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 his tech vision and execution roadmap at play in L&T Finance. This is also followed by a presentation on the digital architecture roadmap implemented by our Chief Digital Officer, Mr. Ramesh Ganesan. Following that, you'll have a presentation on the deep dive of Cyclops and Nostradamus, our proprietary AI-based underwriting and portfolio management engine, and other AI use 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. Ashish Goyal, 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 Loans. 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So, before we proceed, as a standard disclaimer, no unpublished price-sensitive information will be shared during the presentation and ensuing discussions.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documents will be referred to for discussions during this meet, while all efforts will be made to ensure that no unpublished price-sensitive information will be shared.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 has been quite blinding, with new models being introduced with every passing week. In fact, I don't know whether you caught the news, an Indian LLM, voice AI-based LLM, broke into number four rankings starting out of Bangalore, and it was published yesterday only. Indian startups are also catching up. That's a little about me. 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 RACE 24. Tomorrow, we have RACE 25, the second edition of our flagship AI summit. We have done a lot of things, but I'll focus on four 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 its 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 2024. Our overall disbursements, we did our highest monthly disbursements in October 2025 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 two-wheeler 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.
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 pharma finance, again, the tractor business also had a push from the GST 2.0 reforms and clocked a 21% growth rate. 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. Now it's a full-fledged machine working now on our two-wheeler business, on our tractor business, as well as our SME business.
It's a pretty large piece of software right now in version three. Our Chief AI and Data Officer will take you through all the details about it. It's about one million lines of code, 55+ algorithms, 75% data scientists and engineers working behind it, about latency on the various models between 50-700 milliseconds, one million loans underwritten. We have the performance reads for this. Debarag, in this 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. Right now, it's live in our two-wheeler business. Again, 13+ algorithms, about 200+ banking variables getting analyzed every month on an automated basis.
We have about 30+ engineers and data scientists supporting it. Obviously, this number will grow. 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 disbursements to our large partnerships to meaningful volumes. The last time we met, we did about 157 crores. This quarter, we did about 1,138 crores. You can see that the string of partnerships that we have added, and we are going to add a couple more in the next couple of months. This number is scaling at the rate of maybe about between 7%- 10% every month. The numbers are scaling quite well.
We are not only doing personal loans, but we are also doing two-wheeler loans as well as mortgages to 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 a 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 microloans 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 three 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 one single day.
So obviously, the thesis of feeding from the microfinance 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 2026, which means we'll be opening one 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 sub-market circumstances, disbursement growth of 39%. We saw this earlier. Our collection efficiencies in the 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 two 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 a day in L&T Finance looks like.
Welcome to a day at L&T Finance, where data and technology are helping deliver impact at scale. Over 75,000 loan applications are processed daily, with 10,000 new customers onboarded to our growing franchise. This engagement fuels growth, with over INR 300 crore disbursed daily. Our digital partnerships are scaling in a risk-calibrated manner, now disbursing INR 25 crore in a day. Our tech platforms are powering 7 million API calls across the ecosystem, enabling 200,000 transactions every day. Cyclops, our AI underwriting engine, processes 14,000 applications daily through 55-plus machine learning models with zero downtime. Through our Planet app, we service over 200,000 customers in a day. Our 24/7 systems remain vigilant, performing 30,000 risk checks and preventing over 32,000 cybersecurity threats throughout the day. Our AI collection bots handle 100,000 calls in a day, making outreach sharper and more effective.
At L&T Finance, every day is a step forward toward our goal of being a tech-first, risk-first organization.
Peter Drucker said that culture is strategy for breakfast. You can make strategies, you can make large presentations, but strategy execution on the ground does not happen unless you change the culture of the organization and focus the organization towards one common goal. 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, execution-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 onto 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.
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 head structure in the four zones actually has made sharper operations possible with faster time to market for all the decisions. And also, the focus has been on implementing a non-political and a non-hierarchical 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, almost three or four 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're made to understand the AI tools. They're 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 is 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.
