L&T Finance Limited (LTF) Earnings Call Transcript & Summary

I welcome you all to the L&T Finance Investor Digital Day 2025. The L&T Finance Investor Digital Day coincides with the flagship Premier AI and Tech Conference, RAISE 2025 being held tomorrow. We have with us today our Managing Director and CEO, Mr. Sudipta Roy; and other senior management members. The agenda and flow of the L&T Finance Investor Digital Day is as follows. It starts with a presentation by our Managing Director and CEO, Mr. Sudipta Roy, delving on the tech vision and execution road map at play in L&T Finance. This is also followed by a presentation on the digital architecture road map implemented by our Chief Digital Officer, Mr. Ramesh Aithal. Following that, you will have a presentation on the deep dive of Cyclops and Nostradamus, our proprietary AI-based underwriting and portfolio management engine and other AI used cases in L&T Finance by our Chief AI and Data Officer, Dr. Debarag Banerjee. This will be followed by presentations on digital use cases, along with business updates by our Chief Executives, Ms. Sonia Krishnankutty, Chief Executive, Rural Business Finance; Mr. Asheesh Goel, Chief Executive, Farmer Finance; Mr. Jinesh Shah, Chief Executive, Urban Secured Assets and Third Party Products; Mr. Manish Gupta, Chief Executive, Urban Unsecured Assets, Payments and Partnerships, Mr. Abhishek Sharma, Chief Executive, SME Finance. Following that, the last set of presentations will be by our Chief Operating Officer, Mr. Raju Dodti, providing the go-forward strategy of our new product, Gold Finance. The wrap-up to this August event will be with a presentation by our Chief Marketing Officer, Ms. Kavita Jagtiani, shedding light on our branding and marketing strategy. The session will then culminate with a detailed Q&A anchored by Mr. Sudipta Roy, our MD and CEO; and Mr. Sachinn Joshi, our CFO. So before we proceed, as a standard disclaimer, no unpublished price-sensitive information will be shared during the presentation and ensuing discussions. Only publicly available documents will be referred to for discussions during this meet. While all efforts will be made to ensure that no unpublished price-sensitive information will be shared. However, in case of any inadvertent disclosure, the same would, in any case, form part of this meet. Further, some of the statements made on today's meet may be forward-looking in nature. A note to this effect is provided in the presentation uploaded on the exchanges just around half an hour back. Also, before we begin, we request you all to kindly put your mobile phones on silent mode. I would now like to invite Mr. Sudipta Roy, our MD and CEO, to please come on stage and provide us with a detailed perspective on his vision for L&T Finance. Thank you. Over to you, sir.

Good evening, everyone, and thank you all for taking time from your busy schedules to join us today evening. For those who are joining us for the second time, we did the Investor Digital Day for the first time last year. Welcome back. And those who are joining us for the first time, we hope that the next couple of hours will give you a deep dive into our tech initiatives and our growth and execution plan for the next couple of quarters. Now there is a quote loosely attributed to Lenin, where he said that there are decades in which nothing happens and there are weeks during which decades happen. If you look at the geopolitics, that has sort of been stressed on us this year. It's almost like decades have happened in weeks as well as if you look at the tech space, the speed at which things have been moving in the tech space have been quite blinding with new models being introduced with every passing week. And in fact, I don't know whether you caught the news, an Indian LLM, voice AI-based LLM broke into #4 rankings start-up out of Bangalore and it was published yesterday only. So Indian start-ups are also catching up. That's a little about me. But if you were to ask ourselves, what did we achieve since the last time we met? We met in November last year on the backdrop of RAISE'24. Tomorrow, we have RAISE'25, the second edition of our flagship AI Summit. We have done a lot of things, but I'll focus on 4 most important sort of achievements. I think we have accelerated our core business and added new growth engines. Last year to this very time last year, the microfinance industry was going through an asset quality issue, the entire industry. But I'm happy to report that, that is a passing phase, and the industry has more or less resumed this journey on coming back to normalcy. And you will see that our rural business finance disbursements in the month of October at about INR 2,160 crores, a 42% jump over what we did in October '24. Our overall disbursements, we did our highest monthly disbursements in October '25 and crossed INR 8,000 crores for the first time, obviously, on the backdrop of the huge consumer demand that we saw, especially in the Wheel segment driven by the GST 2.0 Reforms. Our Urban Finance vertical clocked a 49% growth on a year-on-year basis in terms of disbursements, and this was largely led by our Two-Wheeler and our Personal Loans business with Two-wheeler crossing almost INR 1,600 crores disbursement for the first time. Our Farmer Finance, again, the Tractor business also had a push from the GST 2.0 Reforms and clocked at 21% growth rate. And SME Finance, after some recalibration has resumed its growth path at about 17%. Our Gold Loan business, obviously, is a new line of business, and it has stabilized very well post acquisition and clocked about INR 457 crores of disbursements. This is one most important thing. The last time we met, Cyclops, our proprietary AI-based engine, credit engine was still in beta phase. We actually operationalized Cyclops for our full Two-Wheeler business in the month of January. And now it's a full-fledged machine working now on our Two-Wheeler business, on our Tractor business as well as our SME business. And it's a pretty large piece of software right now in Version 3, and our Chief AI and Data Officer will take you through all the details about it. But it's about 1 million lines of code, 55-plus algorithms, 75% data scientists and engineers working behind it, about latency on the various models between 50 to 700 milliseconds, 1 million loans underwritten, and we have the performance leads for this. So Debarag, in his presentation and the individual chief executives in the presentations will give you the performance metrics of how Cyclops has been operating as well. We told the market that we will bring our automated portfolio management engine, Nostradamus to a beta mode in this quarter. We launched it in the month of August. And right now, it's live in our Two-Wheeler business. Again, 13-plus algorithms, about 200-plus banking variables getting analyzed every month on an automated basis, and we have about 30-plus engineers and data scientists supporting it. Obviously, this number will grow. And we plan to push Cyclops into personal loans in the month of November and December and lay the ground for building it for rural business finance as well as mortgage businesses next year. We have scaled up the partnership of disbursements through our large partnerships to meaningful volumes. The last time we met, we did about INR 157 crores. This quarter, we did about INR 1,138 crores, and you can see that the string of partnerships that we have added, and we are going to add a couple of more in the next couple of months. And this number is scaling at the rate of maybe about between 7% to 10% every month. So the numbers are scaling quite well. We are not only doing personal loans, but we are also doing 2-wheeler loans as well as mortgages through the large partnerships. And we do believe that the large partnerships help us get a large qualified pool of customers and the trust signals that we consume from these businesses help us actually maintain significant amount of credit quality, which would otherwise not be possible if you were to get them from the direct sales channel. The fourth large sort of block, obviously, the acquisition of the Gold Loans business. And the rationale for acquisition of the Gold Loans business was simple. Our micro loans customers actually had borrowed INR 17,000 crores worth of gold loans. So we thought that, that gave us a natural potential, cross-sell potential to tap into this customer base and build our Gold Loans business. The acquisition of the Gold Loans business has actually cut our time to scale by almost 24 months, and we got 130 branches, about 700 employees, and we finished the integration in a record time of 3 months. In fact, we launched our first new Sampoorna Gold Branch in Ujjain on the 30th of October. The concept of Sampoorna is that we not only do gold loans, but we do other products as well through these branches so that we are able to amortize the cost of the branch rollout across a couple of more business lines, and we're able to sort of harness the cross-sell potential as well. The day we launched the Sampoorna Branch in Ujjain, our local microfinance branch generated 30 leads for gold loans in 1 single day. So obviously, the thesis of feeding from the micro finance channel into the gold finance channel obviously has some merit. And last but not the least, we plan to deploy 200 new branches by March '26, which is we'll be opening 1 branch every day for the rest of the year. Some metrics, 19% year-on-year growth in AUMs with retail book over INR 1 lakh crore, retailization at 98%, sustained ROA at 2.4% even through difficult market circumstances, disbursement growth of 39%. We saw this earlier. Our collection efficiencies in microfinance business improved to 99.57%, especially in a holiday dense month where we had also Chhath in Bihar. Excluding Karnataka, our collection efficiency rest of India hit 99.0%. DPD collection efficiency hit 99.6%. Our Chief Executive of Microfinance, Sonia Krishnankutty, will cover it in more detail. Our Gold Loan Acquisition business has now hit an AUM of INR 1,500 crores. We have a customer franchise of 2.7 crores, 2 million added in the last 2 months. And our investment in the brand building, especially with Bumrah as our brand ambassador has lifted our brand association to a significantly higher level with 32% unaided recall of our association with Retail Digital and sustainable, which is our tagline. Obviously, as we gain scale, how has L&T Finance changed from inside? I thought we will give you a little bit of insight as to how our day in L&T Finance looks like. [Presentation]

Peter Drucker said that culture eats strategy for breakfast. You can strategies, you can make large presents, but strategy execution on ground does not happen unless you change the culture of the organization and focus the organization towards one common goal. And one of the things that we have been trying to do within L&T Finance over the last 24 months is to build a cohesive executive oriented risk-first, tech-first culture. What I thought is that I will unpeel some of the components of that culture change strategy and show you as to what sort of impact that we are getting out of it. Obviously, the first thing which is very, very important for us is credit culture. It's driving down to the field that if you give out money, treat it as your own money and that money has to come back with a return. And this is being sensitized right down to the field force and aided by very rigorous portfolio review processes. We have a growth orientation that has been drilled into the entire organization. The long-range planning process has helped. Everyone knows their targets month in, month out, and it is across various dimensions also implemented through a balanced scorecard across the entire organization. We have moved the organization from a silo to a matrix structure and the introduction of the Retail business structure in the 4 zones actually has made sharper operating -- operations possible with faster time to market for all the decisions. And also, the focus has been on implementing a nonpolitical and a nonhierarchical implementation mindset. Tech mindset, this is very important for us, and we have been working to lift the tech DNA of the organization, including a very successful technology for non-technology managers. Our technology teams run workshops -- and these are almost workshops, 3, 4 workshops a month where people from finance, people from compliance, people from internal audit are actually taken through boot camps on technology. They're allowed to develop apps. They are made to understand the AI tools. They are made to solve AI problems or write prompts properly. So the huge focus has been put on building the tech mindset of the organization. Needless to say, the innovation mindset and the failure tolerance has gone up in the organization where all the leaders communicate that we are a failure-tolerant organization, and this has led to a strong innovation mindset. And last but not the least, we have made a move towards a caring organization with upgrading our facilities and taking care of our women employees to the extent that you will see in the next couple of slides that our diversity ratios also have started improving. In the words of our Chairman, SNS, the entire organization is now focused on one etho, can do, will do, will get it done. But let's hear from some of our employees as well as some of our partners as to what they have felt about our organization over the next -- over the previous couple of quarters, how have we changed?

We will now hear from a few of our employees about their experiences with L&T Finance.

Hi everybody. I'm Sheetal Prabhu. I Head the Legal Team of Retail business at L&T Finance. In my 18+ years at L&T Finance, the last 18 months have been uniquely transformative. We have achieved remarkable growth, while remaining firmly within the regulatory guardrails. Our organization has embarked on a journey of multiple tech driven key initiatives. This includes working with large partners to scale our disbursements, sharpening our credit decision and portfolio monitoring with Cyclops and Nostradamus, and scaling our technology and infrastructure platform. I'm confident that with this sharp focus on building these new capabilities, we're positioned for a risk-calibrated growth with our tech and AI-enabled approach.

Hi, my name is Kunaey Garg and I'm part of the Strategy Team and the Product Head for Gold Loans at L&T Finance. I was given the task of managing the acquisition and post-merger integration of the Gold Loan business. The integration was completed in a short span of 3 months, which reflects the extent of the agility and collaboration present in the L&T Finance DNA. The business has now settled over the past 3 months. We have aggressive growth targets which reflects the top management's investments in the business growth.

Hi everyone, I'm Shruti Shetty and I lead Digital Initiatives for the SME business at L&T Finance.

Hello, I am Balaji, the Lead Developer of the Underwriting Co-Pilot Projects. Our co-pilot is an AI-powered assistant built on Agentic architecture using MCP tools. It acts as a digital co-pilot for our credit managers automating underwriting tasks. The goal is to make the borrower data analysis faster, smarter, and more consistent.

In our Unsecured Business Loan segment, our credit teams often deal with lengthy bureau reports and voluminous banking data. With the underwriting co-pilot, we have been able to cut the assessment time for these activities by nearly 50%.

I still remember during our initial discussion with our MD and CEO, Mr. Sudipta, it's a credit engine which is a next generation credit engine and it is basically product agnostic and as well as it can fetch the data from multiple data sources. It has built from our in-house engineering team into 4 to 5 months. Cyclops is designed for fault tolerance, scalability, resiliency, and observability. To achieve this, we created a common set of APIs that can work with different products. This made Cyclops product agnostic and modular.

Cyclops is developed to handle high traffic. And we have implemented in-house innovative solutions like asynchronous, nonblocking in order to achieve a high throughput and concurrency.

My name is Biswadev Banerjee, and I head, Two-Wheeler business for L&T Finance. For last one year, we have been implemented with Cyclops and what we are seeing is the improved business quality, customer profile, as well as better conversion rate which is helping not only for the disbursement, as well as for the dealers in terms of their profitability.

Hi, I am Rahul Rai, part of Nostradamus Development team. Nostradamus is helping to build a proactive data-driven portfolio monitoring approach and generate early warning signals.

I'm Suresh heading the Two-Wheeler product team at L&T Finance. As a business user, Nostradamus has become our one-stop solution for our portfolio monitoring. With Nostradamus going live, it now provides actionable insight at micro and cluster level helping us with active portfolio monitoring and recalibration of our collection strategy.

We will now hear from a few of our partners about their experiences with L&T Finance, while they talk about culture change and transform.

My name is Mayur Devendra Jain. I am the CEO and Managing Partner of Palladium Automotive, which is a dealership for Royal Enfield. I am very happy to say this, they have initiated a Digital AI platform called as Cyclops. It's a credit engine basically which helps and engage the customers to get faster loans. It is a fully digital process. It has helped the dealership as well as we have scaled up 3 times more disbursements.

My name is Rahul Kapoor. I run India's largest loan distribution business by the name of Andromeda Sales and Distribution. This last one and a half years is the big jump up on the numbers both on the HL side, on the PL side of the business and I think this jump up on the numbers came purely, purely led by the tech initiatives Undertaken at L&T Finance. What it means is the average TAT on a home loan, an average TAT on a salaried PL has come down by about I would say 60%-70%. Today I can safely say sitting here is that, when I'm planning for next 2, 3 years of business growth, L&T Finance is there in the top three; NBFC institutions for me.

At CRED, we've always been focused on credit worthy users and we had the good fortune of knowing Sudipta much before he joined L&T and through his previous assignment, he had clear understanding that we focus on the right set of customers and we were fortunate to meet him in his early weeks of joining L&T and discuss a potential partnership and commendable speed by the team that they actually got the partnership running. The partnership is thriving, we are already doing a very large volume, we expect to probably 5x, 10x the volumes that we are working with L&T and this has been a great partnership, focused on trust, credit worthy customers, usage of technology to create the most magical experience for all our members and super excited for this partnership. I wish all of the members of L&T a lot of success going forward.

