Redefining Banking Operations Through AI

Speaker A: Foreign. So good evening friends and good to see you back. J.

Speaker B: Thanks, Leslie.

Speaker A: Hope you had a lovely weekend.

Speaker B: Oh yes, it was good. I was in Bali.

Speaker C: Ah.

Speaker A: Uh, you must someday have a whole episode on what you do in Bali and all these other places. But uh, for now I think we have a special guest today. Uh, he's uh, Ratan Kumar Keshe. Uh, he's the executive director and chief operating officer at Bund Than Bank. Uh, now he has obviously about, you know, more than three, uh, decades of experience across industries, uh, you know, leadership roles, multiple domains, whether it's operations, technology, transaction banking, affluent banking, you just mention it and then probably he's there. Um, so, um, in this specific role as Ed and CEO, he's been with Bandhan bank for more than two years. I guess it was March, uh, 2023. Of course he was part of the core team leading digital transformation during his tenures at ICICI Bank. HDFC Bank. Yes, Bank Access Bank. He's always been a banker, so pretty interesting stuff. Uh, and you know, he was among the first few in the Indian banking industry to launch you know, a service quality frameworks and service CRM when he was with HDFC Bank. Uh, interesting think he's also, he's a bachelor in Engineering, Mechanical Engineering, very interestingly from NIT and an MBA from Narasimhanji. Uh, and of course he was also certified quality engineer from, you know, the Quality Council of Indiana. Ah, etc. So plenty of uh, and viewers and listeners, as you must have guessed by now, we are going to talk, you know, he's going to talk more from a banker's perspective on the evolving role of, on AI and other kind of emerging tech where I'm also going to ask him about quantum computing for sure whether he, whether Ratan has his opinions on the PQC part of it, PQC readiness or the post Quantum readiness. Uh, as such we will be asking him about that. Um, Ratan, before we, you know, get into the meat of what you are doing at Bandhan Bank, Jayant and I usually have this where we, you know, go over the week's developments. And this also like every week, uh, you know, practically this has also been a very eventful week in the field of AI. Uh, just a few things. Like for instance I was uh, uh, you know, as uh, part of my weekly newsletter on AI for Mint. I typically cover a lot of topics and some of these things that caught my attention was of course that India has stopped, uh, you know, Coursera's global JI learning ranking. Okay, so it is among the top in that but it ranks a very poor 89among 109 countries in skills proficiency. Now this is pretty interesting from the point of view that you know if you look at it uh, so as far as the Genai adoption is concerned Khosrah looks at whatever stuff or how many learners are there. So about 1.3 million enrollments have taken place in 2024. But yet the rank uh out of 109 countries. India ranks at 89 uh in the latest Courseras Global Skills Report. But interestingly it is ranked 46th uh on its uh, newly launched AI maturity index. Um your usual suspects are the top three, four, five, Switzerland, Netherlands, Sweden, uh UK was ranked 22nd, US was 27th and China 39th. Pretty interesting stuff right? And um, also if you know this is a little, why I found it a little strange is that um, uh if you look at the QS rankings which were put out in um, uh January, uh the QS Future Skills Index they tell a different story because India was ranked 25th globally as the future skills contender, uh, it's scored exceptionally in the future of work metrics ranking second only to the U.S. now you know you clearly see a discrepancy in these two reports. But you know the point is that the methodology also is quite different because CORA tracks the actual skill performance on its platform. Uh QS rankings evaluates how well educational systems prepare students for future jobs. So enough in a way they are pretty complementary and of course uh, uh that's how the skills part is concerned. J Ratan, any quick thoughts on where India is uh, as far as ranking? Forget the ranking part of it. Whether it's 89th or whether it's 25th or whether it's 30th really I don't think those things matter at this point in time. But you know you both uh, are uh, seeing you know a lot of skill sets and upskilling etc going on. So some quick thoughts on it.

