do lay a lot of breadcrumbs along the way before the final ruling comes. The idea is that when the breadcrumbs are there, we have to get an idea as to what is happening. So all the fintechs kind of know this because all of us are basically interacting with the regulators on a, maybe not on a day to day basis but at least on a monthly basis when interacting with them we get a feel, a gist of what the regulator understands. But if you are blind to that aspect then you need to basically reap what you sow. Then SROs are there, so there is a collective thought which comes up and based on the collective thought you need to understand what the direction is as in that is there. And generally keep the customer in mind is what it is. Because finally if you look at it, I wear two hats. I am probably an entrepreneur but I am also a customer to basically a lot of other services. So I would want my data also to be treated in the best possible way and I also should treat the customer's data the best possible way. So that is on the regulatory aspect. The tech aspect is basically given fintechs obviously we have been able to respond very fast for all the basically concerns that have been there, whatever the changes that needs to be done on the technology platform. The idea is that basically you need to plan ahead and then have the levers in place at least in the tech aspect so that you can what we call it change the levers as you would say. In terms of tech parlance it is more like change the DB flag or go and change something here or there and then so that basically whatever intended action is supposed to happen you can do it. A good example is recently and another thing is that the transparency aspect which was bought, a good example is the KFS document. Earlier as in the fintechs we were never showing a KFS document but now we have to clearly spell it out to the customers the facts, that is why it is called a key fact sheet. So these kind of things are making it more better to the customer. The customer is also in a lot of ways well informed because these days communication you know one WhatsApp group is there and then the groups are linked the message forward and generally any news gets better across the world in less than a minute. So it does not really make any sense to hide anything. So at least from our side we have been very transparent in a lot of things so that has helped in basically continuing our business without any major hiccups with all these aspects, regulatory aspects that are there. And we are also a bit conservative in terms of interpreting the language that comes out by basically getting multiple legal consultations from the lawyers. So you get multiple inputs from the lawyers and all those things because they will be having a bigger collective thought across other industry peers and even amongst the peers connects etc and all those things. And they will also have the general idea and they will give us a good view and it generally makes sense to follow all these thoughts in a proper way rather than interpreting it and stretching the rubber band more than what is expected. Thank you. So part of the regulatory changes which keep on happening and the challenges there. There are a lot of cyber threats of course involved in the 20th phase particularly and I mean how you as tech heads of your respective company keeping it in the safer side of the consumer and keeping the data particularly in the safer hands. What kind of steps you have been taking and what kind of investments you have been doing in the tech domain? So security is kind of like the number one what is that aspect in our organization. It has been probably depending on the size of things that could be different priorities but it needs to be at least amongst the top one or second basically priority right depending on the size. Because after a certain point in time you know there are only few existential threats which an organization can have and data is close to a basically maybe not a full 100% this thing but it is very close to kind of a threat over that right. So security is an important aspect that we look at. The other aspect is that basically we have been as we come into the RBI aspect and then the regulator has had a lot of guidelines with respect to IT guidelines and all those things. But recently the certain and all these other bodies have come in which have expanded the overall scope of the activities from basically just some of these regulated entities to basically pretty much the entire domain. And even the DPDB bill that is that has kind of basically brought in all the organizations which deal with data in any ways basically under the pursuit of under the capitol of basically data security and basically the repercussions that can have with respect to not taking it in priority right. So obviously it is very important right and again here ideas begin to be conservative in terms of what you do. So generally as in data security has got well-lit rules right as in terms of handbook that you have to follow. So the idea is that you need to have a handbook and you need to follow it to the top 99% of your issues are solved right as in so a process handbook or a security handbook is very much important. And let's again try not to dilute it over there so you would basically not have any basically leaks or anything later on the road right. So that is one thing obviously technology also plays an important role these days in terms of SIEM which can start looking at all these logs that are being generated across your various systems right. As in for example your APIs and then your data access patterns and all those things. Logging becomes very much important because in case of an issue or an event you need to be able to audit what has happened. So auditing capability needs to be done across the system right and these days the audit logs are basically forwarded to the SIEM systems which can give you an alert as the system as they even happens. So I think that's very important now. I think as far as security is concerned I mean there are tried and tested methods right and first and foremost you have to be really good at implementing those. Second is you understand your business and your systems better than anyone else. And hence doing a constant threat modeling right with an intent of protecting the customer right is where you have to invest continuously. Now even nothing is perfect right so there will be holes everywhere. But overall if you have multiple layers of security I mean it sort of becomes like a Swiss