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 ethos: 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 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 Shishil 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 Kanav 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 three 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 three 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. I am the lead developer of the underwriting Copilot projects. Our Copilot is an AI-powered assistant built on agentic architecture using MCP tools. It acts as a digital Copilot 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 lending bureau reports and voluminous banking data. With the underwriting Copilot, 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 Sir. 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 four to five 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, non-blocking in order to achieve a high throughput and concurrency.
Hi, my name is Viswadev Banerjee, and I head a two-wheeler business for L&T Finance. For the last one year, we have been implemented with Cyclops, and what we are seeing is an 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'm Rahul Rai, part of Nostradamus Development Team.
The Nostradamus is helping to build a proactive data-driven portfolio monitoring approach and generate early warning signals.
Hi, hello, 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 insights 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 transformation.
My name is Mayur Devendra Jain. I am the CEO and managing partner of Palladium Automotive, which is a dealership for Royal Enfield . I'm very happy to say this. They have initiated a digital AI platform called Cyclops. It's a credit engine, basically, which helps and engages the customers to get faster loans. It is a fully digital process. It has helped the dealership as well, as we have scaled up three 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 on both 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 the next two, three 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 creditworthy users, and we had the good fortune of knowing Sudipta much before he joined L&T. And through his previous assignment, he had a 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, creditworthy 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 blitzscale 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 dealer 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 have gone up. So they're 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 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 frontline, the adoption is very, very fast. And a complex tool like Cyclops' adoption in the sales channel actually went like a knife through butter. The multi-product Sampoorna branches will leverage the local cross-sell potential. The process has already started.
Last but not the least, initiatives in talent management to build a caring organization have resulted in an annualized attrition drop of 5% and a 30% increase in the 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 a 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 costs 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 way, 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. India is the brightest spot in the world, 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.
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. 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. We are very, very confident that we'll partake in that. 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.
Last but not the least, our technology platforms, our modular NeoStack, which is in-house proprietary built, yields very, very robust performance, which Ramesh, our Chief Digital Officer, will cover during his presentation. What are our imperatives for the next 12 to 18 months? We will announce our Lakshya 31 goals in quarter one FY27. We will drive a 20%-25% risk-calibrated AUM growth. We will achieve an ROA in the corridor of 2.8%-3% by quarter four FY27. 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 will 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 one 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, multi-product retail financier of choice. And I think we are well on that journey to become that. Tomorrow, we have RACE 2025, 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 RACE 2025 sessions tomorrow.
We have two 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 tomorrow. So we have talked about Nostradamus, but the fact is that we have not developed Nostradamus for the world, so maybe let's have a quick look at what Nostradamus can do, but that's not all. Let's give you a look as to how the Nostradamus screens look like. Nostradamus 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. In fact, we have used it for the last one month. Our SME underwriters are using it.
A traditional SME file that takes four to five 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 Ganesan, our Chief Digital Officer, to take you through the next session on our technology initiatives for the last 12 months. Over to you, Ram.
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 the pronoun to AI.
While last year it was about generative AI, this year that's morphed into agentic AI, and within agentic AI, there are two areas that we are watching quite closely. The first is something called 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're 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 1.0 Vision 1.0. In 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 credit 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 two, on top of the portfolio and acquisition layer, as well as the zero 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 four launches that we want 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 the technology team, the challenge was to get the entire integration and the day zero setup done in a matter of nine weeks. Twelve weeks was the entire integration effort. The technology team had nine 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 eight weeks, nine weeks, sorry.
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+ cameras across the 130 branches, growing at the rate of one branch a day, 24/7 monitoring in place with centrally controlled two-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 benchmarks 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 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 two quarters, from three to 10, which is fairly record time with an increased disbursement of two and a half x.
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 pre-qualified leads.
This is in the personal loans business. We rolled it out last quarter. What it also does is, apart from regular calling, we're able to provide real-time dashboards, quick lead filtering, and do this in parallel. The initial results are quite positive, and users have reported 80% improvement in TAT in lead engagement. Let me show you a brief clip of what's possible today. नमस्ते, मैं तन्वी बात कर रही हूँ L&T Finance की तरफ से। क्या मैं नमन जी से बात कर रही हूँ?