Speaker C: See this is look all of us are using AI or getting benefited out of AI in many form in our day to day life. Whether it is uh, you know taking that ride, hailing Rapid Rapido or a Uber or an Ola or buying products from the E commerce or availing banking services that we are getting benefit but the development and the skill is still sort of concentrated with a set of folks set up companies who are developing some of these. Therefore to make it more broad based and really create uh, more revolution in terms of the larger Indian population that is going to take some bit of time, but it's only a matter of time. I would say that as you rightly said, the future readiness is there. And I think, uh, like it happened in many of the other scenarios like digital payments, etc, it will at some point explode. So we are just waiting for that to happen at some point. Ah, as it becomes more rewarding, more people become more aware. I think that has to come very soon. It will happen.

Speaker B: The way I look at it, Leslie, is Indians, uh, you know, Indian professionals, uh, are very attached to certifications. Not just now. If you look at it 20, 25 years ago, CCNA, CCNP or Microsoft certification. Okay. Networking certification. People used to do it as soon as they get into their jobs as a form of putting it on, on their resume to get another job or to get a promotion. Okay. So getting these certifications, I mean, how many times have you and I have been asked which course should I do to, you know, to get better at AI? And my question has been go to a startup, get your hands dirty and do something. Build a project, intern yourself. That's the only way you can learn good AI. But everyone wants to do a course to put it on their resume. I think that's what, uh, you know, the 1.3 million registrations on Coursera for Coursera's courses. Okay. India ranks because everyone wants to put, uh, you know, certified in XYZ AI course on their resume. Okay, that explains it. But in terms of skills and proficiency, two things. Uh, uh, you mentioned maturity index also. I guess maturity index, uh, is a metric of the number of companies in the country that are using AI. There we are 46. It's not an apples to apples comparison. We are a big country. There are many, many industries, many sectors, many companies in, uh, this one and the AI adoption in different sectors. And this one is varied. I mean, you can't uh, compare us with Switzerland, which is a much smaller country with uh, you know, number, uh, of companies. This one. So the concentration and the adoption, the rate of adoption will also be higher because of the smaller base. Right. So if you look at maturity index, we are 46th.

Speaker A: Okay.

Speaker B: But the skills proficiency is also that, you know, the Indian companies right now are not readily adopting AI.

Speaker C: Ah.

Speaker B: How much have we talked on this podcast about enterprise AI adoption? Especially Indian, uh, you know, companies, uh, adoption, uh, this one, Leslie. So I think those two are a little, um, you know, indicators of the 89th in skills proficiency. Also really, really, really good AI engineers are not staying in India. They are, you know, going out of India, working for, working for global companies outside of India. Okay. And, and uh, you know, a lot of people who are, uh, you know, taking courses are also going out of India. I think a combination of these reasons explains some of these rankings. But I completely agree with you when it comes to the fundamental methodology of the two different reports itself that you mentioned. Right.

Speaker A: I guess the current happenings in the US might change this perception of many people going to the places like the

Speaker B: US over the weekend I met someone from Boston, ex Harvard person, growth hacker, working for fintech and uh, uh, ed tech companies in US who wants to work, uh, moving to China or Southeast Asia or India.

Speaker C: Okay.

Speaker B: And it's actually discouraging me to, you know, uh, you know, go to India and build business, sorry, US and build business there. I'm like, why would anyone want to come to us right now?

Speaker A: Absolutely. They make you feel like, you know, a second grade citizen. And I don't think that's uh, a very healthy thing to go out in today's world. But you know, just coming back to the course of the report, you know, India leads in learner volume with 28.4 million users. So this actually surpasses all of Europe for that matter of fact.

Speaker C: So pretty interesting stuff because of the number of people that you have as a country and number of educated engineers that you have.

Speaker A: Absolutely, absolutely, I fully agree with that. And even the skill sets are the usual suspects. Whether it is, uh, AI, whether it is ML, whether it is data analytics, etc, so pretty interesting, of course, from a gender gap challenge that is always there. It's a global gender gap challenge also. And in India, uh, because in India I think women represent only 30% of gen learners, uh, compared to 40% across all Coursera courses. Now remember that this study or this thing is restricted to the Coursera platform as such, Um, I think, but uh, just for stats purposes, uh, enrollment for Jennai, just to show, you know, our viewers and uh, listeners the interest in Genai is that enrollment has surged from 1% per minute in 2023 to 8 per minute in 2024. Uh, and in 2025, nearly 700 Jenny courses average 12 enrollments per minute.