So what has our culture change sort of yielding to the organization? The first and foremost thing, which is very important, is a product mindset developed within the engineering teams, which is actually helping us bridge scale our tech solutions. Our distribution teams have started focusing on risk-calibrated granular distribution. Risk has really become the question whether the customer I'm lending to is a creditworthy customer. Every sales guy is asking that question themselves before they lend. Adoption of Cyclops by Two-Wheeler team, actually, you heard the dealers say, it has actually improved our market share. Our market share actually has gone up because the dealers have more confidence in the quality of underwriting and the LTVs because the loan to values of the customers that -- the customers get has gone up. So they are funneling more and more prime customers to us. Our prime customer penetration during the Diwali period actually went up to almost 86%, 87% of our total throughput. Movement to matrix structure has made P&L ownership down visible down to the branch level. Our percolating the tech mindset down to frontline is helping in rapid tech tools adoption and productivity gains because when you roll out a tech tool, the sales channel actually has to adopt it because there is a lot of digital friction. Because we have been percolating tech mindset down to the front line, the adoption is very, very fast. And a complex tool like Cyclops adoption in the sales channel actually went like night through butter. The multiproduct Sampoorna branches will leverage the local cross-sell potential. The process has already started. And last but not the least, initiatives in talent management to build a caring organization has resulted in an annualized attrition drop of 5% and a 30% increase in number of women colleagues. What will it yield to? It will yield to faster rollout of product variants, cutting time to market. It will help our focus on AI and tech will allow us to leapfrog market share gain resistance. We saw the market share gain in the Two-Wheeler business after we rolled out Cyclops fully. Increased acquisition momentum on the existing lines of business and lower customer acquisition costs because everybody is now focused on productivity. The branch structure is actually allowing the Retail business heads to granularly focus on cost as well as productivity, and we have a detailed productivity council now to look at the spots where we can improve. Lower attrition rates, obviously leading to stability of talent density and longevity, continuously improving risk cost trajectory while eliminating cyclicality because of our focus on prime and prime customers are generally less sensitive customers to credit stress. This -- we are trying to eliminate cyclicality to the largest extent possible. And obviously, all this will kickstart the flywheel for further profitability improvement as we go forward in the quarters. What is the road ahead? Obviously, India continues to grow, right? India is the brightest spot in the world, right? And India will continue to grow. And the demographic dividend of India is real. The young and the Gen Z borrower will drive credit growth. The semi-urban and the rural, what we call Suru in common parlance will drive credit demand driven by government push because the government really wants credit flow to happen to semi-urban and rural areas. And L&T Finance is uniquely placed because if you look at our distribution map, our distribution map is equally strong in urban as well as rural. In fact, we are one of those very few NBFCs who are equally strong in rural and urban, and we're uniquely placed to take advantage of this. And last but not the least, the government push on manufacturing, the PLI schemes and the focus on MSME will actually drive MSME contribution to overall GDP, and we are very, very confident that we'll take partake in that. And we are building focused solutions for the SME business, which I will reveal a little later. The India stack is fully developed. I need not spend time on it. But the fact is that this is helping digitization. Really, if you want to do digitization in the real sense, India probably is the best place in the world apart from China to do this. What are the engines powering LTF's growth story? Obviously, depth of distribution. I spoke about distribution strength, both in rural and urban. Data and AI. I think we are one of those organizations in the country right now who has a full data and AI practice, which is functioning, which is rolling out products and which is rolling out results. In fact, in this matter, I think we are one of the leaders in the industry right now. Risk and controls, we have a seasoned risk policy and compliance team for risk-calibrated scale up, and we have added a model risk management team to ensure that the multiple models that we are using do not add to risk. Obviously, we talked about people. And last but not the least, our technology platforms, our modular Neo stack, which is in-house proprietary built, yields very, very robust performance and which Ramesh, our Chief Digital Officer, will cover during his presentation. So what are our imperatives for the next 12 to 18 months? We will announce our Lakshya-31 goals in quarter 1 FY '27. We will drive a 20% to 25% risk-calibrated AUM growth. We'll achieve an ROA in the corridor of 2.8% to 3% by quarter 4 FY '27. We will initiate and complete the build of the service intelligence layer. As I had said last year, we had an acquisition intelligence layer. We had a portfolio intelligence layer. With Cyclops and Nostradamus, more or less, we have finished a large amount of the build in those layers. The service intelligence layer is built, so we'll finish the service intelligence layer. We'll build an AI-based next-generation collection stack to improve collection yields and optimize collection costs, and we'll drive credit costs down towards a 2% threshold. So when we meet 1 year later, I'm very, very confident that we'll be able to show you significant progress on all these objectives that we are listing out here. What is our overall goal? To be a risk first, tech-first multiproduct retail financier of choice. And I think we are well on that journey to become that. Tomorrow, we have RAISE'25, which is our flagship AI conference, in which we have this year lined up a stellar set of speakers. So those of you who have time tomorrow would be happy if you can join us in the RAISE'25 sessions tomorrow. We have 2 tracks, a plenary track and a tech track, and we have speakers from all over the world coming and talking about the latest advancements in AI and specifically with its applicability to BFSI. We are also going to launch a couple of things tomorrow. And what we decided is we will give you a sneak peek on what we are about to launch, right, tomorrow. So we have talked about Nostradamus, but the fact is that we have not revealed Nostradamus to the world, right? So maybe let's have a quick look at what Nostradamus can do. [Presentation]

Well, that's not all. Let's give you a look as to how the Nostradamus strings look like. [Presentation]

Nostradamus demo is available in the demo session. Demo is available in the demo session. The demo session will open after we finish the investor presentations, and you can have a hands-on look as to how really the screens look like and what they can do. This is the result of hard work for all our data science teams, our technology teams, our risk teams, our credit policy teams and the entire support teams, the legal, compliance, everyone together, everyone works together as a team. And that's why the collaboration culture has actually helped build this in a record possible time. But I'm not done yet. There's one small other thing that we want to show you, which we are very, very proud of because we do believe that it's almost an industry first, and we have been using it for the last one month with fantastic results. [Presentation]

In fact, we have used it for the last 1 month. Our SME underwriters are using it. And a traditional SME file that takes 4 to 5 hours to underwrite. The credit managers are doing it in under 30 minutes right now using the Helios underwriting Copilot. So with that, I would invite Ramesh Aithal, our Chief Digital Officer, to take you through the next session on our technology initiatives for the last 12 months. Over to you, Ramesh.

Thank you, Sudipta. There's a slight issue with the mic. Hopefully, I'm audible. All right. Welcome back. Good evening. This is the second edition of the Digital Day, and I'm happy to be here. We have a lot to unpack today. Before I do that, perhaps we'll take you quickly through what's cooking in the BFSI industry and specifically in the technology space. Like last year, every presentation has to start with a Pranaam to AI. While last year, it was about Generative AI, this year, that morphed into Agentic AI. And within Agentic AI there are 2 areas that we are watching quite closely. The first is something called as multistep workflows with automation. This is about having the ability to use a single prompt to carry out complex workflow tasks without any supervision of a human being. The next one is on conversational and voice banking, pretty self-explanatory, and I will dwell upon this a little bit more in some of my later content. Next up is contextual finance. Again, a person walking into the branch and the bank immediately knows what they are there for or a person engaging with one of the digital channels and the channel is able to immediately service the customer based on what their requirement is before they even ask for it. Finally, multimodal authentication. This is about enhancing security with not just biometric data like face or fingerprint, but also using your behavioral patterns. All these technologies are not distant too far off in the future. They're actually actively shaping how we are looking at our digital backbone and customer interactions. Next up is our futuristic digital architecture, which we unveiled last year, the Vision 1.0. Briefly, it mirrors the loan journey of a customer, starting with the consent layer and then the acquisition layer, which is composed of customer intelligence and threat intelligence, the portfolio layer, which has portfolio intelligence and service intelligence built in, all feeding into a unified customer experience layer that is enveloped by a layer of wrapper of information security. This was the vision that we unveiled to you last year, and active work has gone on each of these segments like Sudipta referred to earlier. I'm thrilled today to announce the next iteration of this architecture, which is as follows. This includes components that we've already developed and you are familiar with, hopefully, by now, as well as newer elements that we've added to be prepared for tomorrow. Of particular interest is the real-time intelligence layer, which is a learning layer on top of these -- on top of the portfolio and acquisition layer as well as the Zero Test Trust architecture framework, which is basically using verification technologies to verify any user or device in the network. With this in mind, I'm ready to move to the next section where we show you how we're able to accelerate business, starting with the acquisition layer. Product launches. We have 4 launches that we wanted to call out. Gold loans, embedded finance, micro LAP for RBF and supply chain finance. I'll walk you through a couple of examples in detail. Gold loans. Like Sudipta referred to earlier, this is a new business that we acquired early part of this year. For technology team, the challenge was to get the entire integration and the day 0 setup done in a matter of 9 weeks. 12 weeks was the entire integration effort. Technology team had 9 weeks to deliver. And I'm happy to state that we were able to deliver that able to build a loan journey collections and a direct-to-customer channel, 100% regulatory and risk compliance with connectivity for all branches. All of this in a matter of 8 weeks -- 9 weeks, sorry. And with gold, fairly unique business, fairly sensitive information and data. We've built an extensive network of multiple live command centers with over 2,000-plus cameras across the 130 branches, growing at the rate of 1 branch a day, 24/7 monitoring in place with centrally controlled 2-factor authentication for vaults. A lot of this was a fairly steep challenge for the team, and I'm really happy to state that we were able to bring the organization acquisition to the technology benchmark that we've set within the company. For us, this was a transformation built on discipline and speed. from gold, another shining example of execution, which is embedded finance 2.0. Embedded finance is about leveraging partnerships as our growth engine. Particularly for L&T Finance, this is about acquiring -- using not just our traditional channels for customer acquisition, but going to where the customers really are. On the left-hand side of the screen, you can see a few of the partnerships that we've listed out. We have marketplace partnerships here. We have OEM partners, and we have fintech aggregators. All of these partnerships have been scaled up quite extensively over the last year. For us, this might seem -- for an outsider, this might seem like a fairly plug-and-play model to develop. However, it's not been the case. We've had a fairly delivered a fairly customized journey for each of our partners. We've been able to do that with complex API orchestration, a unique credit policy and deployment across all of these partnerships. And like Sudipta referred to earlier, we've scaled it up quite massively in the last 2 quarters from 3 to 10, which is fairly record time with an increased disbursement of 2.5x. We've been able to achieve all of this by sticking to some fundamental engineering principles, a so-called ready kit, if you might, about using modular journeys and open API stack. Once the customers or once the partners are integrated, we're able to scale horizontally with using cloud-native technologies and smart load balancing. Again, a stellar example of how we've been able to move the needle on all of these partners in a record time. For us, this is how finance becomes invisible, intuitive and truly embedded. Moving on to a few more examples of now how we've brought in AI to change fundamentally how we do sales augmentation. The first example I have for you is on AI-based outbound calling. So we've taken outbound calling and added intelligence to it. What we've done is we've rolled out an AI-based voice bot for our outbound calling for prequalified leads. This is in the Personal Loans business. We rolled it out last quarter. What it also does is apart from regular calling, we are able to provide real-time dashboards, quick lead filtering and do this in parallel. The initial results are quite positive, and users have reported an 80% improvement in TAT in lead engagement. Let me show you a brief clip of what's possible today. [Presentation]

Our relationship manager, in fact, did call the customer at 5:00 p.m. and converted the person. What you perhaps already have seen is the human-like capabilities that's built into the bot as well as the multilingual ability that's on display there. The next example I have is what we call as a loan offer pod. For us, maximizing customer lifetime value is a key metric that we live by. Earlier, the process used to be fairly cumbersome, manual and not as customer-friendly as we would have liked. What we've taken is we've taken a fairly scientific approach to that. We have an analytics smart from which we have a lot of data that's internal to our databases, push it to a loan offer engine with ML models in place and are feeding that into the customer-facing channels with a feedback loop in place. This allows us to manage the life cycle of an offer fairly effectively, in fact, with very, very great results. This has a full end-to-end automation with live nudges to customers and AI and it's augmented with AI capability. Another example of how AI is really maturing in our sales augmentation process. Moving on in our acquisition layer from the credit intelligence. Credit decisioning is a fairly dense ecosystem that's engineered for scale, complexity and accuracy. Today, I want to again highlight how we've built one of the most advanced credit decisioning engines in the industry. Let me start off with a small example of a loan process. As a new customer comes in, they get in through the log-in process, and we run a lot of event-driven checks, which includes face match, liveliness, DigiLocker, VKYC, as well as contactability checks, which has geospatial intelligence built in. From there, we run into the cross-entity intelligence layer where we conduct our risk assessments, both income as well as other risks that might be involved in the customer. All of this data is processed at record time and fed into the cognitive and risk layer, which is a bunch of models, a complex interplay of all of those models to do your internal dedupes, anomaly detection and so on. Finally, this is fed into the autonomous decision layer that we have engineered. Last year, until last year, this was just the BRE, the business rules engine. Since last year, we have Cyclops at play with its alternate data and trust signals. With Nostradamus that Sudipta just talked about, we will fill the third cog in this massively scalable autonomous decision engine. Dr. Debarag will talk about Nostradamus in his presentation later. I will double-click into Cyclops a little more in my next slide. Before I move on, I just want to call out, right? All of this decision-making happens at rapid fashion, which I'll talk about as well, but 50-plus validations to approve one case, one simple case is what we are talking about here. Cyclops needs no introductions. It's our next-generation credit underwriting engine. What I'll talk you through is a simple case of a thick file processing. Thick file is for customers that have a credit bureau record. And when we feed it to -- for our simple 2-wheeler customer, we feed it through a bunch of models. I'm representing that quite simply here for sake of simplicity. So taking up first is the banking and account aggregator model, which makes a decision, whether it's a go or a no-go and no-go might be because of either incomplete data or missing data. It doesn't matter. We built Cyclops to handle all of that. Fed through a few more scorecards, a few more models. And again, decisioning goes on into payments, account aggregation and credit scorecards and finally resulting in a clear decision at the end of the process. All of these are actually a bunch of 55-plus models, 20-plus APIs happening in the span of 3 seconds, less than 3 seconds, pick file customer. As I said before, right, all of this complex orchestration is about the maximizing our ability to underwrite credit for the most number of customers. That's all this is geared up for. We built Cyclops for volume, veracity, velocity and variety of data. Each of these factors are critical to the way we've scaled our engine from a technology standpoint. As Surita referred again, we've rolled out to a few of the business lines already. Personal loans is under implementation and will be rolled out over the next month or so. All in all, Cyclops continues to be a great engineering challenge for us, and we are up to it. I will move on to the portfolio layer a little more deeper because that's an area of focus as well as we continue our journey. In the portfolio layer on the collection space, we are gearing up to unveil our new futuristic blueprint, which is a completely AI-based full stack collections engine. What you have on one side, starting with the data analytics layer, which feeds into an automated allocation engine, then all feeding off into the optimized channels that is determined by this automated collections engine, finally ending up with real-time alerts and insights in the command centers that we have. We've started to work on this blueprint. And hopefully, in the next few quarters, we'll be able to demonstrate progress in each of these layers. The intent, obviously, is to get -- make collections smarter, resulting in a superior portfolio performance. Moving on, through our flagship product, B2C product, PLANET 3.0. We unveiled PLANET 3.0 last year in this very conference. This year, like perhaps every cell phone in our pockets, it's smarter, bigger, better and bolder. What I do want to show is the different areas that PLANET touches for our customer base, starting with collections, servicing, engagement and the 360 business enablement. Collections specifically, it is one of the most cost-efficient collection channels we have. Over the similar time period of first half of last year, we've doubled the number of collections that we've managed through this app with INR 120 crore plus high-DPD collections as well as 35-plus charge collections purely through app minimal effort. On servicing, we have the industry best servicing numbers through the app. We are at 88% of servicing across all our servicing channels within the company. We continue to strive for 90%, fairly elusive given our rural base, but we're working on it. We have document servicing, payments, profile updates and customer feedback, all feeding into over 236 options that are available today on the app, ensuring convenience for our customers. PLANET is a force multiplier with its DIY journeys. Over the last 2 years, we've accelerated cross-sell volume by 4x and new to LTF volume by 3x which is made possible by our approach to the smart digital enablers, which are integrated journeys, faster fulfillment and enhanced cross-sell. All these improvements have taken a significant amount of effort from the team. But like I said, it's a fairly low-cost DIY channel for us, and we continue to invest in this platform. Moving on to customer engagement. It's been 10-plus years perhaps since Alexa was announced to the world. There have been a few attempts being made at voice and conversational banking and perhaps none of them have succeeded so far. But with advancements in AI, we wanted to take these challenges head on. So today, I'm going to reveal to you the first industry first in industry conversational voice agent built into PLANET. I'm going to stop conversing now and let the agent do the talking. [Presentation]