Haan ji. Dhanyavaad Naman ji. Ji, actually ye call aapke liye ek special pre-approved personal loan offer ke baare mein hai, jo aapko aapke acche repayment record ki wajah se mila hai. Hum aapko INR 300,000 ka loan de sakte hain, interest rate rahega sirf 13% aur repayment ka time 36 mahine tak aaram se. Nahi, mere ko INR 300,000 nahi, INR 500,000 ka loan chahiye. Main samajh rahi hoon Naman ji, INR 500,000 ka loan bhi sambhav ho sakta hai. Hamari relationship manager is baare mein aapko aur bhi options bata sakti hai aur aapki aavashyaktaon ke anusaar solutions de sakti hai. Kya main aapko relationship manager se connect karne ke liye ek samay schedule kar doon? Theek hai, aaj shaam ko 5:00 P.M. call karna na. Bilkul, Naman ji, aaj shaam ko 5:00 P.M. aapke liye relationship manager call karenge. Kya aapke paas aur koi sawaal hai?
Nahin, theek hai, thank you. 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 are 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 the 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 a fairly scientific approach to that. We have an analytics stack from which we have a lot of data that's internal to our databases, pushed 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 lifecycle 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 an 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 to 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 login process, and we run a lot of event-driven checks, which includes face match, liveness, Digi Locker, V-KYC, and 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 in real 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 dedups, 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 in 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 for a simple two-wheeler customer, we feed it through a bunch of models. I'm representing that quite simply here for the 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.
No-go might be because of either incomplete data or missing data. Doesn't matter. We built Cyclops to handle all of that. Fed through a few more scorecards, a few more models. 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 APIs happening in the span of three seconds, less than three seconds. Thick file customer. As I've said before, right? All of this complex orchestration is about 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 Sudipta 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 are determined by this automated collections engine.
Finally, you know, 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 make collections smarter, resulting in a superior portfolio performance. Moving on to our flagship product, G2C 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 the 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-duty collections, as well as 35+ charge- off 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 two 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 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 conversational voice agent built into Planet.
I'm going to stop conversing now and let the agent do the talking.
Hello, I'm Monica from L&T Finance. How can I help you today, Aman Singh? Monica, tell me the number of installments that are left on my two-wheeler loan. You have a total of 24 installments for your two-wheeler loan, and you've paid 19 so far. This means you have five installments left. अच्छा, मेरे को ये बताओ कि कितना आउटस्टैंडिंग प्रिंसिपल कितना बचा हुआ है अभी? आपके two-wheeler loan का outstanding principal amount INR 21,767.52 है। क्या आपको और कुछ जानकारी चाहिए? No, Monica, thank you.
Hopefully, you noticed its amazing ability at interpreting multilingual commands from the user, emotionally aware response, and how it manages interruptions very well. A lot of human-like features, you know, 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, Planet was about customer centricity, but customer centricity also means we look at our customers from the point of different touchpoints 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 sent 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 two-wheeler and tractor farm dealers with amazing capabilities that are built and taken for granted as part of Planet. It comes with their 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 Sujitha referred to earlier as well, and engineering resilience. Starting with agentic AI, I've walked you already through examples of how we've used agentic AI to influence customer experience. We've also looked at how agentic AI, you know, 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 lifecycle is a fairly standard lifecycle, 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 two 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 lifecycle, 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 toolsets 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. A product mindset, like Sujitha 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 it to the next level over the coming year. The final topic in my 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. 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. To that end, we have a four-pronged strategy. The first is about observability. Building real-time observability into our production systems has 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 the slides earlier, given the complex ecosystem, is the need of the hour. Finally, we want to wrap it up with the approach to something what we call as the always-on architecture, where our system reliability, we want to take it up from 99%- 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 this capability, 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, two and a half times the throughput, increased API calls by the same 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 Sujitha 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 seven months, we have a 3x increase in volume, with peak volume days going anywhere from 5-10x 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, reengineering 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 connection 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, you know, because numbers are beautiful, and I don't mind. So, speaking of 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 eating into it, and what you have left is your ROA. Now, we deployed Cyclops.