Speaker B: 1 to 8 to 12.

Speaker C: Yeah, yeah, but pretty interesting.

Speaker A: I mean, so 12, uh, enrollments per minute. That means even as we are talking, so many enrollments are happening.

Speaker C: For Jenny, the other day I was just. I mean sometimes some of those, uh, you open YouTube and some advertisement come and the folks are saying if you don't have AI Gen Gen AI in your resume, M people will not even open Your resume. So as, as Jan said, there's a tendency to say, okay, let's just get enrolled and learn something whether you use it or not. We'll figure it out later. That's also happening. It's a bit of a. Bit of a.

Speaker A: And you must ask us, you know, because in media persons every press release has AI and geniusly being pushed over there for no reason. I asked them, I said, but what is the business problem you plan to solve? Yeah, nobody explains that. Now there's no way that AI, we are using AI and gen AI. But to solve what? Because what could have just been done with plain statistical tools and probably traditional machine learning tools. Why are you using a hammer when you could have just used any other instrument as the light instrument?

Speaker C: There is a joke in Silicon Valley that any company who are going for their quarterly results announcement in the analyst call, they have to get three things right. One, they have to get the JNAI and AI agentic AI right. They must talk about it. Second, they have to get India strategy right. And third, uh, they have to have esg.

Speaker A: Right.

Speaker C: And if you have none of these three,

Speaker B: still is a metric in uh, Valley.

Speaker A: Yeah, that's surprising.

Speaker B: I mean that's the post Trump era.

Speaker A: Yeah, the post Trump era, absolutely. But of course viewers and listeners, before we get back to how AI is being used in the banking sector, uh, uh, to start with, Bandhan bank, just a few More details that OpenAI's annual revenue has touched $10 billion. Of course there' you know, this happens even as a court order has mandated that uh, chat GPT free users, their chats will be retained for the period earlier they had to delete the chats within 30 days. But now because of that NYT, um, New York Times has filed a case against uh, uh, OpenAI and Copilot, Microsoft's copilot for. I mean this uh, actually this uh, case pertains to 2023, but in uh, May last month actually they just reopened the case and the uh, judges passed an order that. Okay, fine, you can, you know, just to uh, check whether there has been plagiarism, especially verbatim kind of stuff, the user chats will be deleted. So free chat GPT users, be careful. All your sensitive information is being recorded. At this point in time, OpenAI will not be able to, uh, delete it. Of course. Uh, other interesting news is that Meta also is getting a little, you know, jittery about the AI plans. So it is buying scale AI. This is a uh, report and the information, I think an Exclusive that they are planning to buy scale AI for over $14 billion and of course they also plan to set up ah, a uh, AGI Artificial General Intelligence Unit. Um, other information of course is that Apple's WWDC it has failed to impress analysts in the AI space. Uh, and um, yeah I guess uh, that's uh, the news of the week.

Speaker B: Apple, the report that they released last week about how the reasoning models are nothing but uh, you know, pattern recognition that they're not actually good at logical

Speaker A: reasons at the talk of the town

Speaker B: actually you know that, that, that needs to be touched upon uh Leslie. I mean uh, uh so for our viewers and listeners, Apple released a uh report a few days back which claims that all the Reasoning models across OpenAI across Meta, uh, anthropic and all of them deep seq, uh, and even the Chinese models they claim that they're not actually reasoning models, they're just very good at uh, memorizing patterns and that uh, in a very fast and advanced way leads to the models looking like they're reasoning and they're doing logical reasoning. But Apple cited a few examples and the case studies that they've given actual uh, puzzles to solve and the models couldn't. Right.

Speaker A: Haven't you and I been speaking about this earlier also? We have been saying that these are next word prediction engines.

Speaker B: Exactly.

Speaker A: Again I myself have written so much about it but I'm glad that somebody is calling the bluff at some point because these are.