Hopefully, you noticed its amazing ability at interpreting multilingual commands from the user, emotionally aware response and how it managed interruptions very well. A lot of human-like features have arrived with bots, and we are leveraging all of that. We built all of that, in fact. We are using this for our existing and new loans. This product is already in CUG as of last month. Very soon, we will be rolling it out into production over the next month or so. We have demo stalls next door demonstrating this capability, and I encourage each one of you to actually try this out after this session. And hopefully, if it works well, I would encourage you also to download our apps and become our PLANET customers. Moving on, in the ecosystem today, from a Planet was about customer centricity. But customer centricity also means we look at our customers from the point of different touch points of their engagement, not just from an organization standpoint and the fintech ecosystem, but partners. Partners in the ecosystem are a fairly important aspect of how customers through which customers engage with us. We send research teams on the field and determined that the dealer community or the partner community has been underserved from a technology standpoint in the competitive landscape. And therefore, with this in mind, we are launching our all new Partner Planet today. This is a mobile app, specifically for our partner community, which is right now rolled out to our 2-wheeler and tractor farm dealers with amazing capabilities that are built and taken for granted as part of Planet. It comes with our own personalized dashboard with one-step trade advance withdrawals with comprehensive details that they're always looking for and never could find a real-time dashboard indicating their pipelines of disbursed customers with a complete portfolio summary that they can review from time to time. For us, the approach we are taking is that if -- when our partners win, we win. From here, I come to the last section of my presentation, which is on the tech modernization for sustainable growth layer. A lot of what we saw earlier is about technology-enabling business. All this is not possible without our deliberate attempt or deliberate strategy of tech modernization that we've taken within the company. And what I will do is take you through a few areas of focus today. Starting with Agentic AI, a product mindset that Sudipta referred to earlier as well and engineering resilience. Starting with Agentic AI. I've walked you already through examples of how we use Agentic AI to influence customer experience. We've also looked at how Agentic AI plays a role in decision-making. What I will do is take you through how it's impacting productivity within our technology organization. The technology development life cycle is a fairly standard life cycle, starting from product design and moving on to development, testing, launch and obviously, requirement creation at the beginning of it. We have augmented all these 2 capabilities within the tech organization with agents built into coding, testing, product requirement gathering as well as migration of legacy applications. We've started the approach late last year, and that shift to adoption, early adoption has given us increasingly positive results. Even with a 60% adoption rate, we have a 30% coding adoption rate and 15% improvement in productivity, and it's early days for a lot of these tools. What I'm also proud about is how we've deployed AI into our production support life cycle, starting with predictive monitoring. We are able to detect inconsistencies in our performance and latency throughout our production support chain. We're able to scale our systems intelligently using the technology tool sets we have. We're able to remediate complex workflows through bots as well as use AI for managing our security. AI-based threat intelligence detection is a fairly accepted use case and being deployed widespread across the industry as well. What I will show you next is how all of this is actually influenced by the product mindset. Product mindset, like Sudipta called out in his talk as well, is a dramatic shift from a solutions mindset that we used to employ earlier. Why take up a product mindset? Because it's an investment in building our strategic competitive advantage. The reasons are fairly obvious. We are able to better manage our user end experience. We are able to go to market faster with full control and customization and keeping data security and compliance in mind. A few examples below with familiar names with Cyclops, Nostradamus out here, but a lot more names that you might not be familiar with. Our investment in productization continues in the last 18 months, and we expect to take this to the next level over the coming year. The final topic in Tech Modernization section is around engineering resilience. Over the past year, there have been fairly widespread large-scale outages that have disrupted customer trust. All familiar names, hopefully. And looking at this, it's a wake-up call for all of us. Building engineering resilience is one of the core pillars of building up our strategic competitive advantage. And in that -- to that end, we have a 4-pronged strategy. The first is about observability. Building real-time observability into our production systems have enabled 70% of incidents to be revealed to us before they actually impacted business. We are able to scale up all our customer-facing apps on the cloud, starting with hybrid, starting with single cloud, moving on to multi-cloud and then today in a hybrid cloud mode. We are heavily focused on maintaining operational continuity through third-party resilience and interdependency mitigation, which is, as you could see from your slides earlier, is given the complex ecosystem is the need of the hour. Finally, we want to wrap it up with an approach to something what we call as the always-on architecture, where our system reliability, we want to take it up from 99% to 99.9% or 100% as well. I'm really happy to report that a lot of progress on some of these has already been done, and we continue to invest in stability as shown by some of these dashboards here. A quick peek at the impact. Over the last year, we have doubled the deployment throughput, in fact, 2.5x the throughput, increased API costs by the similar number and reduced support tickets and disaster recovery time by over 50% in our applications. With Cyclops, we have very strong numbers to report as well. We've already seen benchmark numbers in earlier slides from Sudipta as well. We have average API latencies as low as 30 milliseconds in a lot of our apps. We continue to invest in this capability, and there are very stringent benchmarks that the teams are working towards to maintain as they enhance these models that we've built. Cyclops volume handling capacity, we have referred to it earlier as well. Here are some numbers. Over the last 7 months, we have a 3x increase in volume with peak volume days going anywhere from 5 to 10x on similar days compared to last year. Cyclops has been rolled out sometime in June or May, I should say, last year. Over the last 18 months, it's not had a single downtime. So it's with pride that I can say that Cyclops doesn't blink. And the reason it does so is because it's engineered for zero downtime. To conclude, a few common themes resonate with us across the entire company, and we continue to invest in building scale. Looking ahead, for us, re-engineering business workflows is an important element of what we will do over the next 12 months. We will continue to invest in AI in our collections stack, like I showed you the blueprint earlier and the service ecosystem that we have today. It's going to be heavily influenced with agents and bots. We have augmented our capacity at productization in-house. We will continue to invest in this capability. And like I said, we will be more and more pushing ourselves towards milestones in an always-on digital architecture that we have laid out. With that, I'd invite our Chief AI and Data Officer, Dr. Debarag Banerjee to take the stage. Thank you.

All right. I see a lot of empty tables up here. Why don't some of you come up because numbers are beautiful and I don't bite. So seeing no numbers. Lending is a relatively simple business. You have your onboarding yield as the primary income coming in. And then your credit costs, collection costs and operating costs adding into it. And what you have left is your ROA. Now we deployed Cyclops. You've heard all about it, and you'll hear more. The idea behind Cyclops is fundamentally, it brings down the collection -- the credit cost, right? And I'll show you in which way. The idea we have lost the numbers is after disbursement, it helps monitor the portfolio in a way to make it actionable for the collection teams to bring down the collection costs. And finally, ultimately, the net action of all of the Agentic AIs that we are talking about here today is to bring down the operating cost of running the business. What happens as a result is what you have left, the ROA goes up. Now that gives us the strategic flexibility to go after better customers, optimize our LTV according to a very well-understood segmentation of risk profiles, and that leads to growth both in ATS as well as in total customer volume and books and disbursement and so forth. So AI does not just give you profit, neither does it just give you growth. It gives you both. It's not a zero-sum game. I will now walk through each of these pillars and show you how that happens. So we launched Cyclops about a year ago in our 2-wheeler portfolio. The first version of Cyclops was about bringing in trust signals from banking set data, from payments characteristics of the user and about the affluence of the area that they live in. Since then, we have taken both the individual model and their Ensemble models and have added new data sources, things like digital footprints from the mobile that they are using, things like more indexes coming from credit bureaus. And we have added beyond the original 15 scorecards, scorecards related to dealer grades and past flow frauds. Combination of all this lets us characterize any particular applicant into 1 of 5 segments from the best to the worst. And that allows us, like I mentioned earlier, to give the best LTV to the best segment and the most risk-adjusted offer typically lower LTV to the lower segments. It is not a static engine. We have -- like I mentioned earlier, we have upgraded Cyclops twice in this last year. And we are about to embark on another upgrade where we are going to incorporate a new model called Shikhandi model in the Cyclops scorecards. The idea behind the Shikhandi or Mule is that a nagging problem in the 2-wheeler business is a customer with a bad credit who wants to buy a bike, knows that he will get declined. So he brings forward somebody else with good credit and says that, okay, this is the borrower. And when that happens for any lending agency, it would look like a good customer, you give them the money. And within a few months, the bad customer starts defaulting and you basically lose that loan. Now we have created AI-generated data relationship graphs, combining everything from the applicants info, the assets info, the credit footprints as well as insurance and other related data sources to create this model that can predict if a customer is truly a mule or a good customer. We, through our back testing as well as through live actions on the field by our risk team have seen that this leads to a reduction in the probability of default in those lower segments, the segment 4 and 5 that we talked about to a level of 66%. So stay tuned for that next generation. Now we can talk about models till the [indiscernible] home. But ultimately, like Sudipta said in his presentation, it's the culture that brings it all together, right? So models alone will not be effective in a highly regulated interconnected industry like the one that we are in. We have to have the entire organization run in symphony collaborating across business, across credit teams, across data teams as well as all the compliance-guided risk verifications of the models as well as the policies. And finally, the digital back end that Ramesh talked about as well as the front end from which the customer application is taken. We have made all this work in a way where we can make these initiatives, Cyclops, Nostradamus and a whole lot of others in record time. To do that, we needed to upgrade our AI team. At the time when I came in, we had an analytics team and we had data teams separate from each other, but we didn't have true AI DNA, which over time we built up. And we upgraded our systems. But today, we are fully not only on cloud, but we are going multi-cloud. Not only are we using one kind of LLM and staying there, but we are -- we have an open system where we can bring in any kind of LLMs if needed in our scoring methodologies. And most importantly, we have become a lot more rigorous about our data governance tools and are -- have an ops methodology where we can develop and launch AI models in record time. And of course, if you want to go further rather than faster, you go by building a community, go with everybody. And that's the reason why we do this RAISE kind of events. Now with all this infrastructure in place, the other thing to understand is that this is not just flipping a switch. It took us a long time to build and refine Cyclops. From the time when we had the original concept to the time when Cyclops could go GA and actually this was a loan to the first customer, that was about a period of 8 months. Now after that, if we went Cyclops 100% of the book, that would not really be the best scientific way of doing it because then we would not know how does Cyclops VF compared to non-Cyclops previous ways of underwriting. So what we did is we did a graduated A/B testing where we slowly increased the number of cases that were coming through Cyclops comparing against non-Cyclops performance, looked at the difference and then when we are convinced that, yes, this truly is dropping our credit cost down. At that point, we went 100% Cyclops from December, January onwards. Since then, we have upgraded Cyclops twice, like I mentioned. But even then, a loan is about a 2.5 year loan. So how do we convince ourselves that, yes, we really have a credit engine that will produce lower credit losses? The answer lies in looking at early indicators. For example, if you look at GNS, which is the percentage of people who have missed their first payment, that number before Cyclops to where it is after version 2, we are about -- at a level that is about 2/3, actually less than 2/3 of where it started from and keeps continuing to go down. Other way of looking at it is the vintages. And there, there's a very interesting story here. So when we build Cyclops, we backtested it with data from more than 12 months ago -- loans that were disbursed more than 12 months ago. And when we launch Cyclops, like I said, that we had a Cyclops and non-Cyclops experiment going in parallel. Now the deep blue line is the vintager of how 2-wheeler loans were 12 months before Cyclops was launched and the gray line is how it was when Cyclops was launched, but this is non-Cyclops performance. Now what you notice is that during that time, in 2024, the industry had actually worsened, right? That's the reason why the vintager went up. But then we made subsequent versions of Cyclops fine-tuned to adjust to the reality and also add more score cards and so on and so forth. So where we are today, our 30-plus DPD rates are just about 1/3 of where the non-Cyclops rates would have been. So now if I assume that the flow through -- the flow forward rates and so on and so forth are going to be the same, which, by the way, they would not because of loss of the actions that I will talk about later. But even if I assume that, this leads to a logical a priority conclusion that, yes, everything being as is, we would end up with 1/3 the credit cost that we would otherwise. Now Cyclops was not a onetime success in only 2-wheelers. A very different part of India, the rural farm India, right? We deployed Cyclops with very different sorts of models for farm tractor lending. And there, we needed to access very different kinds of data in order to create an agro model that by looking at soil types, by looking at rainfall and other patterns, how it would determine the yield of a field. By looking at other unstructured geo intelligence data, things like how far a village is from the mandi, how is the infrastructure look like, how does the mandi prices over there look like, et cetera, we needed a different type of data sets. And finally, unlike 2-wheeler loans, farm loans are very high-touch loans where there is a field investigation for every loan that is processed. And there's a rich ton of data that gets generated there that until recently, people did not think of as data, but now we can actually extract the right features from it and then build a model that can make the right decisions. How do you know they are the right decisions? So Farm Cyclops launched in December. It is still in that A/B phase that we talked about. And it's about -- right now, it's processing about 30% of the portfolio. By looking at the early trends, the NNS trends, which is the number of people who are missing their first month of payments, we, as a result, can see the comparison between Cyclops processed and non-Cyclops processed loans, and we can see that the NNS for Farm Cyclops processed loans are about 8x lower compared to non-Cyclops processed loans. Early results, lots to work on, but it proves the basic thesis that Cyclops is more general than just 2-wheeler loans, and it can, as a result, be applied on pretty much all LOBs in our portfolio. Now like I mentioned in the first part of my talk, it's not just about increasing profits, it also increases growth. And here is proof. So 2-wheeler portfolio went Cyclops from December, January onwards. Since then, the disbursements have actually not fallen. So profit per deal has increased, but the number of deals themselves have increased. And in fact, this huge spike is because of the last fantastic festival month that we had. The increase that you see in Farm Cyclops, part of that is because we are increasing the ratio of Cyclops processing in the portfolio, but there is underneath the numbers, there is also a generic uptrend as well. SME loans have went 100% Cyclops last month. So September to October, that is a pure growth of the portfolio, while at the same time, we are applying this multifaceted checks underneath. And our PL portfolio is due to go Cyclops very soon, followed by HL, followed at some point of time by the RBF loans. Now every loan looks one way before you disburse it. After disbursals, the customer chooses real [ toughers ]. How do you differentiate within the customer, let us say, a college kid who just bought a bike when he is in college and looks just like any other customer with no credit and history and all that. But then later on, he graduates from college, hopefully builds his own business, takes an SME loan, does well in life, does many purchases with his PL and credit card and so on and so forth and then finally buys the house and perhaps upgrades his house and so on and so forth, eventually becoming a prime customer. So how do you discover this undiscovered prime early on and differentiate him which is a more important aspect, differentiating from a customer who, at the time of disbursements looks like the same customer, but you see that, okay, this is actually somebody who is teetering on the brink and a little bit of cross winds from one way or the other during the -- as the macro economy goes up or down can throw him off the board and he ends up going into default. And how do you do that before the default happens? So our answer is that we, again, look at multiple data sources every month on each of the customer categories and try to understand how up to the level of every single customer, how are they doing. And we do this not only based on credit bureau data, which is a little bit post priority, but we also, with consent from the customers, look at their bank statements, which gives a better signal about their income and expenditure, look at how our peer companies are working, how the macro economy is working based on open data sources, not only at the national level, but all the way down to every single district and how that mixture is changing. And of course, given the penetration of Planet and other data sources in an increasingly digitally connected India, we can bring all of that device data, all of those digital footprints, understand the location and the location mobility of the customer and bring it all together into this Nostradamus engine. The Nostradamus engine allows us to look at the portfolio, understand early warnings as they go up or down and identify those clusters where we see that the portfolio has real growth potential, can grow more, and those are our green shoots where -- which enables growth. But it also highlights the areas of concern, whether that's a geographical area or a particular customer segment where we may want to tweak our collection methodologies in a way that we get the best outcome. It is brought to the credit and the business teams in the form of a very intuitive dashboard that we took quite a bit of time designing in our data and BI teams. Here's, for example, a typical indicative view. So for example, here, I'm looking first at the pan-India picture and trying to see how the 30-plus DPD moved within a space of 12 months for the 2-wheeler portfolio. Now here, you see that, that number, the 30-plus DPD has actually improved by more than 220 basis points. This, by the way, if you paid attention. And a year ago, there was very little Cyclops, right? Now there is more. So it kind of is an indication -- yet another way of indication that, yes, Cyclops is actually making the portfolio better. But what is more interesting is that the lift is not the same all across India, right? So some regions are a little yellow compared to the greener parts of India. This also allows us to drill down to those districts that are green -- sorry, that require attention that are yellow. Here's an example. So let's say, we look at the district of Coimbatore. And you notice that although the 30-plus DPD here has improved, actually improved more by 388 basis points, but it is still not at a threshold where it is better than the national average, right? Now that would typically mean somebody needs to look at it. So they can drill down into it and try to understand why that is so. So it turns out that, oh, by the way, when we look at the dealer-by-dealer distribution, dealer category by dealer category distribution, there's more of the deals coming from the bronze dealers who typically exhibit less affluent customer characteristic compared to the platinum dealer, which leads to a potential idea for that area's business to look into, which is can they make a shift in terms of their sourcing patterns more towards the platinum and less towards the bronze. This is not generally applicable, right? It is only for that specific segment for that specific business. So that is the level of granularity with which it enables decision-making. Now we could go at it with the dashboarding approach where ultimately, it takes a human to click through and think in his mind and get to all these decisions or we could use AI. And we could use AI on top of all the data that is underneath. So we basically put an MCP server on top of our data lying in BigQuery. You don't need to know what that means. What you can see is it essentially enables a ChatGPT type of interface, but you can basically chat with your data. You can say, "Hey, tell me how is my default trends going? What kind of customer segments do I have more early default contingencies? And it can finally summarize all this for you and say, hey, here's a snapshot. This is the sliced and diced segment where you may want to look at. And then ultimately, it's up to the credit manager to decide what to do with that. It is not an autopilot. It is only a CoPilot. Now how will all this come together, Cyclops and Nostradamus? So Cyclops, like I mentioned, it starts from data and it ends up with an onboarding decision, yes or no and how much. That loan gets disbursed many times over, create a portfolio, which is looked at every month by this Nostradamus engine through this multidimensional scale. And from there, either decisions can be made by somebody looking through dashboards or with this chat interface, you can call it that way, which we call Orion, and the decisions that the person makes, whether it's a credit or a collection or a business executive would be, for example, what kind of collection actions to take, what kind of agencies to allocate what to, how to build roll-forward models and so on and so forth. Moreover, the original Cyclops engine, as I mentioned, it's ultimately a very flexible engine where you can actually modulate the dials and change the thresholds on each of those models and make it better to respond to what you have discovered, right? So that is also driven by the insights that you gain from Nostradamus. And you can say that, okay, looks like, for example, this particular LTV category in this particular sets of customers who have this type of data, but not that type of data, I can see an increase in the early warning score, which is basically yet another model under that gives you that forward-looking score. And you can say, oh, by the way, I'll tweak Cyclops so that for that specific consumable model, I'm going to change some of the credit decisions about the type of rate or LTV or conditions or documentations and so on and so forth that, that category will have. Now this you can do with humans, but we have 20 million portfolio of about, what, 20 million-ish loans. What is the optimal portfolio that you can create from that? It's actually a very hard problem because of many, many reasons -- many mathematical reasons why standard techniques can only get you an approximate result. And if you truly want to find out what is the actual optimal of that many loans, that's the number that the fastest supercomputer in the world will take the life of the universe plus more to get there. What we are exploring, and this is still very much an early exploration, it can be possible, not possible, but some forward-looking companies -- some forward-looking BFSI companies in the U.S. and Japan have -- are trying out quantum computing to arrive at that optimization. We are in the very early stages of thinking about applying similar methods to see whether we can optimize our portfolio even better. Coming to the Agentic AI part of it. So Sudipta showed in the Copilot launch, how we have built Helios, the AI CoPilot for underwriting SME loans. I'll just summarize it. Essentially, you have tons of data from Credit Bureau records, from deviation -- from bank statements and GST records and so on and so forth. These are things that come to an underwriter, hundreds and hundreds of pages from which he has to come up with a handful of metrics and decisions and questions that they want to go and ask the customer. All of that is done by -- by Helios and is presented in a very easy to consume form, which saves hours and hours of the underwriters' time. And that time has two dimensions, by the way. One is, of course, you can do more with less. But more importantly, this means you can have a faster TAT. And in the SME business, if you are able to give an answer quicker, then the more of the better customer deals will come to you, and that's another way in which the portfolio improves. Not only are we using Agentic AI for internal usage, we're also opening it up to the end customers for both existing loans that they may want to service or new loans for which they may want to understand how it happens. So on the left, you see our KAI chatbot, giving answers to somebody who has a home loan and -- a farm loan and says, okay, "I want a statement of account." It asks, "Which loan are you talking about," and it delivers that specific thing to them, entirely driven by a RAG-enabled agentic chatbot. The same exact chatbot, by the way, you can ask questions about, hey, I just heard, you have a new gold loan business, I have some gold, how can I get a loan? Is it going to safe and all of that. And it gives you those answers in a very human form. So this today is available on the website. You can check it out on our website or the demo with over there. Or -- but very soon, this will also come through a lot of other channels to the customer like WhatsApp and other messaging facilities. Not only are the chatbots available in English. This is live today in 11 Indian languages, and it's went in all of them. And not only does KAI talk to chat with you as text. But when a customer is delinquent on their loan and it is time for them to get a call from a collection agent, KAI, the agentic chatbot, the machine can actually be that collection agent.