You've heard all about it, and we'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 behind Nostradamus 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, you know, 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 two-wheeler portfolio. The first version of Cyclops was about bringing in trust signals from banking 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're using, things like more indexes coming from credit bureaus, and we have added beyond the original 50 scorecards related to dealer grades and fast flow frauds. Combination of all this lets us characterize any particular applicant into one of five 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. 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 a Shikhandi or a Mule is that a nagging problem in the two-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 applicant's 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 backtesting, 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 four and five that we talked about, to a level of 66%. So, stay tuned for that next generation. Now, we can talk about models till the, you know, till cows come 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, you know, 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 backend that Ramesh talked about, as well as the frontend 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 analysis team, and we had data teams, you know, 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 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 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, you go with everybody. And that's the reason why we do these RACE 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, you know, actually this was a loan to the first customer, that was about a period of eight 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 compare 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 were convinced that yes, this truly is dropping our credit costs 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 two and a half 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 NNS, which is the percentage of people who have missed their first payment, that number before Cyclops to where it is after version two, we are about at a level that is about two-thirds, actually less than two-thirds of where it started from, and it keeps continuing to go down. Other way of looking at it is the vintage charts, and there, there's a very interesting story here.
When we built Cyclops, we backtested it with data from more than 12 months ago loans that were disbursed more than 12 months ago. When we launched Cyclops, like I said, that we had a Cyclops and non-Cyclops experiment going in parallel. Now, the deep blue line is the vintage chart of how two-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 is the reason why the vintage chart went up. Then we made subsequent versions of Cyclops fine-tuned to adjust to the reality and also add more scorecards and so on and so forth.
So where we are today, our 30+ DPD rates are just about one-third of where the non-Cyclops rates would have been. So now, if I assume that 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 Nostradamus actions that I will talk about later, but even if I assume that, this leads to a logical a priori conclusion that yes, everything being as is, we would end up with one-third the credit cost that we would otherwise. Now, Cyclops was not a one-time success in only two-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.
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 geointelligence 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. Finally, unlike two-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 we 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 two-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. The two-wheeler portfolio went Cyclops from December, January onwards. Since then, the disbursement has actually not fallen. Profit as per deal has increased, but the number of deals themselves have increased. 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, you know, multifaceted checks underneath.
Our PL portfolio is due to go Cyclops very soon, followed by HL, followed at some point of time by the RBS loans. Now, every loan looks one way before you disbursed it. After disbursal, the customer shows his real colors. How do you differentiate between a customer, let us say a college kid who just bought a bike when he's in college and looks just like, you know, 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, you know, does well in life, does many purchases with his PL and credit card and so on and so forth, and then finally buys a 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 tottering on the brink and a little bit of crosswinds from one way or the other during the, as the macroeconomy goes up or down, can throw him off the board and he ends up going into default. And how do we do that before that 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?
We do this not only based on credit bureau data, which is a little bit a posteriori, 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 macroeconomy 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. 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 which enable 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+ DPD moved within a space of 12 months for the two-wheeler portfolio. Now, here you see that that number, the 30+ DPD, has actually improved by more than 220 basis points. This, by the way, if you paid attention, a year ago, there was very little Cyclops, right? Now there is more. So it's kind of an indication, yet another way of indication that, yes, Cyclops is actually making the portfolio better. But what is more interesting is that the list 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 Pandharpur, and you notice that although the 30+ 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 characteristics compared to the platinum dealer, which leads to a potential, you know, 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 tendencies?" And it can finally summarize all this for you and say, "Hey, here's a snapshot. This is the slice and dice 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 co-pilot. Now, how will all this come together, Cyclops and Nostradamus? So Cyclops, like I mentioned, it starts from data and ends up with an onboarding decision, yes or no and how much. That loan gets disbursed many times over, creates a portfolio, which is looked at every month by this Nostradamus engine through this multidimensional scale.