Speaker C: Once you type, give you a prompt which is slightly longer, you actually end up finding that the answer comes only from the last couple of lines because it could only relate to that. So it's not a reasoning model in the, in the largest sense of things for sure.

Speaker B: Yeah, yeah, yeah.

Speaker A: I mean see they are definitely improving but they cannot reason the way humans using your props and how um, it also depends on how effective your prompts are also at the end of the day. So I think there's a lot of learning that goes. I think this is, you know I think one section always gets carried away by. What's that term that we use that San Francisco

Speaker B: consensus?

Speaker A: Yes, the San Francisco consensus. Because this everything is, you know it, it hinges on the San Francisco what It's like the people dictate in Rome, you know for the Catholics, whatever it says that like the Dom has spoken the causes. So it's like San Francisco has spoken and everything is like you know, ankido in the field of. No, it doesn't work that way and I think people like Ratan Are now going to tell us how in actuality when you actually building AI and JNAI and so called AI agentic systems and marry them with legacy applications and regulatory hurdles. What exactly happens? Ratan, give us a quick overview for the viewers and listeners as to you know what uh, are you doing anyi a bit AI agentic system so that our viewers feel good about it and then you can always sort of crash the expectations if you wish.

Speaker C: No, it's simple. Look it is a fact that if you go to do digital transformation and most of the banks have moved out of the brick and mortar model to more modern digital transformation LED technology and we have no choice but to do that simply because while we are in this regulated industry, um, we are being compared with experience that the customers are getting in the E commerce or in the larger segment of the more sophisticated AI driven companies, the customers do compare with that. Therefore if a customer is coming to open an account for an onboarding experience, they expect that the experience has to be far more refined and simplified. Now how to make it simplified? You have no choice but to use AI and what does AI do is it's simple. You just give you one of your ID proof, you just click a picture, take a selfie, rest everything at least 60 plus validation are done real time and then you can set up an account instantly and give it to the customer and see your account is set up. Now this you couldn't have done in the earlier scenario because you had to go through each of these checks and balances with a human being doing it. Today the systems have ability to do all of that. That's number one the underwriting decision that we take today. There are so many models that have been using but the model scan actually you can only apply for a particular model and you have to keep tracking against each model and scrub it through. Whereas today AI has an ability to churn millions of data and create models and also look at alternate data sources to do credit uncredit decisioning. AI can use fraud detection. For example if Leslie is someone who withdraws 20,000 rupees from Bandra every week and the moment Leslie tries to withdraw 40,000 rupees from uh, Bangalore city, AI understands that there is out of pattern transaction and raises an alert and say looks like it is a fraudulent transaction. It doesn't stop at that. If it finds that Leslie tried 40,000 in Bangalore in one location and it goes on uh uh to another place, another few kilometers away from another atm it has an ability to say it is definitely fraud and will Stop the transaction. Also AI is actually making life lot more simpler in the larger sense of thing. And I think we've been talking about whether it is a reasoning model or whether it is.

Speaker A: I'm uh, not interrupting you but I just wanted to you know you to expand on certain things because the way I understand it all this has been happening for almost a decade now because this is, this is basic traditional machine learning seeking patterns and putting out uh, the stuff uh yeah, you could just tell us how you do the newer refinements. Whether it is Genai, whether it's agent system, how they are enhancing all these offerings. So what probably do earlier as far as the. So maybe the red flags are being found out in real time uh or what. Exactly. And the kind of you know that would be pretty interesting for us.