cheese right. There are holes but then the entire cheese if you look at it is a solid block I think that's through it right. I think that's the way to sort of look at it right. You need to have sufficient guard rails at each layer to make sure that nothing passes through right. And then you know you have to respond really quickly. So being very systematic you know establishing and implementing the tried and tested methods very well and then making sure that constant threat modeling right is where you should focus on. Hello I think with the category that we operate in all of us definitely a lot of sensitive personal data for customers which is just required for the business that we are in. There is no sort of getting away with that. At the same time because there is a sensitive data that we are operating with we have to be very very careful in the way sort of we are handling data. And I think it's a combination of tech and the checks and balances that you create in the system at every layer. So at the critical sort of junctures where the data has to be handled you need to have good controls and good processes both automated ways of looking at it. And also the manual checks and balances to sort of make sure that make sure that the whatever process that you are putting in place is actually effective and working. And it's basically a edge that you are working all the time. On one side you need the business to operate seamlessly. On the other side you have to make sure that the data is protected. There are standard sort of card rails. I mean if you just follow just the basic standards well I think large part of it is taken care of. After that it's basically you have to do a little bit of testing here and there, do some on the spot testing, check your controls, separate the ownership and sort of the authority who is responsible for different data. If you do that well I think you are fairly in good state. Also there are lots of products now. I mean of course on the cloud environment, so basically two environments right, there is a cloud environment and then there is the edge devices. Both you have to sort of look at it separately. On the cloud environment there are enough tools now to do stuff well. And you do your audits regularly, you do the certifications regularly. I think that pretty much takes care of most of the issues. Even after that you will have some or other sort of concerns, address them quickly. It's the speed at which you address the issues is essentially sort of important. So we've done enough, I mean we've done alerting and monitoring. There are specific people who are responsible for looking at alerts. Make sure that there's no noise in the alerts so that only the high-fidelity alerts are getting addressed and coming to write set of people. I think a lot of it is just basics and common sense but I think that takes care of most of the challenges and issues that come in. I'm actually helping all of you in regulating it and helping with cyber fraud detection. To an extent, yes, I mean running the business for sure. Our entire credit underwriting for example is model driven. And I think when it comes to cyber security for example, that is not our core specialization. But the products that we use and the technologies that we deploy for cyber defense, they do involve AI and machine learning. Not probably generated yet but yes, I mean there is model driven anomaly detection for example. There is identification of patterns and there is this point of view from a very perspective that we deploy and use. Yeah, I think as in the code expert is obviously not security but we do have a proper security team there. But it's more of the leveraging on the product that we use which in turn uses basically machine learning and probably these days the modern name for it which is AI and like whatever it is there. So it helps in basically logs and log analysis and pattern recognition across the logs to identify if there is any deviation in the way in which the systems are being accessed kind of thing. So this is again where we call it as a SIEM which is slightly bit more advanced with the traditional logging systems where you log it and then post an event has happened. You come and see, try to see the log and then try to decipher what is happening. But in this case basically the what do you call it the partner security system would be revealed the logs on a millisecond basis or close to a few seconds where the logs come to them. So basically looking at the access patterns and basically highlighting them. Obviously the systems are also basically you know each customer will have different logs and the pattern of them will also be different. So you need to be ready for a bit of a basically noise there in terms of certain things being highlighted. That's why the proper SOP team having which is monitoring at 24 bar 7 is important. Again it depends on the scale at which that is there. As in we have dedicated SOP team as you can also leverage the partner SOP team etc. So multiple strategies there but idea is that you have something running that at least so that the systems are alerting and then you take action rather than you post event action. As in this is more like a preventive rather than a reactive action basically. So alerting tools and basically these tools which use this machine learning and modeling and AI to understand the logs and basically learn and tell you that some changes happening system is always good to have. Yeah I mean we use some tools in SIEM essentially at the scale of the data that you are looking at without machine learning and pattern detection. It's just almost impossible to sort of identify any threat. So the SIEM tools that we use definitely do that. We also from a sort of certification perspective sort of managing the evidences and we have to sort of go through quite a few audits by different authorities throughout the year. And the questionnaire may be different but the evidences are fairly similar. So we use a product called Sprinto to collect all the evidences, keep it at one place and kind of make it available to different auditors in the format that they expect. So there is some sort of structure to the madness which you need to sort of this actually generates and this is a throughout the year activity. So between the tools which are available on the cloud, public cloud that we use, the tool like this and SIEM. All of these places there is a little bit of machine learning, it's just there, just very part of it. Any new tech tools which all of you