[Foreign Language]

This is a real-feel customer who went -- who missed one of their payments, and...

[Foreign Language]

The exact status of the...

[Foreign Language]

And can [indiscernible] to get a promise to pay.

[Foreign Language]

And once the customer says, "Yes, I'm ready to pay," can also send a payment link [Foreign Language] and then walk the customer through the process of making the payment itself. [Foreign Language] We have turned it on less than a month ago and so far, about 50% of the time KAI is effective in getting those payments and it does it in 11 Indian languages. [Foreign Language] Not only are we using voice as our only multimodal form of GenAI, one of the other things that we are using it for is, let's say, there is a Microlab customer -- Micro-LAP lead who being in the rural area is tens of kilometers away from where our branch is. And so the -- when this lead comes in, somebody has to make a decision whether to go all the way there, look at the house and decide whether the house is good enough to be worth further pursuing the loan or it's a shabby little hut where the value simply isn't there to bother. Today, what we are running a pilot on is where we can send a link to the customer that walks him or her through the process of taking the right kind of pictures of the house, which are scored by a GenAI engine and then it can give a suggested decision under to the sourcer whether to take forward the deal or not. Ultimately, we leave the last decision to the sourcing, but this is a great way in which the operating cost for sourcing for any lapse can come down. Another end of the loan cycle is settlements, right? So when a farm tractor has been reposessed because somebody could not pay the loan back, then one of the big decided -- big decision points of the value is whether the tires on the tractor are good, bad or ugly. And at the time when these tractors are taken into the yard those pictures are taken by the field agent. Today, we have a GenAI engine that with 88% accuracy can say whether the tractor is in a good or poor or somewhere in between condition. And this is not just theoretical. The resale value of the tractor can go up or down 12% based on whether it's good versus poor. And making that decision both allows us for the case of good tractors to bring in the full value of the tractor and for poor tractors to discount it enough so that, that inventory flows out quickly enough. Our gold loan business has started. There's hundreds of branches where theft is a big concern, and we have multiple video feeds with which we look for it today with humans. We are actively looking at video anomaly detection solutions, which can alert these humans to possible break in so that they can actually react faster and more accurately. We are also bringing AI and democratizing its usage to all 35,000 of our employees. Why? Because there are so many ways in which today AI can help an employees' life by doing what they need to do faster, perhaps more accurately and in a way where they can refashion their own workflows that we can't possibly design all of those from a top-down manner. So instead, we are a big believer in enabling AI to all employees within the constraints of a protected cloud so that PIIs cannot pass away and other guardrails are there so that misuse and dangers from AI does not happen. Finally, -- we all talk about AI. But recently, there was an MIT report that came out that 95% of AI deployments fail. So if you go deeper, 55% of them fail because the data is not of good quality. Fields are missing or they're wrong or there's a gap in the data stream or something or the other. We have built data pipelines that are very rigorous, entirely run on the cloud, very high availability, and we are in the process of putting data governance guardrails and business catalogs -- business name catalogs so that it becomes easier and easier for not only people who are building AI agents, but also normal business users to find the right data and build the right models. At the end of the day, without good data, you can't have good models. And because we started from data first, that is one of the reasons why our model simply work. Thank you very much. And I would invite the Chief of our Rural Business, Ms Sonia Krishnankutty to come. And call to action after all these talks have done great demos. Please take a look.

Good evening, everyone, and welcome to this event. After a lot of data, tech and AI presentations that we have seen, let's have a look at how all of this has helped businesses scale up and build superior quality portfolios. Before I start off the presentation, let's have a quick recap of when we met last time, some of you would have been there. It was the midst of a large looming crisis in the JLG industry, caused partly by external factors, partly by internal factors, overleveraging of customers, and there were large questions about whether this industry will survive this crisis and how long is this going to last. And it was at that time we came in and we showed you a very granular view of how our business has been built on very strong controls, robust structure, very stringent policies and credit rule engines, which work behind the business and how we have built this portfolio. Now with a year into the crisis and since we have completed 1 year after the crisis started, let's have a look at how our portfolio has trended during this time and how have we been able to keep our portfolio resilient and manage collections in a superior manner. So where are we today? Right now, we have around INR 26,000 crores of book spread over 61 lakh customers. We have a pan-India rural presence across 17 states. The portfolio has exhibited extreme superior collection performance as well as quality. And after having built this portfolio, we -- when we realized that we are very deeply integrated in building sustainable rural livelihoods in the women customers, 2 years back, we also felt the need to be embedded in the entire rural ecosystem by providing financial service to the non-women customers, probably by way of a secured finance. And that's when 18 months back, we launched this product, which is Micro-LAP, which was a large opportunity out there in the field, primarily focusing on male customers who run small businesses, which have stable cash flows coming in and who would want to use the money for their business needs. So that's how we started this business, and we are building on this business. So let me take you through details of the JLG business before we come to this product. So if we look at the entire presence, as I said, we are present in 17 states, around 350-odd districts, 2 lakh-odd villages, and we have around 2,100 points of presence spread across the country. As you can see in the map over here, we have a couple of geographies which are pretty large, which are our points of presence, which are real strength areas, I would say. We have some markets which are in the medium to large category, which are INR 1,000 crores to INR 3,000 crores and a few markets where still the book is below INR 1,000 crores. So this actually provides us with the geo strategy the way we expand in large geographies where we have a calibrated growth, where in geographies where we are medium to large where we deepen our presence and in geographies where our book is still below INR 1,000 crores, these are the geographies where we expand. Now if you look at the Pan-India growth, which has happened over the last 8 years from INR 7,500 crores to INR 26,000 crores, which is a 3.5x growth in the last 7.5 years is what we have achieved. And we are the second largest financier today in the JLG industry with a 7% market share. Now if you look at the growth phases, it's very typical of what the JLG industry goes through, as all of you would know that it goes through a very good cycle for 2, 3 years, then comes a downturn and then again, it's an uptrend. What you notice over here is a very unique pattern wherein we have been growing during the good times. And during times which are difficult, we have actually used the time to strengthen our processes and controls, putting systems in place to ensure that when the next spurt of growth happens, we are there ready. Now if you look at this, this is the last 1 year, 1, 1.5 years of credit cycle looming. We have invested a lot in data and technology, as Debarag and Ramesh had explained, so that we are well poised for the next growth. Now if you look at the same picture, if you look at it from the market share perspective, this is how it pans out. So our Pan-India market share being 7%, we have 5 markets where our market share is more than 7%, around the same number of markets in the 5% to 7% and some markets in the less than 5%. Now this offers a very multipronged strategy wherein our market share is on the higher side. Again, there is a very calibrated growth in these markets which are 5% to 7%, we deepen our presence and in the less than 5 percentage market share are the geographies which we plan to expand. So together, if you look at the market share has moved from around 5-odd percentage 3 years back to around 7.6 percentage. This is a time when we had invested a lot in our processes and which actually scaled up our business in the last 2 years. Now with this, we are well poised for a very calibrated growth in the geographies with deep presence and in other geographies where we have a lesser presence. Now if you look at how this entire last 2 years, what all we have been building on as has been explained in the previous presentations, through the life cycle of the loan, which comes in your system through the different checks that happen, ending in disbursement, data and technology are strongly embedded. So when at the time of data entry, whether it is about validation of the KYC of the customer, validating the customer, whether the customer is genuine, geo tagging of the customer to ensure that the geo limit is as per the prescribed norms, there are different partners who help us in ensuring that this validation happened real time. And now when it comes to the credit check, there are various rule engines which happen behind the entire loan system wherein there are internal dedupe checks which happen, which checks whether the customer is already a customer in any other product line, whether there's an employee check done to see whether there is any dummy applications coming in. So there are a lot of risk checks which happen over here. And what passes through the credit rule engine grants real time to check the exposure association as well as the DPD norms. And within a minute, the credit sanction is given. So this helps us ensure that we are communicating the decision with the least possible time. Now the third point, this is about the huge repeat base of customers. It's very important to know that these customers who are already on the book at what time they become eligible for a loan and what sort of offers can be given to them. That is on time -- real time, we are checking with us with the bureau data, and we are ruling out preapproved offers for these customers. So that is one of the reasons why you have a very strong repeat franchise. And this is the in-build system calculator, which actually derives the income based on the trade activity. Now coming to disbursement, if you look at all of this, the penny-fuzzy checks validates the bank account of the customer to ensure that the funds reach the right customer. This is something that we have implemented last year, data-driven analytics to ensure that we give the right risk-based pricing to the right customer, wherein a repeat customer who has a lesser exposure gets the best interest rate. Customer onboarding, we're doing through e-Sign so that there is no paper which is signed by the customer and servicing of the loan through Planet App. So a rural woman in the rural area who can access her the entire loan details through the comfort of her mobile, see the loan outstanding, make digital payments and also get an SOA downloaded is something that we provide through Planet App. Now as a result of all of this, what we have been able to achieve is what is captured over here, a monthly disbursement increase of around 42%, a percentage of disbursement happening through straight pass-through, which is 25% of the total business that happens through this from sourcing to disbursement TAT reduction from 4 to 2.5 days and 100% paperless disbursement. This is on the disbursement and productivity side. While we look at the collection and the portfolio side, this is what you see. For collections monitoring, we have automated route maps, which is developed so that we can track the entire collections team who is doing the collections on the field and monitor that, this real-time receipting. For the Micro-LAP business, we have 100% mandate registration and account aggregator data being pulled to ensure that the customer is right for funding. In terms of digital collection, we have given the customer options -- multiple options to ensure that we do digital collections very methodically on the field. And portfolio tracking, this is extremely important because it when we enter a pin code, it's very important to know whether that pin code is something that where we would want to do business and whether the pin code has any trends which are pointing towards a lack of safety in operating in that pin code. So this is where we ensure that we do a lot of data analysis. Once the customer comes on board, it's important to monitor the customers' behavior in terms of increasing leverage, in terms of delinquency outside and while still being regular with us. All of these are very predictive methods by which we see -- we try to understand the customer much better before the customer shows a behavior in the book and helps us ensure that our portfolio monitoring mechanism is very robust. With all of this, what we have achieved is this 99.5 percentage collection efficiency as on September, which we have reached 99.6% in October, 96% 0 DPD book, 96% full group collection, 2.6% 90-plus DPD and 35% digital collections. So this is the impact that digital and tech has caused in our book. Now when you look at the journey forward on some of the things that we are working on as well as some of the projects which are taking shape are this. So the one on customer identification, real-time alerts, this is majorly for ensuring that our exclusive customer remains with us. And whenever there is a bureau inquiry from an exclusive customer, we are able to reach the customer in time to provide a loan. There is a propensity model, which works on -- works behind the entire system, which will tell us out of the entire repeat base of customers, which are the customers who are likely to take a loan and hence, help us improve our efficiency. We have built DIY journeys on PLANET app so that repeat customers can approach the company without waiting for a feed day to reach her. In terms of customer appraisal, we have built alternate data access as well to ensure that we know much more about the customer before giving her any exposure. In terms of collection, there has already been details about bot calling the customer to ensure that we inform the customer about the due date well in advance. On AI capabilities, we are working on a couple of things along with Debarag and the entire team. In terms of geo expansion, how do we overlap the entire socioeconomic patterns in the area, along with the bureau data to ensure whether we should open our presence in the right markets or not. We are also working on a score based on the customer lifestyle so that we know whether the customer whom we are lending is the right customer. This is also used to track group collections as well as provide a real-time help line to the sales team to resolve any queries. So this is on the tech and the data enablement in the business, which has helped business scale up. A quick look at the business updates. These are graphs that you would have seen as part of our market disclosures. So when the industry went through the largest crisis that we have seen, one of the most important things that was being tracked is the percentage of portfolio, which is having more than 3 lenders. So when -- it was when we started off and when this entire crisis started off and when this was being measured in the market, we were at 17 percentage. From 17 percentage, it has now moved to 4 percentage over the last 5 quarters. So now it's just 4 percentage of customers who have more than 3 associations, which is a very minimal portion of our portfolio at this point. So if you look at how it has fared compared to industry, we are here. Here I have marked Q1 FY '26 because I'm comparing industry with Q1 data. Q2 data is not yet available for the industry. So this is how the other major players pan out. So you can see the percentage with more than 3 lenders is around 7 to 11 percentage, whereas we are at 5 percentage, which shows that our portfolio is underleveraged. And if you look at collection performance, as was explained in the first presentation, our collections has remained steady, except for this blip where the Karnataka crisis happened. We have been able to maintain collection efficiencies well ahead of the market, both in 0 DPD as well as 0 to 90 DPD collections. And the result of this collection efficiency can be seen in the 0 DPD book and the 90 DPD book. This is indexed across industry. So you see this industry representation of 100%. Our 0 DPD book has been at 111 percentage, climbing up to 126. So it has been increasing steadily. And the 90-plus DPD book, when the industry is at 100, we were at almost at 1/3 of the industry. And today, we are at 12 percentage, which is much lower than what we used to be in the past. So this shows the portfolio quality as well as the resilience. Now if you look at the business, as has been explained in the first presentation, we have been able to maintain the disbursement momentum during the crisis and scale up quietly at the same time, working on the various digital as well as tech initiatives and understanding who are the right customers to lend, and we have been able to come back to the INR 2,000 [ crores ] trajectory of business. And if you look at this business, how has it enabled AUM growth? Our AUM has moved from INR 25,795 crores to INR 25,962 crores between Q1 to Q2 in the last 1 year. As I said, we did not degrow the book. We have been able to hold the book. But when you compare it with industry, the industry picture is very different. There has been serious degrowth in the industry. That's one place where we have been able to maintain our book as well as the disbursement trajectory. Now moving on to -- that's all about the JLG business. Moving on to Micro LAP business. When we entered this product, what we looked at is which are the more potential markets in terms of market presence, and we saw that there are southern markets which are pretty large. There are markets on the western side as well, which are pretty large, whereas the market was pretty minimal in the eastern part of the country. So our geo presence started off from Tamil Nadu. Right now, we are present in almost 175 points of presence in these geographies. Now we are at Stage 2, we are expanding these 5 geographies where there will be another 75-odd branches coming up. So by the end of the year, we would be having 250 points of presence. Now how has this business scaled up? From the start in Q3 FY '24, the book has built to INR 843 crores and is well positioned to touch INR 1,000 crores by end of this quarter. Quarterly business trend of INR 200-odd crores, which is around INR 70 crores of business per month is what we are clocking in right now. And the key portfolio metrics are captured here. We operate at a FOIR of 47% and LTV of 52%, regular collection efficiency of 99.9% and portfolio yield of 18% to 20%. So that's on the Micro LAP business. To conclude, the JLG portfolio that I explained gives us benefit in terms of strong rural penetration, strong customer connect and resilient performance and the Micro LAP business, which gives us a secured product to balance the unsecured mix on the JLG side, higher ticket size and lower credit costs. Together with both of these business, we are ready for the future growth. That's it, and I would like to invite Asheesh Goel, the Chief Executive of Farmer Finance, to come on stage.