Speaker C: It's like this, you need still need an OCR capability, optical character recognition capability. The accuracy five years back used to be 72 to 80%. Now it has reached 99% accuracy even on a handwritten. Now if you have a higher degree of accuracy on an OCR and that then picks up that this field appears to be pan or it appears to be other then uh. The ability of the AI to churn it and give you the better outcome goes up automatically. That's one thing that has changed. Second the good news about AI is that without doing anything just because of usage of that model, the model keeps getting better. Now that's the beauty about AI it. It learns from every single mistake. It learns from every single uh correct uh thing job uh that it does. So it learns from the successes, it learns from the failure and therefore it keeps getting enriched time and again. That's the second piece that the third is it has an ability to give you decision real time if you allow it to do that or it can say okay I have decided this much and I will flag it up and say I believe it's 95% correct. Either you go ahead with this or you can not have any human being touching it and allow it to go ahead automatically or allow a human being to touch it. That's the third thing that is happening. The fourth I would say given the fact that it is more open banking ecosystem today. It used to be very very clean cut which is your on prem systems and therefore the learning from the outside surrounding systems or from the other partners not used to get loading it was only on the release the next version release which will be coming after some time you to get the benefit of it. Now with open banking and the cloud capability you are exposing to a lot of other partners to come and develop and a lot of other partners. You are opening the APIs and therefore you are getting the benefit of the multiple ecosystem levers to get the benefit of the entire full scale banking. And what it does, therefore it makes banking bit more unobtrusive in a sense. I don't go to Jayant and say Jayant, take my credit card or why don't you take uh, have this offer. I actually know that Jayant is now in Phoenix mall in Mumbai. It is 2pm and he's probably looking uh, to go for a lunch. And I also know that Jayanth actually prefers Chinese. And I can make an offer to Jayant saying that look, you are looking for an offer in Chinese in my credit card in so and so restaurant, you got a 20% offer. Now this open banking ecosystem of multiple players coming and developing, uh, cool ecosystem players are actually helping us to get better, to offer the contextual solution. So these to me are few of the changes that had happened compared to what it used to happen few years back.

Speaker A: And you're also doing some work with Salesforce, right? That's on AI agentic system. So et cetera, lending.

Speaker C: Yeah, yeah, Salesforce. We are building our loan origination system. And of course along with that it comes with multiple other levers and uh, Agent TKI is one of them, Agent Force as they call it. Agent Force is one of them that we are developing at this stage. But uh, you know, honestly I would say that we are not trying to go straight to the customers. I am trying to develop for my employees. I've got 80,000 people, most of them. These 80,000 people are asking lots of questions to the head office on behalf of customers. So I am not still fully into it. I don't, still don't trust it fully. So what I'm trying to do is that I'm trying to open it up for my employees and say, can I answer your question correctly? Because I still have some degree of risk taking capability with my employees and once I get that better and it reaches a particular level of accuracy, then I'll open it up for my customers. Only thing is that the responses will be slightly different in customer scenario. Visa is what I do for my employees.

Speaker A: That's what you are doing, eating your own dog food. Yeah.

Speaker B: No, he just explained it. A typical enterprise AI development M and rollout gtm, uh process. He's explained it without us asking the question, which is our normal question to every guest almost on every episode. Ratan uh, there's one interesting thing that you said, the first thing that you said is the experience of the customer. Uh, you guys are competing with the E commerce sites, the cab hailing sites, the ride sharing sites or, or even the zeptos of the world. That's very interesting uh, because you're saying that essentially the consumer is the same or the customer is the same and he gets AI enabled this one from consumer Internet, uh, applications that he uses across uh, his life on a daily basis. And when it comes to the bank, uh, this one is sort of uh, expected. And that's why the banks, or at least your bank is trying to catch up and deliver on the same experience. That's very interesting.

Speaker C: So tell me.

Speaker B: And when it comes to banks you're uh, a very regulated uh, entity, uh, uh, and an industry as such. And still as much as Digital transformation, digital KYCs and everything are taken into consideration still because of regulation there's a lot of paperwork which is involved which might not make the AI driven or digital uh, technologies driven experience that much seamless and uh, you know, um, that much seamless or smooth. Okay, so for the same consumer who's used to getting extremely seamless, you know, everything in 10 minutes when he, when he orders, sitting at home to the banking experience because of regulations, because of the number of you know, checks and balances that you guys have to do, how are you overcoming that? And you know, uh, how is AI being used specifically to smoothen some of those processes?