must maybe are experimenting with right now which you think are further going to bring in more changes, revolutionize the industry further? I think I mean not in cybersecurity but of course in other places as the large language modules are becoming more and more prevalent. I think there are a bunch of use cases that we see. If we look at our business, it still involves a lot of human interaction from the time the customer shows interest into the product that we are offering to all the way the policy is issued, especially in the health and life category. The product collaterals are actually very technical which the end consumer actually do not understand the technicalities of the insurance. But sort of those are required from a regulatory perspective and just the way those products are structured. We are experimenting with actually, I mean some of them are in a fairly advanced stages where the newer technologies on the large language module and deployment of those either to sort of automate some of the human interactions. And then the summarization use cases of course where there are large techs sort of converting that into a simpler English and making it available to the customer. So I think there are definitely, we will see a lot of use cases which are getting built and will be built both on the servicing support side and on the sales side also. Of course, develop productivity is another area where there are lots of tools which are getting used and experimented. I think yeah, there is quite a bit happening, right? I mean qualities for example, one big area for us, right? And the conversations that someone is having say with a sales agent and a customer or someone who is negotiating an industry or amount or someone who is having a conversation with a credit manager when it is not straight through process or someone is having a collections call, right? There are a lot of aspects of it which are, you know, wide range and it is pretty much impossible to understand how good the conversation was. The customer promise to pay, at what point in time they promise to pay, right? I mean are we losing business because of certain behavior issues or the way we are pitching the product. All of this, right? Automated speech recognition, identifying of moments, understanding the sentiment of the customer. A lot of this, you know, we have started experimenting and deploying some machine learning and AI there as well. Post dispersal customer servicing for example, right? There is some bit of it there. We also have to be very, very careful, right? I mean, you know, there is coverage as in, you know, what population of your conversations or what population of your customers can be covered through that. And what are the use cases that are covered? And, you know, is there any danger to it? You know, you cannot accidentally respond, right? I mean, for example, hallucination, right, with some of the new models, right? I mean, that is not done. And regulator also says that if you do something, you'll have to explain why we are able to do that. So just saying that I've deployed this model and the model said this, I don't know why, but that is not allowed. It has to be transparent. So those are some of the guardrails that we have to take care of, but definitely a lot of use cases are working. Yeah, so I think obviously this view would basically broadly classify this into probably three buckets, right? One is basically underwriting and then maybe four buckets underwriting, fraud detection, operational efficiency improvement and developer efficiency improvement, right? These are the four broad areas where we primarily run our, basically, whatever the new word of AI basically that is there. So in terms of basically underwriting, I think we have been doing this for quite some time. As in pretty much all the fintechs do this because the sheer number of customers that we basically provide the services to requires the fact that at least you need to run it through these systems to come up with basically some decisioning aspects. But the idea is that obviously there is a manual oversee on top of basically what the system is deciding, right? And there is a monitoring aspect and then there is a comparative aspect also which happens as to in the sense that they would be running through a lot of different models, but you would be basically having different models, giving different outputs and then you are trying to understand why each one is behaving as it is and which one is behaving the better, right? For example, certain events basically can change the customer behavior and all those things. So getting these things to be understood also important, right? And the model might not be aware of such a thing, so you need to have a human oversight, especially in these aspects, right? The second in even here, what happens is that the model doesn't take the 100% decision, but it's more of a supportive action that happens with respect to what we do, right? So this is something that you can do and then we decide based on whether to go ahead or not based on certain parameters. The other one is with respect to what detection, right? So obviously in terms of, you know, the face comparisons and all those things, we need to do a lot of these things, especially because you cannot have an agent sitting manually, but I'm checking each and every customer's face against his ID faces and all those things. But there has to be again thresholds where you are able to basically identify what constitutes what and basically push them to a manual bucket and then get them done. So definitely a lot of work there and a lot of work on in terms of basically looking at the data and all those things. Customers are also getting innovative, you know, as in I think all of you would have written news about, there is this some town in Rajasthan where primarily a lot of people are involved in this, this is their prime business. So the idea is that with all these things you will have a certain set of people who will be looking at trying to get an advantage of basically the systems that we are having. So definitely these systems help in identifying such patterns and blocking them basically at the onset itself, right? That is the second thing. And the systems are getting better, right? As the technology improves, definitely the systems are getting better. So that helps. The third thing would be in terms of the operational efficiency. So as we said, customer support and obviously tele sales and all those things are places where you can enhance augment