Good evening. You guys have heard a lot regarding the interventions that we've taken over the past 1 year. Ramesh and Debarag took you through what's under the hood, what's the work that's gone behind getting that to fusion. I'll try and dimension out as to from a business standpoint, what is it that we were looking at or aiming to gain out of it and also try and quantify the same. However, before we do that, it's only prudent to look at how the year has been so far. And because the business is actually subject to the vagaries of the rural macroeconomics, let's look at the 4 vectors which really impact us. I mean, whether it is the rainfall, whether it is the kind of water reservoir levels that are there, the kind of harvest that comes into the mandis or the kind of sowing that's happened and all of them are green, all of them tick off. This is very different from what it was 2 years back wherein all were in the red. And the result was a year back, we saw the stress that we did in the market. Now this augurs well for us. And as a result, our AUM has already crossed INR 15,000 crores. We've serviced over 11 lakh customers and our active dealers stand at greater than 2,500. What you see over there is productivity, which has increased by 14% and the new tractor disbursement by 13%. Now this is one of the outcomes that we want from the interventions that are there, which is efficiency is going up without adding manpower, and that's something which is clearly evident here. The new business of WRF that we launched, that's moving in the right direction, and we saw a 55% growth, although at a small base, but it's again directionally going right. And also the product is working well because the MTM on WRF was sub 1%, of which also 40% got auto settled. But like we've been talking about risk-first, tech-first, let's look at a few things as to how the portfolio is doing. And we'll have a lead indicator and we'll have a lag indicator. So if you look at the lead indicator, that's the nonstarter. And if you look at it, this is indexed against March of '24. And from that, it's constantly on the way down, currently standing at about 35%, but from a peak down big time to about 35%. Then if you were to same time map it against the industry delinquency, and this is 12 MB 90-plus. The thing to be noted is that while there is already a difference, this 79% actually refers to this period. So as this period and the impact of that keeps on coming in, we are very confident that we will start to see a wider separation between our performance versus the industry. Let's look at a few more collection numbers. And what has this been driven by? So obviously, the e-NACH penetration stands at about 88% and the clearance at about 72%. Again, this number is relevant because 2 years back, this used to be a single-digit number. Again, the overall bankable pool has gone to 70 with a clearance of 66, meaning 2 out of 3 people are paying me through their bank account. The result is collection efficiencies, which have seen a clear uptick versus the last year performance and are up by about 100 basis points for the first half of the year. And also the on-due date collections and the 0 DPD book has moved up by 2%. Again, a thing to note that while my on-due date collection has gone up by 4%, my cash collection has come down by 4%. Now moving on, again, impact of Cyclops. So one of the things that we wanted to achieve -- so there were actually 3 things that we wanted to achieve. We wanted to have a differentiated offering. We wanted a better TAT, and we wanted a better portfolio. Now let's see how we fared against each one of them. So we actually launched Cyclops in April with about 16% of our dealers, 91% -- 90% got a better offering than the standard of what they offered for and 2% were the ones that got lesser than that. Our biggest concern was as we go 100%, will this hold good? September, we were at 100% and the offers continue to be better for 91%. The LTVs moved from 67% to 72% -- 70%, about a 3.5% upside. And the ticket size also went up despite the decrease that we saw in the cost of the asset due to the GST release. On the TAT front, we had 19% that was STP, 81% still had manual underwriting interventions, which has now come down to 64% and 36% moves through, leading to much better TAT. On the GNS side, if you look at the portfolio, 20% of the book was Cyclops underwritten and about 80% non-Cyclops. The GNS was 23% on non-Cyclops and 17% appears exception of 6%. And obviously, on the non-starter, a miniscule 0.2% against 1.9%. Although these are early signs and the book is yet to come to full maturity on the Cyclops written book, but it's moving in the right direction and showing us the kind of improvement that we expect it to bring to the table. The second thing, if we move on, from a customer, there is another customer which is very, very important for us and which is the dealer. And Ramesh, in his presentation, talked about the fact that we've launched something for them. Here, again, we wanted to give them a one-stop shop with a lot of transparency and control and allowing them the ease of doing business and not being dependent upon the frontline guy to be able to do that. As part of that, we've launched the partner PLANET app. Let's see exactly what that brings to the table. [Presentation]

So again, I don't want to go into the key features. They have already been discussed. The proof of the pudding is in the eating of it. What did we see? We saw a 25% increase in the TA disbursement with 12% more dealers taking trade advance from us. Let's look at a few collection numbers. And from a collection numbers as to what is the impact of the digital interventions that we've taken. Here, again, the objectives are very, very clear. We wanted to give the frontline a unified platform wherein all the details would be available to them for them to be able to do their job more effectively, to be able to track them on a real-time basis and also to have a litigation management processes, which allowed them to do the settlements at a better pace and at a better number. Three of the things that we started, we obviously gave them a BRAKE app, which was a single point contact for them to be able to get all the information; an activity monitoring app, which was there that we could -- the first level and the second level supervisor could track them; and the third was obviously a litigation management tool, wherein all the notices were automated and would go through and be tracked. And from there, it would move into the BRAKE app and be available to the front line once we met the customer for any discussions. Like I said, it was simplified, paperless journey, allowed for the resources to be optimized to the maximum. Here, again, if you look at it, we had a touch-free collections going up by 9%. The productivity increased by 5% and the settlement TAT came down by 2%. One more thing that we wanted to do was use analytics for our collection strategy. So a lot of times, we look at customers from a sales point of view, wherein we say we have a personalized offer to them. We try to take the same concept and say, what if you were to do it from a collection standpoint? So we wanted to actually have a personalized collection effort depending upon the customer profile and actually have the right channel, the right time going to the right customer. Two things were used into this. One was the customer risk score and one was a predictive EMI bounce score. Now both of this were put together to be able to come up with a decision engine, which would tell me as to where this would go. Obviously, what was used in this was repayment behavior, both on us, off us, loan parameters, crop details, demographics, customer banking details. And this allowed us to be able to come up with that predictive model. Ultimately, again, it's the result that brings us that this is something that's working well, 156 basis point betterment on collection efficiencies on the 0 DPD book and the bounce rates coming down by 9.44%. Now all I'd like to say is with all the lead indicators for the macroeconomics taking off well, all the interventions helping us take us into the right direction. It only makes us believe that the risk calibrated growth that we are looking at is just around the corner for us. Thank you. And I'd like to invite Jinesh to please come and talk about the urban secured landscape.

Good evening, everyone. I'm Jinesh, and I look after the urban secured space. Happy to speak to you all here today. What I'm going to do is take you through what we are doing on the home loans, 2-wheelers and the insurance side of the business and how that is actually changing the shape of L&T Finance's overall book. A quick snapshot in terms of where we are. This is what the book looks like today. On an overall basis, the portfolio is today heading at about INR 10,000-odd crores of disbursements that have happened in the first half of the year with overall disbursements and book leading to about INR 40,000-odd crores. This is broken up with mortgage and 2-wheelers in the ratio of 54% and 46% and disbursements in the ratio of 68% and 32%. Also from an insurance perspective, there are 4 broad products that we are offering to our customers today. They are basically linked across life, health, motor insurance and EMI protection. Today, 95% of our book has some of the other insurance that has been cross-sold to the customer. Also, if we look at the income per case, it has actually gone up by about 23%. So clearly, we are reaching out to customers using the database to actually sell insurance to customers. One of the new initiatives that we will be launching as part of our quarter 3 initiative this year will be actually to launch the retail cross-sell model with far more API integrations, better commercials that are there for our customers. Now where are we on the overall urban secured business? We have close to about 90 lakh customers, 87 lakh plus to be exact. That's part of the overall number of about 2.7 crore customers that are there in L&T. We currently lend in 160-plus markets. We have close to about 9,500 touch points in terms of sales for dealers, DSAs, et cetera. And we have close to about 6,000-odd employees who are working with us on the overall piece. All this is building up a large distribution across the secured lending business. I will now break this up into 2 parts on the home loans and the 2-wheeler side. While I start on the 2-wheeler side of the business, we have broadly about a decade plus of history in this business that we are running with. And one of the big advantages that we bring here is the 100% digital journey that is there. You heard Debarag and Ramesh speak about Cyclops. 100% of cases from January onwards this year have been only going through a credit decision engine. Nobody can overrule what is there as part of it. And that's, in fact, one of the biggest advantages that come out of this. We have close to 80 lakh customers who have been serviced on 2-wheelers on an overall basis within our business with 8,500 dealer touch points that we are actually servicing today. That kind of gives you the idea of the number of places that we can actually cater to customers. These 5,000-odd customers who are part of our 2-wheeler business are actually catering to all these touch points, leading on to some broad indicators of 28 months of tenure and a 1.1 lakh ticket size that is there today. Now what does this translate into from a Cyclops perspective as well? Our book today is close to about INR 13,800 crores. And of this INR 13,800 crores, INR 8,700 crores is actually sitting in a Cyclops-driven credit engine. That means close to about 2/3 of our book is today being underwritten through a digital process. Some of this book, like I said, has been prior to Jan '25, which obviously is today running off by itself. What is also important for us is if you look at our peak months, which are typically during the festive periods and when most people buy 2-wheelers, we see actually productivity per employee go up from about INR 18-odd lakhs to about INR 25 lakhs. That by itself is a significant number which has gone up. And this increase is despite the fact that we are getting a much larger prime share from our customers. Today, 87% of our customers are actually prime. That means their behavior on credit, behavior on payments, et cetera, is significantly better than what we would term as non-prime customers. Because of this kind of behavior and higher banking penetration, which I will discuss later, you will also see that our yield on this profile of customers who are relatively better has marginally fallen from 19% to about 18%. But this drop has actually resulted in a risk-adjusted return going up by 1.61% or 161 bps. Now clearly, the quality of the customers that is coming in through the door is significantly better and giving us a better return as part of our overall strategy. This is part of our journey, and Debarag kind of covered this in terms of his earlier presentation that was there and how we gradually increased our penetration on Cyclops on a test basis. We gradually started off with 4% to 5% of our dealer counters actually starting off with Cyclops. And by January '25, we moved on to 100%. Now what is really the outcome of this is what I also want to share with you. The outcome of this is our GNS today on an indexed basis, which last year was at about 100%, if I take November number at 100%, we are today sitting at a GNS of 57%. So we are close to about 43% of our peak number of what we saw in November. What is our bounce rate? If I was to take November again indexed, we are sitting at a 69% today. That means close to about 31% has actually improved the quality of the book that's coming in through the door. These 2 things are a large indicator of what we are approving and what the behavior of the customer is actually. Another important indicator for you to look at from a Cyclops perspective and where I spoke earlier about 63% of our customers are coming in with banking. That means the customer who is, let's say, walking into a Bikaner or into a Nashik or into an Aurangabad showroom is sharing with us his banking information for us to decide whether to lend to him or not 63% of them. That is an indicator of the shape of things to come. That means we are able to do a much better assessment of these customers on a forward-looking basis. When we do a better assessment, we are seeing that the bounce rates have actually fallen. We are looking that the GNS is much better and the collection efficiency is also much better because we have assessed these customers differently. What is also happening is from a partner perspective, we have much better acceptance of L&T at the dealer counter. That actually tells you that we are getting an 89% chance to actually review every case coming in through the door at a dealer counter where we are present first. We are getting a much higher approval rate from 55% moving to about 64%. That's close to about a 20% higher approval rate. Our LTVs that we are offering customers now who are coming to us with banking and a better profile is also much higher. That means close to about 80% average LTV we are offering to these customers who are coming to us with this information. And our GDP dealers, now what do we define as GDP dealers? These are our top-end dealers whom we term as gold, diamond or platinum dealers, GDP, unlike the GDP that you referred to as part of economics, these are our gold Diamond and Platinum dealers. We have 1,800 such dealers today with us who are doing their business. We actually wanted to test what is really the outcome of our selection of customers versus that of the market. We took help from TransUnion CIBIL out here to look at what is the 2-wheeler segmentation of their base and how are we performing against that. We identified 3 segments out there, low risk, medium risk, high risk. And typically, with TransUnion, they have these kind of bad rates. So low risk has a 1.5% and a high risk has greater than 12%. Now what is defined as a bad rate? Bad rate is basically 90-plus at 12 MOB with an 80% LGD. LGD is loss given default. If we were to compare our pre-Cyclops book to our post-Cyclops book today, we are clearly seeing that where we used to be at a 5% average bad rate, we have now moved to about a 3% average bad rate. So that's actually telling you that we've been able to shave off 40% of the bad rate that was coming in through the door from an earlier selection. And this is not vetted by us. This is also vetted by an independent external established credit bureau. That kind of tells you how things are really changing. While that is on the business side, on the collection side, what are we doing? We have now focused very differently on the collections front. We are reaching out to customers a lot earlier. We are using the help of digital channels to reach out to customers and make sure that the payments come in, in a much faster manner. Now you will ask me what is really the impact of all this? Well, the impact of all this is this. My X bucket, if I was to look at it in an index manner, has actually fallen down by close to 16% at a time when my book is actually growing. My buckets 1, 2 and 3 have all seen a downward trend compared to what we were in April. And in all these months, we have only increased our book size. So in a growing book size, I have lower delinquent book, which is actually telling you a good story about what is happening on the 2-wheeler side. I've met up with some of you during some of the reviews that have happened, and this was exactly the story that I had spoken to you about, and now we are seeing it in pure numbers over this period of time. What are our strategic levers that we are looking at in the next few months? I won't delve into all of them given the paucity of time, but one of them that we are trying to bring in is obviously our differential QR journey, where customers can actually just scan a QR, fill up an application, get approved instantly. Another one is where we are trying to do a lot of integration. Today, we have an integration with 2 large OEMs. Hero and Honda, as you know, control more than 50% of the market today. With 2 of them, we are having an integration, which has been done, and we have a live in September '25 to ensure that we are able to cater to them in a much better way. During the festive season, 2 other large manufacturers, and I don't need to tell you the names of them because there are really only 4 people who control 85% of the market, but we've also gone live during the festive season with them. One of the big things that has worked out for all these things is the kind of relationships that we have with our dealers and what we do. Ramesh spoke something about the dealer app that is there, the partner PLANET app. Asheesh just before me spoke about some of the capabilities that are there. We've enhanced some of these capabilities for our 2-wheeler dealers because the requirements of a 2-wheeler dealer are relatively further different from what is there. The next video will actually take you through this. [Presentation]