Speaker C: Yeah, so there are two parts to this. One is we call it in banking as you know, we call it three types of customers. One is ETB which is existing to your bank. There is NTB new to the bank and then KTB known to the bank which is they are not my customer but they are using my payment systems etc, etc right now, the treatment for these three will be very very different because if someone happens to be my ETB customer, I know the customer so well that I don't have to really collect fresh set of documentation anymore because whatever documentation I have, these are all stacked in the rightful manner saying that I have collected passport, passport is valid in August 2028. I have all of those details available and from the field the sales guy can actually fetch the detail and say here is this passport which is available with me or there is something more, that additional document that I need for this particular purpose. So the requirement of the number of bunch of documents has come down significantly because you only collect only the one that is needed for your ETB guidance customer, that's number one.

Speaker B: Fair enough.

Speaker C: Number two, because the customers are now using UPI. And UPI, as you know, almost 50% of the global transactions are happening in India and it's almost alive, uh, uh, for all of us.

Speaker B: It's a living organism.

Speaker C: It's a living organism, yeah. Therefore the number of transactions that are flowing through your system, you know so much about the customer, what the customers are doing with the money every day, how much is coming in, how much is going out, for what purpose. You get it, get to know that very, very clearly. That's the second element of it that gives you. So you have got uh, identity taken care of and you have got transaction pattern. And the profiling is a lot more enriched than what the customer would have told me four years, five years back. So it's a lot more enriched because the financial uh, profiling, I know it a lot better. Therefore I have an ability to stamp a customer for any product. And if I am able to stamp the customer for a product, if the customer is okay with that. So PL for a 10 lakh or a car loan for 14 lakh if it is meeting that requirement, if it is within that, I can do it like instantly because then digital loan origination system allows me to do instantly. Not only that from the field I can connect to uh, an expert agent who will come and explain the nuance of the product. And that agent could be a real agent, it could be actually a bot agent. And the bot agent talks far more seamlessly in their own vernacular language than it used to be done before. So this ability actually makes it a lot simpler. But what I'm trying to do, and we're actually revamping the entire uh, digital platform, the Internet banking, mobile banking. I'm trying to create some bit of a card that you apply, you want a debit card, you put it in your card and then you check out uh, whatever, paying 250 rupees etc. So many banks are actually trying to get to that similar e commerce like

Speaker B: ecosystem Marketplace, uh, Marketplace UX user experience

Speaker C: so that you actually do that. And then we also have an ability to say if you buy this or if you're availing this, you may need this. So m. If you are buying, uh, taking a pal ah, and you are making a credit card payment for your overseas ticket, you may need a health insurance, travel insurance or you may need forex. I have two more products which are linked to that. So these are some of the stuff which AI can give you, uh, an ability to do A lot better, you know.

Speaker A: But tell me something. You know, typically at the back of the mind when you know so much about the customer, there's always at the back of the mind how much do you actually know and how is, is there a very good chance of you misusing it? Also when I'm saying you, I'm not talking abundant bank, I'm not talking of any specific bank as such. I'm talking as a sector. It's a problem of, you know, this cross trading of uh, information that takes place. Now some part of it could be anonymized data even then, uh, or some part could be something else, but even then it's a little tricky.

Speaker C: So two or three parts to this most of the times a we are highly regulated industry and therefore how the data can be stored, data integrity, how the data can be transmitted, it has to be always mass. It should be encrypted both in the static and in dynamic environment. All those are taken care of as part of the multiple rounds of audit that happens. So that's one second. I think it's fairly a thin line between, let's say giving convenience to the customers and data sometimes getting misused. And there is a very thin line, same customer who may love it, saying that look, it is so great that you came back to me offering me something which I was just looking for. Visibly somebody is saying why are you having this information with you? So you've got to be extremely careful of it. But what we do is that most of the time we churn the data, uh, and nobody is physically touching it. The models are getting enriched time and again using AI and then you offer it to the customer and tell the customer to do it themselves rather than making too many outbound call and say, why don't you take this product? Why don't you take this product? That extent is coming down of too many outbound calls and hounding the customer almost to the level of harassment. You are actually doing it for the segment of customer wherein you want them to use the data or they use the offer themselves. So it's bit more self driven and if the customer like it, then an individual comes up and say you tried doing it but you dropped up. Is there anything that you need that I can help you with? And then an agent comes back. So that's the second element of it. But for the lower end of the segment, wherein let's say we also have a microfinance or you also have the common uh, sort of uh, retail customers there we don't use so Much of it. I mean because we believe that there you've got to be a bit more careful, uh, with heightened sort of chances of fraud happening around, we are slightly more careful. So we are using the data in the specific buckets, specific segment of customer for specific product. But the data integrity is taken care of end to end and that is always there.