the productivity of the agents and that is how we are looking at it. As in we are not looking at it as a replacement tool but more of an augmentation tool where basically 1x can become 1.2x or 1.3x. So that basically on an overall we kind of increase the efficiency of the entire organization, right? The fourth one is in terms of developer productivity. I think in terms of basically the, at least in Bangalore and I was also talking in terms of the salaries and the cost that is paid to the developers. This is basically a very negligible cost to basically put it on to the, give it to the developers so that efficiency again can improve that. And then the idea is that an engineer who is coming into the system newly can he become productive in two weeks rather than one month or two months or something like that, right? That's what we are looking at and even the members were existing. How can they review their code faster? How can they look at search for bugs or understand a piece of code that has been written, summarization of the code, etc. These are all various different aspects that these technologies now do very well. And the idea is that again increase the efficiency from 1 to basically 1.2, 1.3 where you get massive scale of benefit across the entire base of people that are obviously there in your organization. This in turn will lead to better customer satisfaction and also being able to release products at a faster pace. So it is more in turn, more feedback loop which keeps coming in and then as you find these things, definitely it's for the betterment of basically customers and organizations. This is how I would look at it. So before I throw in the floor for audience questions, a quick comment from each one of you. What actually do you perceive as the future of fintech in the current times? I think if you look at insurance, the market size itself in India is so huge. There is a very small percentage of India today in short. So I think just the category is just very, very large. And I think in short text, just because of the size of the pie, it's going to keep growing and growing. That itself is a good sort of growth factor for the category. More than that, I think also more tech as we see deployment of more devices, more imagery. The whole ML side of things of processing some of this data and ability to make some certain decisions, whether on the sales side or on the claims side, all of that. I think in short text, I think has a very, very good future in India. Both on the production side of manufacturing newer kinds of products, more personalized products and also on the distribution side of reaching out to larger customer base in India. Like for an example in India today, the postman who, there are 350,000 postmen across India. The India Post Payment Bank, which was created as a sort of semi-private entity, they're leveraging postmen to do insurance distribution across India. And we are seeing, so Turtlefin, which is our tech platform, powers some of that. And we are seeing small ticket insurance products getting distributed in the remotest parts of India and massive scale of those kinds of products. See if you have to be a multi-trillion dollar economy, technology is the only way to do that. And I think there is no alternative. It's an imperative. And financial inclusion is a big thing for the country to grow. And the only way to grow about it is technology. I'll give one personal example right from Lightning Card is that we deal with small business loans and we cover about 5,000 towns and cities and locations. Now in that 5,000, there are small businesses in 4,000 and a one-way trip for them to a bank or an MDFC is 20 kilometers. The rest of the 1,000 are within the 20 kilometers. So even if someone has to go submit an application for a loan for running their small business, it might be a cloth shop, it might be a big making outfit, it might be a jaggedy thing. They have to take at least a 20 kilometer. They have to close their business and they have to travel 1 by 20 kilometers to be able to do that. And it's a multi-stage process. I mean there is a real impact of technology there. And I think, and this is one very small example, you barely scratch the surface. And if you have to become a multi-trillion dollar economy, this is the only way to grow about it. I would sum it as basically democratization of the services that are available to the customers. So as we just said, the good example is this 20 kilometer trip. And there is no way basically in the traditional space of basically how some of the systems run that you would get any service done in the first trip itself. So this is where the democratization happens, be it in terms of fintech services. Just to give you, if you take two steps back, the actual democratization started with respect to try when 3G and 4G basically kind of, the bars were set to such a good level that basically all of India basically switched over to that. That's basically the basic foundation. On top of that, I think another good example is UPI where payment space, five years back we were nowhere, probably a small spec, but now we are pretty much competing with China or basically in terms of number of digital transactions that happens on a daily basis, right? And that too on an open platform, right? That's where the democracy aspect comes in. And further to that, basically the next level of basically the services that comes is financial services like what we provide, insurance services like what is provided, right? Wealth tech services, market services, all of that becomes democratized, right? It's very well seen in the number of D-MAT accounts that are being opened and in terms of basically the new basically insurance, basically the number of people who are getting to the insurance space, the customers basically getting insurance, right? So all services are getting democratized. In terms of wealth generation, as in D-MAT accounts, then market products like probably NCDs and all those things which there are some platforms which offer, which is available to all of them basically on the mobile platform and none of those would have been possible without technology. And given that we have such a solid base of basically digital, this thing of basically wireless UPI payment space and that basically fintech space on top of that, I think digitally India is going to be kind of like a flag banner across the whole world as we have seen as we are exporting already UPI to a lot of other countries. I think soon you would just not be basically