So that's the capability that's coming in as part of our PLANET 2.0. The verbiage that you actually saw from some of our dealers of Hero, Honda, et cetera, is actually what we are doing with them and what they have actually said as part of the overall delivery. Now just to explain to you where are we, finally, it is the financials in terms of where we are from, where the numbers are growing, how much has disbursements grown, how much has book grown and how much has collections improved. If you see out here, in line with the industry growth, our new application log-ins have increased by 5% and so have our disbursements gone up by 5%. Our ATS, because we are focusing on a better profile of customers coming in through the door has also gone up, which is the average ticket size, while our tenor largely remains the same. While this is on the new business generation front, this is how the collection side has actually performed. Our 0 DPD collection efficiency has gone up. Our self-cure, which is a very, very important factor, which decides our cost to collect has actually improved by 280%. Our 0 DPD bounce index has actually fallen down by about 21%. Our credit cost has actually fallen down by 47% and our collections cost by about 11%. So all these are indicators of a growing book, good profile of customers with much better collections efficiency that is there in some manner. That's in sum total, our 2-wheeler business. I'll move on to the mortgage side. Mortgage, once again, a decade-old business with us, close to about INR 28,000-odd crores of book. But what is important is that we are really lending to a pristine set of customers that are there. If we were to compare our delinquency with that of the market, we are significantly better. Our loan to value is also significantly lower, as you see out here, and the average ticket size is very nicely pegged at about INR 67 lakhs and INR 90 lakhs. We are not really in the low ticket today. So clearly, we are in a much better segment that is there. All this based on a strong footprint of close to about 60-plus markets that we are lending into, a better profile of customers and what we have done is we have expanded our suite. I will be covering this a little later to focus on high-yielding customers of mortgage who either take home decor or take our mortgage plus product or take our LAP product. All these things are actually going to increase our revenues by our focus out here. Needless to say, customer is king. We will continue to optimize as far as that is concerned. But one of the best things that we've been able to do on the mortgage side has been our entire revised digital process that is there. This entire digital process is 100% paperless in terms of where we are. So let me throw some light on that. This is where we are as far as our Neo2 platform that we have just launched. The entire process end-to-end is completely digital, right from the application form all the way to signing the final loan agreement that is there. And what we do is we do a triangulation of client data, whether it is to do with KYC, whether it is to do with income, whether it is to do with bureau, it is to do with banking. You saw the entire spectrum that is there. When Debarag spoke about what is the capability of actually encapsulating all that together as part of the copilot that he has launched for the SME business, that's exactly something that we are also going to be doing as part of our mortgage business to improve the efficiency even further. Today, we have an automated CAM, which is already happening, which is the credit approval memo. That's already improving efficiency. If we see today, 83% of our base is 730 plus on the bureau. Our TATs that we have are anywhere between 4 and 5 days. Our lowest TAT across both is actually about 1 day. And our underwriter productivity has gone up by about 22%. That's just an indicator that the investment into technology is actually helping us improve turnaround time, client experience and underwriter capability. So where are we on the financials? On the overall book, the book has grown up between last year and this year by about 26%. Disbursements are up 7%. I spoke about high-yielding products. We are at close to 6x growth on high-yielding products that we are working on. We are not doing the pure -- I would not say not doing, we are less focused on the pure pristine low-value home loan products that are there. We are trying to shift our focus towards high yielding. Our Net Promoter Score from a client perspective has actually shot up because clearly, we are getting digital data, we are able to quickly respond to customers and handle them in a much better way. Now if I look at collections efficiency as well, which are all these are collection parameters, our collections efficiency has improved marginally. And if we look at a 30-plus DPD has actually fallen, overall 30-plus and 0-plus DPD has also fallen. Credit costs gone down by about 46% and collections costs gone down by about 22%. Both these are indexed. That's just an indicator of the shape of things that even the mortgage book, despite its aggressive growth over the last couple of years is actually telling us a better story than before. Moving on to what is our strategy as we move on. This is broadly at a very 30,000 feet level, what is an ideal process that should be there for a mortgage perspective. These 5 green boxes that you see out here are today all digital. The part that we have not been able to digitize and which is what we are working on is how do I get the entire property details. Now Debarag spoke about trying to get photographs of customers, houses, do an assessment and take that forward. That's something that we will build on. Legal documents, which are basically the property papers that are there of the customer, how do I assess them to be accurate, and how do I get a valuation done in the best and the fastest possible manner. These are data integrations that we are working on to increase our overall market space for home loans. From a long-term perspective, these are the various areas of backward and forward integration that we actually want to do as part of our home loan business, where we will try and reach out to customer requirements that are prebuying as well as post buying, trying to make sure that the entire experience of buying a house is very, very different from just saying that this is a plain simple transaction to moving to a transaction plus service in some manner. So that's the end of my presentation. This is actually encapsulating the entire piece that is there on the home loan side. I would request Manish to please take forward the unsecured lending side of the business. Thank you.

Thank you, Jinesh. My name is Manish Gupta, and I lead the Unsecured Business for L&T Finance. Now we've talked -- we've heard Debarag, Sudipta and Ramesh kind of talk about multiple initiatives on the tech and the data side. And I think personal loan has actually been one of the biggest beneficiaries of that particular transformation that has happened in the last 1 year. So let me kind of just go back to about a year on what we promised and what we delivered as a part of our -- in the last 1 year. So we actually -- our whole idea is to actually embrace technology, embrace innovation to actually build a very strong quality portfolio. Now we talked about 2 main things last year. We talked about fostering mega partnerships. We started off with CRED last year. We launched Amazon as a part of our Digital Day as a part of our RAISE last year. And basically, we talked about a next-gen underwriting architecture. Now I'm very glad to say that from a 2% contribution of these mega partnerships, we are now moving to almost 43% contribution of these mega partnerships, which are with PhonePe, CRED, Google and Amazon, which are now contributing to a very large book that we are sourcing today. Now not just that, we are also -- we've also revamped our underwriting policy last year, and that has shown us amazing results that are there. Cyclops, I think Debarag did talk about that we are going live. It's under implementation, and we'll go live in the next 1 or 2 months. What has it given us? Now if I look at the salaried customer base at an overall book level of what we are sourcing, that number has actually moved from 37% to 57%. Now this includes the cross-sell portfolio. So in our new book, it's actually much, much higher. If I look at the customers who are 750-plus CIBIL, which are actually very, very good, that number has moved from 54% in FY '25 to about 69% in FY '26. And what is also interesting is that basically our customers who we are sourcing and who have a secured trade line. Now what do we mean by secured government trade line? The customers who have a home loan or a 2-wheeler before, and they've shown a certain behavior on that has actually increased from 21% to 37%. So we are getting quality customers across all -- across our book and not just from our large partnerships, but across all our partnerships that you see here. Now if I go to the actual numbers, September '24, we were actually disbursing about INR 476 crores of personal loans. Now that number has gone up 146% to about INR 1,170 crores. Now how did it go? We actually see that the orange line that you see here is the mega partnerships that we fostered that has grown from INR 12 crores contribution to now INR 500 crores contribution, which is actually even higher than what we actually disbursed in a month across all our channels. So that has given us some very good results of the growth that you see in the personal loan book, but the growth actually has to come with a very important factor, which is risk. Now if I look at what has it given us, it has given us a very important portfolio quality, which if you see the 4 big parameters that I'm talking about, bounce rates, our bounce rates have gone down from 14.1% to 8.2%, which is a decrease of 4.2 -- almost 42%; self-cure, which Jinesh also talked about, which actually gives about almost [ 1/30th or 1/50th ] cost as compared to a field collection has actually gone up by 119% from a 15% in September '24 to about 34%, and it's keeping on increasing. And our GNS and NNS, which is gross non-starter and the net nonstarter, which is the customers who are coming in the first time, are they bouncing or not? That number has gone down from 5.2% to 2.9% and 0.53% to 0.16%. Now how has that happened? On the bounce rate side, we are actually doing a lot of data. So pre-delinquency management. So we've built multiple models, which are effectively then -- through what kind of digital channels through that particular PDM model is what we are trying to kind of remind the customers that they have to keep money into their account. Self-cure, you've heard Debarag and Ramesh kind of talk about the multilingual bot that is there. So basically, we actually contact the customers in multiple languages through a bot, which actually gives us this number. And of course, on the NNS side, there is -- again, there are multiple models that we've actually deployed, which has given us these numbers that you see right now. If I go into what exactly did we did in detail. Now the first thing that we did, and that's how we had kind of launched all of these partnerships, which has given us great results is actually built a very strong tech layer. Now that tech layer effectively is our core systems, which are our LOS and our LMS. On top of that, we built an internal API gateway and an external API gateway, which is built on micro services. And that connects to both the lead generation partners, which are basically Google Pay, PhonePe, INDmoney, Amazon, CRED, et cetera, and our underwriting stack, which is through our -- which is our internal stack, which is there, which connects to external partners, which actually then gives us a completely digital underwriting process through multiple data points that are there. And of course, that then finally converts into once the customer is onboarded, then how do we do collections on these customers, which are, again, through some of these partners through which all of this is completely integrated. So that's the architecture that we've built, which effectively is robust and can help us scale up to whatever number that we want. Now in terms of details, distribution, we've already talked about large partnerships. We've talked about native journeys. Now for each of these partners that you actually saw, we've actually built native journeys, which effectively give the customer best-in-class experiences. Basically, even on our cross-sell side, which is almost about 25% of our portfolio today, we actually see data and technology, again, using multiple models to actually target which customers are actually wanting to take a loan and therefore, basically which are our best customers and then go after them rather than actually going after all the customers which are eligible for the loan. And of course, the micro services stack, which I've already talked about. In terms of risk and policy, we've got digital underwriting, use of alternate data, which will actually go into Cyclops as we move forward. And of course, with all our partners, we've actually done a data room exercise to ensure that we are taking the goodness of what the data that is there of these customers to actually build a great underwriting process and models that are there. Collections, GenAI bot, we've already talked about. We've talked about PDM management. And of course, the 2 things that I have not mentioned before, which is basically our allocation to which channel the customer who's bounced goes to is actually all technology and data-driven. We've got DIY journeys for settlement for customers who want to settle. So if the customer actually comes on the PLANET app, he can actually -- and if he is eligible for a settlement, he can actually go and settle as well. So these are various transformations that we've done as a part of H1 2026. As we move forward into the next half of this year, our objective is to create a scalable business, which is of quality, but also kind of create a great customer experience. And I think that's what we are focused on in the next half as well. So what are the things that we are doing? You've heard Ramesh talk about the multilingual bot for sale. You heard -- you saw the demo as well. We are doing an in-app agent. I'll actually talk about that, an agentic voice-based chat interface, which is there, a revamped consumer journey, which, again, you'll see in the next few days. Our journeys are actually better, but we are keeping on improving that. And of course, the loan offer pod, which again, Ramesh kind of talked about. Risk and policy, Project Cyclops and Nostradamus will get launched very, very soon. And of course, we are kind of doing preapproved black boxes so that we get the best customers, not just at the bottom of the funnel but also at the top of the funnel. In terms of collections, customers who finally go into a 90-plus, for example, there is a dedicated effort through agency maximization that we'll be doing a mandatory of banking. Of course, Nostradamus, you've seen that it actually requires a lot of banking information. She can monitor the customer, what he's doing, how is his bank account information is changing, et cetera, with consent, we'll actually be putting mandatory account aggregator and of course, automated contactability through customer consent. So all of those initiatives will actually help us build a very, very robust business as we move forward. Before we actually close, I just wanted to play a small video, which shows you a glimpse of what that in-app chat agent actually looks like for our existing customers. [Presentation]

Yes. So this is a very short glimpse of what exactly is coming. There's a demo out there. You can actually go and experience this in real time, which will give you a detailed output of exactly what is happening. So that's the end of my presentation. Thank you. I'll now hand it over to Abhishek for taking you through the SME business.

Thank you, Manish. Good afternoon, ladies and gentlemen. So in the next 10 to 12 minutes or so, I will take you through SME sector, our understanding and our customer segment where we play in, the portfolio performance over the last 4 years because we are just above a 4-year-old business per se. And then a couple of use cases around how exactly AI is being used in terms of enabling underwriting for the business. So with that as a backdrop, if we see the industry per se, it's sort of a bit of a recalibration or sort of adjustment mode as we speak, with inquiry volumes being muted and originations falling down. However, if we break it into further parts and try to understand that which sector is slowing down or which segment customer cohort is slowing down and which one is growing, then the segment which is around INR 1 crore to INR 10 crore aggregate exposure. And by aggregate exposure, I mean all sorts of exposure that the person has, that segment, which has grown in the last 2 years from around 41% to 43%. And this is the segment that we play because our average ticket size is around INR 29-odd lakhs. And the minimum ticket size is around INR 10-odd lakhs. So that's up to -- from INR 10 lakhs to INR 1 crores is the segment that we play in. And hence, this becomes a close proximate. Another defining factor about this segment is that this segment from a delinquency perspective also has been doing pretty well. And in last 2 years, the number has dipped from 2.3-odd percent to around 1.8%. So per se, this segment gives us confidence. And what helps us is that we continue to be focused in this particular segment and continue to bring features, which makes our product attractive for this particular segment. Now what exactly is our business model? How exactly do we play in this particular segment? So point one is around it's a digitally native business, which what I mean is that there is not a single wet signature or paper that we take in the overall journey itself. So that gives us ability to expand our locations, expand our area of operations, expand our channel partners exponentially in a very, very short time frame. So just to give an example, in March '25, we were operating around 110-odd locations. In next 2 months, we were able to double. And today, we are present in around 200-plus locations. So that was possible because the complete play is around digital and hence, it allows us to expand the geographical presence of our business. The second aspect is around strong underwriting standards. What we do via Cyclops is break the customers into 3 cohorts of customers, which is around premium, value and core and then try to tailor our offering. This is the customer segment that they belong to, and we'll talk about it a bit more in detail when we cover the Cyclops in detail. Last but not the least is we have been very, very focused on our collection efforts. We have been able to maintain our CD CE at 99.5% level over the last 4 years. And the intervention of AI has enabled the self-cure and call center piece to go up. So today, as we speak, around 45% of our cases gets covered by self-cure mode, which is through either nudges, which is sent by AI or through the intervention of call center. So to that an extent, the field flow is less, which has its own impact in terms of the overall cost of collections that the business has. Now with these 2 cohorts, if we map it against the industry, I would like to draw your attention to the right top table. The industry and indexed to INR 5 lakhs to INR 10 lakhs being the core delinquency, then the delinquency as we start moving up the ticket size starts declining with the sweet spot starting from INR 25 lakhs onwards. If we see the ATS of the industry, ATS of LTF is slightly higher than the industry. So we continue to be focused on greater than INR 25 lakh segments, and our ATS keeps growing quarter-on-quarter as seen through that particular table. Also, in the overall portfolio, we do not have anything which is NTC. So basically, we do not underwrite NTC. We do not underwrite thin file. So the customer cohort is completely around medium risk or high or low risk, that's the customer segment that we play in, which is seen both through the CMR ranking of the customers that we have as well as their own individual retail bureau that they have. That you see the portfolio is focused on prime, prime plus and super prime. As an outcome of that, what happens is that if industry standard is indexed at [ INR 100 lakhs, ] our standard book is at [ INR 109 lakhs, ] which then subsequently in the buckets, our numbers start becoming less. So that's the kind of play which has happened. And hence, what the outcome that we like to drive home is that because of the appropriate identification of customer segments that we play in, our overall retail -- overall performance of the portfolio is a bit more risk resilient, if we can say so. Coming to the business highlights. We are at around 6.23% market share. This pointer is not working. 6.23% market share. And what we intend is to increase this market share to around 7-odd percent. We continue to be focused on this sub INR 1 crore ticket segment by adding features so that our products becomes much more attractive to this particular segment. As a next stage, what we want to build is build on the partnership and direct sales team. Currently, in the overall volume, around 13% of the contribution comes from direct business. We would like to take it to around 25% level in a medium term, 12 to 18 months duration time frame. So that's the aim that we are chasing. In terms of collection strategy, we have a 2-pronged approach. The first approach is that understand our portfolio performance and wear out the segments that we do not want to entertain. So around that, various policy interventions keeps coming, which is at the top of the funnel that, that customer cohort should not get into the portfolio itself. However, if they have entered, then we also keep investing in our collection setup and bring better structural changes so that our collection efficiency remains strong as seen in data that our collection efficiency at the 0 DPD level has been in the range of 99.5% for the last 2 years. That's the approach that we have carried, and that's the outcome that we have. Now a bit of a deep dive into Cyclops on the -- with the lens that how exactly it gets used or what exactly use case it was solving for. So if we see SME underwriting, the first thing which happens is that there are various data points that needs to be seen to appropriately bucket the customer that how risky that customer is. Those data points would be related to his retail bureau, the commercial bureau, the banking, the GST and his income statement. And one by one, the credit manager used to go through all these things and then start understanding the riskiness of the customer. The second aspect was that while these data points were available, they were not available in one shot. And hence, a unified view of the customer was not seen at a point of time that used to be seen over a point of time, but not at the same point of time, resulting into higher turnaround time. So what Cyclops solves is that it takes all the data points and bucket the customer into premium, value and core offer and itself. So that allows us to have a better customer segmentation and enables underwriters to then have a more consistent, more accurate assessment of the customer. What it has already helped is in terms of productivity enhancement that their turnaround time has reduced by half as seen by the data and what we expect is that going forward, it will also lead us to have a better credit cost control. Now the month of November was a unique month in one aspect that the Cyclops went live on 12th of September. So in November, the first billing happened. And there were 2 kinds of customers whose billing happened in the month of November, one, which were underwritten prior to 12th of September, which we call it as a control group, if one may say so, and the portfolio, which was underwritten post 12th of September. Now once their due date happened on 3rd of November, the GNS performance of Cyclops underwritten portfolio is around 160 basis points lower than the non-Cyclops portfolio. So that's a good data point to have. Also, in a single month, the contribution of premium that if we back calculate our complete portfolio and see it via the lens of Cyclops, then our premium used to be is around 37% in the pre-Cyclops era. In a single month, it moved in terms of incremental sourcing by around 300 basis points. So that's another good thing because the likely PD rate of the premium segment is around 1/5 of the value segment. So that's the outcome that we expect coming out of Cyclops at a going-forward basis. Further to Cyclops, we talked about Project Helios as well, which was the underwriting Copilot. It has given good results. Demo is available. I would encourage all of you to see that. We have seen the outcome, very encouraged with that. So on this, we are further building on Phase 2 and Phase 3 of Helios, which is around other documents. So Phase 1 is primarily focused on bureau. What we want to build on is banking and GST as well on top of it. Also take AI to other aspects of documentation, which is around legal and technical checks as well as the universal partner API, which is -- which will help our partners to have a better lead sourcing and give a better customer experience. So to understand that what exactly were the challenges which this AI solved for, we have seen the video of it. So I will not spend much time. But primarily, the data compilation part is the thing that is solving, which allowed the overall TAT to come down because it is able to bucket the information into appropriate places, which is around overall credit summary, the derogatory information, which -- if available in that bureau report, if it is made there. And most importantly, it allows the underwriter or it nudges the underwriter to ask certain specific PD questions basis the deviations or variations that it has seen in the bureau track record of that particular customer. As a going-forward basis, we would continue to be completely focused on attracting the customer segments which are risk resilient, while having incremental features in our core loan offering, which is through various feature addition, tapping into the new markets, riding on to the existing distribution of either gold or farm or 2-wheeler, keep encouraging and having more and more partnerships, both on the UBL side as well as on the supply chain finance side. We already have onboarded 4 anchors and 2 partners which we have onboarded. So overall, just to sum it all, we'll continue to be focused on attracting a customer -- risk resilient customer segment in a digitally native play. With this, I would end the SME portion, and I would request Mr. Raju Dodti to take us through the journey of gold. Thank you so much.