Speaker B: Yeah, um, I'd like for you to explain, uh, elucidate for our audience. You gave an example earlier of how um, a person, a customer, uh, in a Phoenix mall, um, at lunchtime and you know that uh, he likes Chinese so you can give specific this one and you attributed that to open banking. Okay, we understand what open banking is, but for the uh, you know, for the benefit of our listeners and audience, can you explain what open banking is and what it has enabled and why that has enabled the right uh, you know, extremely high profiling and the right way of servicing uh, the customers. Not just by the banks, by other entities as well.

Speaker C: So let me give some, some other example which is bit more common, sensical in the rural market where we are very, very strong at. If you go to a rural area and say can you take a personal loan or can you open a savings account? That's one way of taking your banking open a branch and then start doing that. There's another way of doing it. What is that farmer looking at? The farmer is looking for the right crop that one could go for number one. Farmer is looking for the right quality of seeds. The farmer may be looking for a bridge loan to be able to go through this whole process of agriculture. The farmer may be looking for a tractor loan, uh, at the end of harvesting, farmer may be looking for the right price for the product. And if there is surplus money, the farmer may be looking for getting a right sort of product wherein one can generate bit more money for the next year. Now rather than trying to do traditional banking, can you be part of the and weather forecasting? So can I be part of this ecosystem and say I will provide you all of these in an ecosystem which is open banking and I will bring banking solution as and when needed. So I'll give you the right bridge loan, I'll give you the tractor loan, I'll also help you with any other product that you may need and I'll help you open a savings account, but I will not disturb your day to day life. I will bring a platform layer wherein you get all of these other services coming to you automatically. And I just remain somewhere beneath this whole layer of your ecosystem. So the farmer is liking it because they want someone to be part of the ecosystem. Not intrude into the sense of why don't you take a uh, tractor Finance. That's what it is. Open banking. So there are Reuters of the kind, there are other fintech players of the kind, there are pharma producer organizations. They're all part of that ecosystem. They're all building that layer thanks to the India, Indian government ecosystem. Let me put it that way. I think these are all helping.

Speaker B: So open banking is essentially a system where not just banks but um, other players, business correspondents, merchants from different industries can get together and service the customer wherein bank comes and offers its uh, role of a loan provider, credit provider and different products and financial services provider as such. But a lot of um, transactional data and a lot of profile, um, intelligence is uh, anonymously and under a certain uh, you know, data regulation and masking being uh, shared between these players on the same platform too, uh, with the objective of servicing the end customer in a much more seamless and a holistic way.

Speaker C: Look, look, Nandan Nilakani said that we used to talk about jam, Trinity, Jandan, Aadhar and uh, uh, mobile. Now you've got upi, Uli and jam. That's the next level of Trinity. Now these are all facilitators of this whole open banking ecosystem.

Speaker A: Uh, you have the whole stack, basically the India stack, whether it's the account aggregator, the ONDC, etc. I mean a pretty interesting time. Before we conclude, uh, Ratan, one question which I wanted to ask you is about quantum computing. I mean people are saying that data could be harvested at this point in time. That fear is very legitimate in many circles. Now is it scaremongering? Is it fear? Uh, and if so, what are banks doing to become you know, sort of quantum ready? Or is it something that is a