constrained within India and then move on across the rest of the world. Just to give an idea back in 1600s, the GDP of India compared to the entire world was almost 30%. I would say that we would be reaching similar levels at least. That's my wish to basically reach at that level where we are at 30% of the world's GDP. Thank you. On that note, the floor is open for questions. If anyone from the audience have any questions for any of our panelists, please raise your hand. The mic will be handed over to you. Sunshack. Ladies and gentlemen, we are opening the forum for question and answer session. We will be taking about three questions for this session. Which are the audience? Thank you. Thank you for such an insightful session. Question on, we spoke a lot about AI and ML models. And I think there was a mention about threat model. Okay. So there was a mention about threat model. Are there still any use cases which we feel are still not very future proof in terms of, let's say, secure? Which can still pose a rest to the fintech industry or fintech as a whole. I think there are still a lot of issues on the underlying secure supply chain also which has to be addressed. So some light on what kind of threats do we see on the learning models? But we are still open. That can be very helpful. Okay. So in the digital world, what happens as I said, when this democratization aspect happens, you know, it's for everybody, right? It's for the guys with the real intent and with the guys with the non-real intent, right? And just to give a good idea, I think voice cloning is something that is there. And even these days, even video cloning is also there, right? So these are all, it's become more of a cat and mouse game, right? So the idea is that anything digital can be altered. But the point here is that when digitally something gets altered, you can also have tools to basically identify the change in the patterns, right? I said, for example, if you do a JPEG kind of, if you edit a JPEG and then you do it, that's what probably step one of this thing, where somebody tries to modify a pan-card image with a fake image and then try to get it through, right? How do you identify that? That was step one, right? So these days, people have been, so they are upping their game. The idea is that you also need to be aware of the fact that digitally whatever is happening, it can be modified and then how do you identify the modification or what are the boundaries around that, right? So models definitely have to be built for that and looked at it. There are some models which runs for all of these things, right? So it's a cat and mouse game is what I consider. I think your question is, are there still problems that are unsolved? There are a lot of problems that are unsolved, right? And there are a lot of threats out there. If you look at any advancement, you know, how quickly can you understand the problems with that advancement and hence respond to that is very important. Think about, say, using machine learning and artificial intelligence in credit and writing, right? Typically, you know, you give someone a loan, they pay after 30 days, right? The first payment and then, you know, they have to pay for about three years, maybe, right? So if your model has said that, okay, this person is credit worthy, then it's only after a few months that you realize they are credit worthy or not. And hence the rate at which you change those models need to keep pace, right? Accordingly, and you need to have that kind of data to be able to backtest those models as well. That is very different from, say, a consumer segment where you are testing the price change in a mobile phone sale, for example, right? Where in, you know, you understand customer behavior instantaneously. I think there is a lot going on. So there is quite a bit of work that needs to be done. But the fundamentals, right, of whether you are doing the business on the basis of models or on the basis of, you know, traditional, you know, maybe intelligence, right, is the same. Hi. This would be the last question. Yes, for now we can take the rest questions off the stage. Thank you for that last one. So we are currently building a market base for whole-base savings. And we are trying to change the idea of savings outcomes. Like when people are depositing their amounts for various RDS or FT's or mutual funds or SIT's, obviously they are saving for some particular both, so we are working on that. Obviously B2C category and we both cater to the same. And you mentioned penetration as a major sort of opportunity because we have not gone hardly, especially in tier three. So how are you people working on that? How are you exploring in terms of insurance as well and credit loans as well into that category? Because agents is one way to work on that and then telecollers is another way. But to bring out that trust in that particular way, the belief is towards the government provided products in terms of FT's or loan. And even though, like, you might have seen, they generally do not get credits because of Sibyl score and everything, and they are going for a very higher interest rate. So that trust with, like, the crunch of the question is penetration part. How are we doing it and how we can do it, how we can actually penetrate towards that segment? Yeah, I think first and foremost, any product that you are building, you need to identify that unique value proposition. Where you are very different from someone else. For lending card, it was digitally delivering credit to a new to credit person, someone who does not have Sibyl score on the basis of cash flow in their businesses bank account. So we started there and then it sort of expands from there. And it's a lot of hard work. Clearly I came in the middle of that journey at lending card, but it's a lot of hard work. And from there on, there are several distribution channels. What you are talking about as penetration is really distribution. You would want to be able to distribute your product. Now, then you have to look for a channel for distribution. In his case, it was, you know, one of those examples was India Post. In lending cards case, and in a lot of other tech companies' cases. It is search engine optimization and search engine based marketing. There is agent-based networks that you will have to tap into. And then you have to grow from there. How do I put it? It's a lot of hard work, but there are no shortcuts to it. And you have to do it day on day and inch by inch. With that, we call it a day. Thank you so much.