Good evening, all. I think amongst all the products, this is a new kid on the block. World finance business is something which we just started and I would just take a few minutes to take you through why we acquired this business, how we integrated this business, what have we done so far and going forward, what we would do in terms of expansion of this business in our product suite. So when it comes to the business, we were looking at finding a new product. As you saw somewhere in 2021, the last business, which Abhishek just presented, SME was started. And it was felt necessary that we have a new business and that is where the high yield secured products in the form of gold finance was found a suitable products. It was also necessary from the following key aspects. As a whole in banks and NBFC have learned close to INR 13 lakhs in the gold finance market. And NBFC share straddles between 20% to 30% in that. We do have a physical presence in almost 2,200 locations in the country already. And we believe that getting into a gold finance business would be the right fit for us. And that is how we ended up scouting a boutique firm in Chandigarh, who is present in North and West, and we onboarded that particular business by way of a slump sale in somewhere June or 9th June exactly. We did the usual protocol of completing due diligence, both finance, technical and legal. And we also ensured that all the gold packets of that particular entity were physically verified to ensure that we get the best-in-class business transfer done on our side. And that is how we ended up beginning this business. On the day when we acquired, it was close to 700 employees, which came on our side, 130 branches, largely predominantly based in North and in West and 1 lakh customer with almost INR 1,300 crores of an AUM, which we acquired. In the period -- a short period that has gone by, we have ensured that we maintain the disbursement trajectory. In fact, we have improved on it. You can see from the graph, we have had the highest amount of disbursement in the month that has gone by. And we have seen close to INR 200 crores of net growth in the business. So what were the key principles of integration that we focused on? Ramesh in his presentation, gave a detailed backdrop as to how the technology integration was the key for this business. And it was important because we were keen to have a seamless onboarding of all the customer on our ecosystem without disturbing. And as you can see, when we closed the deal on 9th of June, the business when it got transferred, right from the time these branches became L&T Finance branches at 9:30 a.m., the customer walked in and they got, in fact, an enriched experience, and that happened because of seamless integration of IT systems, which had. We also had to give a proper orientation to all the employees which came to L&T Finance. And lastly, all the policies and processes of L&T Finance, we being an upper layer NBFC were required to be deployed onto this business. That too was seamlessly ensured. When it comes to the security aspects, we focused on physical as well as Infosec security. We already had one command center in Chandigarh, which was acquired at the time of business. We created a mirror command center in our Mumbai office. This is to ensure that we do have a redundancy as well as when it comes to the expansion of this business, and I will briefly speak about it in the next slide. We have parallel center in place in anticipation of that, that we ensured. And last but not the least, it was important for us to ensure that the Infosec security protocol for this business are at the same pedestal where we are currently operating on. And that too was ensured as part of the integration. So this is when it came to the security, we kind of did it. In 100 days of a plan, we also wanted to give the customer something new when they came to L&T Finance branches. So on the day when we started, the offices were made paperless. So during this implementation phase, we actually implemented the e-sign protocol for all the customers. And I'm very happy to state that the customers journey right from the time the customer walks in was given in the e-signing format. And we also used APIs, got the Indian Bullion and Jewelry Associations' gold rates live embedded into our system so that the LTV calculation and the loan quantification is done automatically. So these are a few of the things which we could manage during that implementation phase. So what are the technology themes, which we want to focus on? So everything that we spoke about, the progress of AI and digital, which we have inherently developed in L&T Finance over the last 12 to 18 months period, and that has been implemented, be it Cyclops or Nostradamus and all other digital tools, we want to have use case of all those being deployed for this business as well. Obviously, we would not be doing that in one shot to ensure that the business is not disrupted in the process. But these are the 5 key themes on which we will definitely be focusing on. I will not go through all these, but the one aspect which we want to genuinely focus on is the 12-plus security protocols, and that is important given the fact that we understand that gold is quite close to customers' hat is being given the jewelry. And we want to ensure that the faith with which customers would be reposing this gold in our custody, we would like to take care of that gold as much as that lady in her house would take. As far as the command center is concerned, it's bit of an archaic system, if I have to use that word. Nonetheless, it is a 2-factor security protocol, which is there. But Debarag, in his presentation spoke about having the video AI being deployed over here. So we have already experimented as to how the AI-based systems will be in a position to glance through videos. We have close to already many cameras in all of these branches and how the video feed would be analyzed with the help of AI as an additional layer of security that would come. This is also important from the point of view that as we will expand our branch network, we will not be forced or compelled to set up equivalent number of command centers across the country. We acquired close to 130 branches. On the left side of the screen, you can see geographies where these 130 branches are based at. And on the right side, you can see we have already finalized 200 branches across the country. The centers which we have finalized is an approach taken on the cluster-based approach. One, it must have the potential to grow the business. Two, we must have an existing L&T Finance presence. This will ensure that our ability to scale up is pretty faster. So these 200 branches is something which we intend to operationalize by end of this year itself. So in addition to 130 branches, and I'm very happy to state that the existing company from which we got this business, they took 9 years to set up 130 branches, but the power of the brand of L&T Finance has made us capable to inaugurate close to 200 branches in this financial year itself. So we have a confidence and we are actually striving to have an aspiration that from April of 2026 onwards in 2 years thereafter, we need to have a 10x growth of this business. And that is possible because in the first slide that we looked at, you must have seen that our existing microfinance customers have got close to INR 17,000 crores of gold loans, which they have availed from other financiers, and we want to make use of tapping those customers into system. Last, when it comes to the branch network, it is the stated policy of L&T Finance that this is the 11th product in our product suite. We want to have a branch setup, which we would like to call a Sampoorna branch, as the word denotes in complete sense. All products of L&T Finance would be offered from that physical place. We did not have that kind of a structure so far. We are moving towards there from the whole city of Ujjain. We have already inaugurated a branch. This is the branch wherein the gold loan as well as all other products would be catered to and service for our customer. Here is a quick video of what's there in that. [Presentation]

So as the culture of the country says gold shines, it is also considered as an auspicious metal. We believe that this particular metal and the business with which we are embarking on our new product journey should augur that much needed shine for L&T Finance armor as well. Thank you so much. And I will request Kavita, our CMO, to take us forward on the marketing initiatives.

Thank you, Raju, and good evening, everyone. This is the last presentation before the Q&A session and dinner. So please bear with me. Well, since we all know that Jasprit Bumrah is a brand ambassador for L&T Finance. So I'm going to begin with a message from him for all of you. [Presentation]

Well, this video looks so close to real. But I'd like to share with you that this entire video is generated using AI. This is one of the limitless uses of AI and technology in marketing that I'm going to cover during the course of my presentation today. Well, India is a cricketing nation, it's truly a passion for us and especially now with the Indian women also lifting the World Cup. So when it comes to bowling, Jasprit Bumrah, our brand ambassador, is known for his speed and style as well as he being a game changer every time he's on the field. Now these attributes resonate very well with our product categories. Hence, L&T Finance, Just Zoom 2-wheeler loan campaign is [Foreign Language] or our L&T Finance business loans [Foreign Language]. Zooming into our campaign, which is Just Zoom 2-wheeler loans, we were essentially the associate sponsors for the Asia Cup, which we incidentally also won. In addition to airing our spots on the India matches as well as the non-India matches, we also took up outdoor branding and branding on transit media and great presence on the digital and social media. However, we were able to drive a higher level of impact through technology interventions in our media execution. Let's see how. [Presentation]

Well, in addition to our TV spots, we also played our aston bands during the cricket match. What are these aston bands? They're essentially bands that come at the bottom of the panel of the TV screen. Now since Bumrah is on our creatives, we wanted to maximize the visibility by ensuring that these aston bands are in sync with every time Bumrah is on the screen. So that was enabled using a live stream technology through which either through recognition facial or jersey recognition, jersey number recognition of Bumrah that these aston bands were timed exactly when Bumrah was on screen. Here's how it looked. [Presentation]

Moving on. We also took up advertising in transit media. Now here, too, we wanted to make sure that our advertising is contextual to the time of the day where people are traveling to or from work. Now here, too, we leverage the technology. So on a real-time basis, audio ads were placed in the metro basis the time of the day. So here is what, let's say, a traveler returning back from work would hear. [Presentation]

So as you can see, this is perfectly timed because of technology. This is one of the ways we did it. Moving on, we also wanted to appeal to our target audience, which is in the younger age group and what better way to do it than to enable them to bowl like Bumrah. So we created an entire AI-generated initiative on digital, which is called Bowl like Bumrah. And what users had to do was simply upload their bowling action and get a score generated through the AI model, both on their speed and their style of bowling as well as a personal message from Jasprit Bumrah. [Presentation]

To experience this you should visit the Bowl like Bumrah in the demo zone, which is to my right. Moving on, of course, these are the results of the 2-wheeler campaign that we did. We had high levels of reach because of the Asia Cup. We had 40 million views, and we had very high levels of engagement. Moving on to the next campaign, which is the game changer campaign for business loans. Here, too, we wanted to celebrate with the business owners as game changers for their respective business. So we took up advertising through TV media on news channels and outdoor hoardings as well as digital video, and we generated high levels of engagement with a very close business community. How did we do that? Well, we created for the very first time an AI-led micro site, which enables these business owners to simply get onto the site and create business posters with Jasprit Bumrah along with a customized message for their line of business. All they had to do was to, of course, upload the picture and the AI model would create a business poster exclusively for them. And they could also place this in their social media or at their shops to try to promote their business. We didn't stop at that. On a real-time basis, these business owners also had the opportunity to see themselves on an outdoor digital hoarding that also helps to give them visibility. All of this was enabled through technology, of course. And this generated us great business results. So 11 million reach, 29 million views on digital platforms and high levels of engagement. Not just that, in terms of the consumer metrics too, our brand awareness moved from 12% to 31% and the consideration for L&T Finance business loans improved from 8% to 23%. This is carried out through research through Kantar to measure the consumer metrics. Moving on, what we saw right now were examples of using AI for awareness and engagement. Here's a great use case of using AI to drive conversions. So again, through AI, we create -- using Gen AI, we create topical and festive videos all around the year, which have high frequency. And here's what it looks like. [Presentation]

So who would believe this is all Gen AI generated? Not just this, Gen AI also helps us to create personal lodges in the customer journey each for at an individual level. Let's see how. [Presentation]

And of course, this helped us to drive higher click-through rates. So we improved that by 3.4x and were in the range of about 1% and also higher contribution to cross-sell business for personal loan. Moving on, not just for topical campaigns, but an entire thematic campaign can get done using Gen AI. So that's exactly what we did. For our Diwali campaign, an entire campaign generated through Gen AI, both from visuals to sound to graphics, all of that done using Gen AI. Let's have a look. [Presentation]

So this generated high views on Instagram and engagement. So here's wishing from all of us [Foreign Language]. Thank you. I now request Sudipta and Sachinn sir to please come on stage and address the Q&A.

Hello. Yes. Thank you, ladies and gentlemen, for a very patient hearing. We now move on to the Q&A session. That's the last part. I now invite questions from the audience please. [Operator Instructions] We have volunteers on both the sides of the hall who will assist you with mics. We request you to provide your name and organization before you ask the question, please. So Sudipta, sir and Sachinn, sir, please.

This is Ramesh Bhojwani from Mehta & Vakil. First and foremost, many heartiest congratulations on such a wonderful presentation encapsulating the each and every aspect of L&T Finance so beautifully, we have moved from physical to digital and used technology, particularly AI. Going forward, we will be doing a much faster, much smoother, but much stronger growth trajectory. The question rather than the thought was we have two master pieces communicated. One is Cyclops and other is Nostradamus. Nostradamus was a man who saw tomorrow in France in the year 1527. So would like you to expand on the Nostradamus as well as Cyclops? What I understand is it stands for cyclical loan operations. Would you like to comment or correct me?

Thank you. So these are like -- one of the things that is that when we build products, and I talked about the product mindset, we like to give names to the products that we build primarily because it makes it very personal for the development teams, and they tend to identify with. We also build mascots around this. Cyclops obviously is sort of derived from the mythological greek character with one large eye. The philosophy is that Cyclops never blinks. So basically, no credit risk can get past Cyclops. So that's the original Cyclops. It's the ever-seeing eye, which sees each and every small bit of credit risk. So that's the philosophy behind it. And as you rightly said, Nostradamus was a French philosopher and a clairvoyant who saw the future. So the thought behind Nostradamus is an automated portfolio management engine, which actually predicts the event before the event is supposed to happen, right, which helps the teams to monitor the portfolios more granularly and sort of get to the customer before the customer probably becomes unrecoverable, right? So that's the process. But it also has to do -- Nostradamus also has a very large cross-sell angle along with it. Nostradamus also identifies customers cross-sell potential much earlier than others do. And we are also piloting an undiscovered prime concept within the organization where a customer might not exhibit signs of a prime customer right now, but will be a prime customer, maybe 2 to 3 months, 3 years down the line depending upon his income growth. So Nostradamus helps us to identify those also. So together, these systems are supposed to operate in conjunction. Cyclops is obviously the origination layer, Nostradamus is a portfolio management layer. But Nostradamus in his ultimate avatar will be designed to send feedback to Cyclops, and that will happen automatically. Debarag covered it in his presentation. It will take us another 12 to 18 months to build, where the signals from Nostradamus will go and automatically tighten parameters in Cyclops to that extent that really the credit managers can go on holiday and sit on a beach while the system operates on its own, auto correcting itself.

My name is Nitin Naik, Director of Naik Consulting. And I would like to ask the question that the last division which you have, that is the gold finance division the gentleman said that when it was started, the very first day itself you could -- your customers would get the feeling of paperless office. So does it mean that while the negotiations was going on for the takeover, at the same time, you have in parallel having the systems put in place for a paperless office?

Not really. But the fact is that more or less when it became clear that we would probably be the successful bidder, that's when we started, right? And what we had done was that we had actually -- this part of the due diligence process, actually mapped the software that the company used. And we had also very clearly mapped what are the process improvements that we need to do. So it was very, very clearly mapped out as to what we need to do. And during the 3-month integration process that we had, we built everything.