Speaker C: little far off I would say for Indian banking industry I'm sure. I mean we are highly regulated. At the same time we have the most pragmative and proactive regulator in the form of Reserve bank of India. We believe that the regulators are already way ahead of uh, each, each of the individual banks in this space simply because they have got data about all segments of customer for the entire industry, accumulative data for all the banks put together. And I believe that regulators are already working on quantum computing of that sort. But individual banks are far away from that. But I'll tell you one specific use case and maybe something that can draw attention to you, to you and your viewers. See I Created something not here, I'm just creating at some point here, but in another institution I created something called a service Data Lake. Now what Service Data Lake does is that on a real time basis my customers are talking about my products and services. As we speak they're talking to in social media, they're calling up my call center, they're walking into the branches and saying something which is getting recorded in CRM. They are sending an email, they're talking to a relationship manager who is recording in their CRM as well. So the call chatter is happening across and then of course there are certain messages which are coming, all of that are happening. Can I bring all of this data on a real time basis to a data lake? I call it SDL or a service data lake. And then every time a customer is calling my call center, I refer to the data lake with an intelligence to say customer must be calling for something because the debit card has not worked in a particular atmosphere. M. Now if I know that then the bot is listening to the conversation happening between an agent and the customer and it knows who is customer, who is agent and if the conversation is going right to solve that particular problem, why the customer is calling me, it's great. If the conversation is not happening right, it goes and pops up to the agent to say that here is this knowledge portal, can you refer to and do a better response? If even that is not helping and the conversation is really terrible, the bot goes and tells this agent and say hang on, you are not able to handle it. Let me transfer to the senior uh, SME, uh, uh, uh, person which is subject matter expert and transfer this call who handles it better. And these are all happening real time. Now if you're going to do all of these with at a sophisticated level, pure play AI cannot handle that. So you need far more complex algorithm, far more strong hardware systems to be able to get to that. And then you look at these at a lot more scale environment and I think probably quantum computing could be helpful. But I would say I'm not highly competent on the subject. Um, based on my little understanding about that, I would believe that it is little far away for individual companies to be using it. But it is definitely something that can happen very pretty soon than what we can even think of.

Speaker A: Yeah, I was referring more to the uh, encryption being broken by quantum computers when they become fault tolerant at some point in time. But of course as you said it's a little distant at this point in time. Um, and Indian banks and banks, uh, the world over are Preparing for that. And of course we're also talking at this point in time about a uh, kind of emerging uh, or melding of AI and quantum computing technologies. But that's a little distant at this for now viewers and listeners. I think the good thing is that you know banks, I mean as uh, Ratan has clearly pointed out and emphasized time and again is that banks are using your data but they're using it for the good purpose. Now you need not be you know unnecessarily uh, you know there's been a lot of.

Speaker C: But let's be. Sorry I still tell your viewer one more time. Banks never ask for customers pin customers, account detail customers OTP and therefore all of the viewers and your friends, all your acquaintances please remember if you're getting a call saying that there is a KYC requirement etc, bank is asking for OTP, asking to come on a video call on WhatsApp that is absolutely fake. Please do not encourage that. Please, please, please do not do this. We are trying hard to protect your money.

Speaker A: She said it with folded hands. This is exactly what Amitabh Bachchan does on kbc. I think viewers and I think you got the message that you're information to the because this is a highly regulated sector. RBI uh takes a lot of care uh on the data and um, I mean yes there will be slip ups at some point in time because we're dealing with human uh made systems at the end of the day you uh, know not everything can be automated and the sun. But there's always a human in the loop and there are checks and balances and I, I think Ratan also pointed out basically that there's a slightly, the lines are blurring you know between how much benefit benefits you get and it's also your perception whether you're being you know you're getting a benefit or you see it as a nuisance. Now you have to take that decision for uh, you know one, one man's food is another man's poison as they say. But uh, I think uh, uh clearly Ratan has shown how AI gen AI AI agent existence etc are solving business problems in the banking sector. So thank you very much for your time Ratan. Ah, hope you have a lovely weekend and Jen wishing you also a lovely weekend. Don't make us envious every time by going to Bali. We'll go to

Speaker C: or rather take us, take us along whenever go next time.

Speaker B: Yeah. Almost every guest, almost every guest on this podcast says that every time I go to God. Yeah.

Speaker A: Thank you viewers and listeners. You all also. Have a lovely weekend.

Speaker C: Thank you. Thanks, Jayant. And let's see. Thank you.