Rahul Maheshwari here from Dolat Asset Management. First of all, great insights from the entire team. Just one question after implementing Project Cyclops and all the AI tools, how much on an overall basis, the cyclicality you expect has reduced from a business perspective? And if you can mention some sensitivity to it, that will be very helpful.

See, it's very early days, right? So Cyclops, for example, 2-wheelers is about 14, 15 months in operation. In the tractor business, it's about 6 to 8 months in operation. And in the SME business, it's only 2 months in operation. But whatever I can say, gleaning the learnings from the 14 months of 2-wheeler operations, you see 2-wheeler is a very, very aggressive credit product. It's a very, very difficult product to do, right? And if I were to give you some stats, average 2-wheeler portfolio industry bounce rates are anywhere between 20% to 22%. This is the average bounce rate. Our Cyclops underwritten 2-wheeler portfolio this month, the gross non-starter bounce rate was 7.15%. That's actually a prime personal loans bounce rate. But the fact is that we are not underwriting prime personal loans customer. We are underwriting 2-wheeler customer who are actually the sort of, I would say, near-prime customers, right? So obviously, the system does a very good job of finally dissecting customers who are actually have a high intent to pay rather than customers who have sort of limited sort of commitment towards sort of completing the tenure of the loan. So it actually separates out those customers. So I would say overall, on the long run, on a cyclicality, probably it will reduce by about 50% to 60% is what my gut feel is. But actually, we would need to run Cyclops continuously for 3 years, right, to have a real answer to that number and probably run it through a cycle to get the real answer out. But leaning on my experience of over 27, 28 years in the retail financial sector in India, I would say it would reduce cyclicality by 50%, 60%.

Can I just add. So you're looking at cyclicality for ultimate end result has to be a reduction in either collection cost or credit costs. And as Sudipta mentioned, the first impact of this will be felt in fourth quarter as far as 2-wheelers is concerned because there has to be some seasoning involved. And 2-wheeler was the one which was taken up the first. So you'll see in fourth quarter, the book, as you are seeing that almost 60% of the book has been -- has gone through Cyclops, underwritten through Cyclops. The balance 40% over a period of time will start running down fast, and you'll start seeing the impact from Q4. But FY '27 will be the year when you will start seeing significant part of the books have routed or being underwritten through Cyclops and accordingly, H2 of FY '27 will really be, I would say, the full impact of credit costs where Sudipta also mentioned in his presentation that what we're targeting is ultimately to move from about 2.75% to 3% down to about 2%. So directionally, you have seen that barring the micro loans, which we -- the challenge which we went through over the last about 4 quarters, I think now each and every business by end of December, even personal loans will actually start getting completely routed through Cyclops. So next financial year is when we will start seeing full impact of it on the credit cost.

I'll give a little more detailed answer. The fact is that this is something -- the efficacy of this tool is something which we are discovering with every passing month. Typically, we have not run Cyclops ever in the festive month. And typically, what happens in festive month, your volumes go up. And normal commonsense logic says, as your volumes go up, normally, a large number of bad customers might slip in, right? Because what happens is in a large manual process if you run a sort of credit administration system that is largely manual, as volumes go up, generally, trade discipline tends to fray on the edges. In the month of October, we processed about 65,000, 70,000 2-wheeler loans -- sorry, in the month of September. In the month of October, we processed, we actually disbursed about 130,000 2-wheeler loans. So actually almost a double size jump in volumes, right? So obviously, our feel was and my worry was that we will see the gross non-surplus spike up. It came in at 7.15%, the lowest ever. So we did high volumes and saw the lowest ever GNS, which means the machine did exactly what it was designed to do, cut away the bad guys and let the good guys in, right? So it is -- so this was actually aha moment for us, right? So as I said, it's a machine. We have tried to give as much disclosure as possible. The presentations were long. Probably some of you got bored during the presentations because it was too much of data or too much of -- but the fact is that the reason we wanted to give the disclosure to such detail was to give the investor community a good amount of understanding as to what we have built, how it is operating and how it is manifesting it in its business results.

My name is Parikshit, I'm from Pkeday Advisors. I have a 2-part question. If I understand correctly, Cyclops is fully automated underwriting system. And if that is the case, then why Helios even exists because that -- it's both targeting the SME loans. And if Cyclops is well on its way, then Helios would be redundant.

A very good question. The fact is that Cyclops fully automated is very good for loan ticket sizes to a certain part, where we have the confidence saying that we don't need a single underwriter to look at it. For example, 2-wheeler, average loan ticket size is INR 1.1 lakhs, INR 1.2 lakhs, right? So we are comfortable with machine taking the decision. But when the loan size is INR 25 lakhs, INR 30 lakhs, INR 50 lakhs, though the machine is giving you a decision, we want a secondary underwriting to be done. Because if you by chance go wrong, because you can understand that in machines, you can have bugs, right? When we build Cyclops, we figured out that there were a couple of bugs in the original stages, right? And they led to some anomalous results, right? So we underwrite the file through Cyclops. The Cyclops gives the first level of underwriting and then we want a human to look at it also and corroborate that underwriting because the ticket sizes are much larger. When you're underwriting a loan ticket size of INR 60 lakhs, INR 70 lakhs, INR 1 crore, still we would like our human to look at it, till we are confident enough after a history has built of about 12 to 18 months to say the human is not needed anymore, Cyclops can do it. So then Cyclops will do it, but we need to travel that distance. So that's why the Helios copilot is currently helping our underwriters to supplement the work that Cyclops is doing.

And if you don't mind, I'll just ask another question is a lot of the data that you showed shows a downward trend. But how have you controlled for the overall macro situation because even the macros have improved, right? So how should we look at your data from that angle?

So see, macros have improved in pockets. Macros again move in cycles, right? So one of the things which you saw in the Nostradamus tool, was that the Nostradamus also consumes macro scorecards right? So there is a set of macro scorecards which have been built for Nostradamus. And those scorecards are overlaid on a particular geography as well as a particular line of business. For example, if there is a flood in that one particular area or if there is a drought in the area, the macro scorecard will overlay on that and sort of predict the sort of forward-looking delinquency in that area. So in that way, we sort of try to address the macros through Nostradamus right now. As of now, for example, yes, business growth, there is a GST 2 impact, on tractors, there's good rain. So generally, that has a positive impact. So macros give you positive tailwinds, macros give you headwinds as well, right? So you cannot pan for macros, unfortunately. So that is why in part of Nostradamus, the tool is that at least -- as and when macro step on happening, at least, it gives you an advanced warning as to what might happen.

This is the continuation of the same question. If Helios can do it, then what is the difference between Cyclops and Helios?

Helios for manual underwriting simplifies your workflow, right? Like, for example, you will see in the demo. There is a bureau report which is 900 pages bureau report. Who's got the time to look at a 900-page bureau report? Helios translates a 900-page bureau report into actionable actions in exactly flat 10 seconds and also give you the PD questions, right? Similarly, Helios has got many parts. Helios had got a bureau analyzer. Helios got a banking statement analyzer, Helios will have a GST analyzer. Helios will have legal, technical as they talked about it. So all these modules have come in-built, right? So for example, Helios right now we have done for SME, but Helios will really be liberating for the mortgage generator because at times in mortgage underwriting, you have to look at many things, right? So average, it takes the industry anywhere between INR 18,000 to INR 20,000 to underwrite a single mortgage file. My estimate is that Helios in full flow, when we implement it for mortgage, will crash that number by about 1/10, right? So a file that takes INR 2,000 -- INR 20,000 to process should be done at INR 1,000.

Good evening, it's Kaitav Shah from Anand Rathi. So my compliments to you for the presentation. I think you've been batting on the front foot today versus last year. It's a commendable job. My question is you touched upon culture, I think a couple of times, we heard that. So how have you implemented this across the organization because that's a very difficult thing to do, right, change your culture?

See, culture doesn't change without executive sponsorship and executive demonstration. So the management committee knows that these are the couple of things that we need to do. And the fact is that the discussions with the management committee is very, very transparent. So each and every member of the management committee is tasked with that task of culture change, right? So basically, when the top end of the organization, speak in unition. And frankly, when you bring down -- and this is something which is probably appreciated in very large organizations, more than smaller organizations. But when you bring down the sort of the -- or when you bring up the outcome orientation of the organization, then the politics that might happen between intergroup or interdepartmental politics and it happens in every organization, it slows down the organizational velocity, right? It creates friction within the organization. Now when the management committee speaks at one unison, when the entire organization is focused on one key outcome goal and when you are trying to granularly distill this desirable behaviors in each and every person, right? Obviously, HR plays also a very, very strong role. So you use HR to force multiply your messages. But the fact is that HR cannot alone do it. It is the line managers, it is the support unit managers, it is the middle managers, right, who have to speak in unison. And that will only happen if the senior managers actually get out from their office and travel to the nook and corner of the city of the country. And that is what we have been trying to do. Our people have been literally been on the road almost like 2 to 3 days a week, right, our senior teams, right? And the branch sort of structure, introduction of the silo to matrix structure also has helped. So is it a job fully finished? I would say no. It is still a work in progress, and it will take a couple of more quarters for this to sort of come to a level of completion where we think that we have done a reasonable good job.

This is Namit Arora from IndGrowth Capital. Sudipta and team, compliments on the phenomenal progress over the last year, and thank you for a very detailed set of presentations. My question is company has made a lot of progress on various areas like technology, strategy execution. But sitting today, are there any things that you worry about as a company and as a management team?

We -- there are many things we worry about. And I'll be very, very candid about it. We are present in some cyclical businesses. We're present in some cyclical businesses. We are present in businesses where there are event risks. For example, J&J lending is a business that has event risk. We saw policy decisions in one particular state, suddenly tank collection efficiencies and our team struggling to keep a float. These are the things that we worry about, right? So one of our objectives is also to try to reduce cyclicality or exposure of cyclicality to our lines of business. So that is one of our major objectives. The second thing, obviously, is that we are -- we -- one thing which keeps us awake is the specter of losing our best people. Over the last couple of quarters, we have become a magnet for technology talent and because of some of the work that we have been doing. And trust me, what spreads very fast in the communities that this place you do -- get to do good work, right? And people have been coming and joining us. In fact, for that matter, this year, we have actually got very good people from all the top 5 IITs, and we really didn't struggle to hire them because we felt that all of them looked at the RAISE'24 videos and realized and decided that this is a place where work can happen, right? And so they decided to join us. And when we ask them why -- what pushed you, they said RAISE'24 videos on YouTube. That's what pushed us. So the other thing that keeps us awake is losing some of our best people. And obviously, again, geopolitical risk and sort of any other sort of risk we cannot measure or we cannot anticipate sort of worries us. Apart from that, I think we have a fair bit of control on how to underwrite, how to manage customers, how to manage credit risk. Yes, we have drag from our back book, which is underwritten by legacy algorithms. As and when that clear, the real cost of -- or the real impact of what we have been doing over the last 18 months will start appearing in the balance sheet and in the profitability numbers.

This is Himanshu from Aditya Birla Sun Life. Sir, just firstly, congratulations on the great Investor Day that you have hold and also the way you have delivered in the last 12 months. But can you put some perspective, I have a very basic question. Here, on the JLG when this entire fiasco started about the higher leverage and particularly exposed to most entities being exposed to greater than 3-plus lenders. At the start of the fiasco, we are also having somewhere between 15% to 20% of the book. But the outcomes the way you have delivered has been far better. Can you just help us by giving some what prudent measures in your underwriting, which ultimately results to a better outcome?

I think if you go back to some of our presentations, you will get that. But a couple of things we have been following always. If you look at the MFIN guidelines that came in June, July of this year, which said that greater than INR 2 lakhs threshold should not be breached, not more than 3 loans should be breached, et cetera, and all those things that came in, we have been following those guidelines since 2022 or 2021, right? So what MFIN introduced in 2025, we have been following those guidelines since 2021. The second thing that we did is that we realized early on that a crisis is coming. We realized sometime around October, November of 2023 that things are going to get worse. So what we started doing was that starting from Jan of 2024, we started off cutting off repeats and started taking our income thresholds up, especially for some of our repeats, right, and also tightening some of our through-the-door ingress of customers. The other thing that we did was that we beefed up our collections because one of the things we realized that our accounts per collector was at about 580 before the crisis. We say that we need to bring down -- because we had a sense that something was coming, we said that we need to bring down our accounts per collector to about 450 to 480 levels. So we actually pushed in 1,000 more people into collections before even the crisis hit to bring down our accounts per collector from 580 to 480 or 450-odd depending upon the markets, right? So -- and plus, at the end of the day, the JLG business is about brutal discipline, right? The fact is that unless you are able to maintain the discipline of the workforce and a large amount of workforce on the ground. And unless you digitize to the last and don't leave anything to interpretation, you will not be able to manage this. Our workforce is very, very disciplined. We have an extremely disciplined management team. We're an extremely hard-working management team. And our -- every process in the JLG business is digitized, right? Through all this and obviously, through proactive portfolio management, we have managed to weather the storm, right? And the fact is that our next step will be to make sure that we take out cyclicality as much as we can from this business, though we cannot plan for event risk. So obviously, we have said during our analyst calls that our objective will be to build back the macro prudential provisions over a period of time, and we will do that.

Sure. Sir, second is the project Cyclops and some of the rule engines which you have developed in the last -- basically sometime in the last 1 to 2 years, I think -- is it like -- because many of your team leaders and many of you have come from a larger bank where they are the early implementer of some of the rule engines, is that experience really helped you in setting some of the things here, setting the stage here?

No. I'll be very emphatic about this. Cyclops is a ground-up natively designed machine. What exists in Cyclops does not -- as far as my knowledge, does not exist in many large organizations. So the philosophy of building those scorecards, the philosophy of ensembling those scorecards, the philosophy of fast response from the scorecards using streaming data is something which has been probably built in the country for the first time. And to pull data from 7, 8 disparate sources, make sense of it, manage latency and deliver an experience to the person at the dealer point in an acceptable time frame is really the challenge of Cyclops. And what Ramesh said that Cyclops does not blink. I'll tell you this. Cyclops is a large machine now. In the last 15 months that Cyclops has operated, it has not had a single day of downtime. It probably has not had even 10 minutes of downtime. It is so robustly built, right? So I would say it doesn't exist in any large organizations.

I think we have room for just one question, if at all there is any?

You mentioned zero down time, so I was just wondering [indiscernible].

It is zero downtime, that's it. That does not qualify the question, but -- so maybe take one last question. One last one or two questions, if there are.

My name is Vipul Shah. I'm an individual investor. I just want to know what does it cost to build all these 3 platforms? And what will be the cost going forward every year?

Yes, the cost of building, I mean, the amount that we have spent till date on Cyclops and Nostradamus put together is about INR 60-odd crores, but it's still a work in progress. So Nostradamus is still in the making. We have just one particular business, which has been implemented. So this will increase further. But at this point of time, what we have actually capitalized is about INR 60-odd crores. Sorry, you had something more?

Next year [indiscernible].

Next year, see, out of this INR 60 crores, about INR 35 crores to INR 40 crores has been spent on Cyclops. The balance INR 20 crores has been to build Nostradamus. This will further -- so next year, we will have to figure out because it's a combination of time being spent by various engineers to actually build. As Sudipta was mentioning, it has been built in-house. So it will -- it's basically the time that we are spending on building this. So the cost of -- the employee cost, all these engineers are working around the clock. So it depends on the time frame required to actually go through that implementation is what we will...

[indiscernible].

Significant manner would have gone into this. So we just don't count that.

So it took flat 4 months to implement Cyclops, and it was done through the shift. So it was not that it was a 9 to 5 kind of thing. People worked in shifts to deliver it so fast. Otherwise, it would have taken at least about a year or so.

Probably have taken more.

Thank you, sirs, for the detailed discussion and clarification provided. With this, ladies and gentlemen, we come to the end of this Investor Digital Day. On behalf of L&T Finance, I thank you again for spending your valuable time. For any further clarifications, request to contact the Investor Relations team. Thank you very much. And please the tag that you're wearing, the access code that you have is valid for RAISE'25 tomorrow also, so please do visit us. There will be more stalls here coming by for RAISE'25, there'll be interesting speakers coming here for RAISE'25 tomorrow, which is our AI tech conference. So please be there tomorrow and attend it as well. With this, ladies and gentlemen, thank you very much.

I know it's late, but there are demos of Cyclops, Nostradamus as well as Helios and Bowl like Bumrah in that side. So in case any one of you have time, I'll urge you to spend some time there. It'll get a hands-on feel of what some of the solutions are. Our teams are waiting out on the demo area. The demo area is also available tomorrow as well for RAISE.

Thank you so much, and thank you for coming and attending our Digital Investor Day. On behalf of L&T Finance, I'd like to thank all of you for a patient hearing, and we hope to see you again next year.