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. 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.
You can say that, "Okay, looks like, for example, this particular LTV category in these 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 Nostradamus that gives you that forward-looking score. 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 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 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 it's as used in the Nostradamus in the Copilot launch how we have been telling you about the AI Copilot for underwriting SME loans. I'll just summarize it. Essentially, you have tons of data from credit bureau records, 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 Helios and is presented in a very easy-to-consume form, which saves hours and hours of the underwriter's time. 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, just heard, you have a new gold loan business. I have some gold. How can I get a loan? Is it going to save?" 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 over there. But very soon, this will also come to a lot of other channels to the customer, like WhatsApp and other messaging facilities. Not only are these chatbots available in English, this is live today in 11 Indian languages. And it's fluent in all of them.
And not only does Kai talk to you, chat with you as text, but when a customer is delinquent on their loan and it's time for them to get a call from a collection agent, Kai, the agentic chatbot, the machine can actually be that collection agent. नमस्ते, मैं Kai L&T Finance से आपके लोन अकाउंट के बारे में कॉल कर रहा हूँ। क्या मैं धर्मेंद्र कलिता जी से बात कर रही हूँ? This is a real PL customer who missed one of their payments. बहुत अच्छा। कन्फर्म करने के लिए धन्यवाद। मैं जल्दी से कॉल का कारण बता देता हूँ। आपकी INR 5,760 की EMI, जो कि को ड्यू थी, अभी भी पेंडिंग है। एंड आप आज ये पेमेंट कर सकते हैं। The exact status of the कल कर सकते हैं। And can not. Customer to get a promise to pay.
दिन से बाउंस चार्ज लग सकते हैं और आपका क्रेडिट स्कोर प्रभावित हो सकता है। क्या आप आज ही पेमेंट करने की कोशिश कर सकते हैं? And once the customer says, "Yes, I'm ready to pay," can also send a payment link. लेकिन क्या आप बता सकते हैं कि आप कल किस तरह से पेमेंट कर सकते हैं? Payment itself. एक-दो बजे। We turned it on. तो आप कल 1 बजे पेमेंट करने का वादा कर रहे हैं। And so far, about 50% of the time, Kai is effective in getting those payments. And it does it in 11 Indian languages.
Namaskar. 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 Micro LAP customer, a Micro LAP lead, who being in the rural area is tens of kilometers away from where our branch is. And so 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 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 MLAPs can come down. Another end of the loan cycle is settlements, right? So when a farm tractor has been repossessed because somebody could not pay the loan back, then one of the big decisive decision points of the value is whether the tires on the tractor are good, bad, or ugly.
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. 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. 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 the inventory flows out quickly enough. Our gold loan business has started. There are 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-ins 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 employee's 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 do not happen.
Finally, we all talk about AI, but recently there was an MIT report that came out that 95% of AI deployments failed. 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.
Because we started from data first, that is one of the reasons why our models simply work. Thank you very much. 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, over-leveraging of customers, and there were large questions about whether this industry will survive this crisis and how long is this going to last. 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 one year after the crisis started, let's have a look at how our portfolio has weathered during this time and how have we been able to keep our portfolio resilient and manage collections in a superior manner. Where are we today?
Right now, we have around INR 26,000 crores of books spread over 61 million customers. We have a pan-India rural presence across 17 states. The portfolio has exhibited extremely superior collection performance as well as quality. After having built this portfolio, when we realized that we are very deeply integrated in building sustainable rural livelihoods in women customers, two 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. 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.
That's how we started this business, and we are building on this business. Let me take you through details of the JLG business before we come to this product. If you look at the entire presence, as I said, we are present in 17 states, around 350-odd districts, 2,000,000-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-INR 3,000 crores, and few markets where still the book is below INR 1,000 crores.
This actually provides us with the geostrategy, 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 eight years, from 7,500- 26,000, which is a three and a half times growth in the last seven and a half years, is what we have achieved. 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 two, three 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, put in 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 one year, one and a half years of credit cycle looming. We have invested a lot in data and technology, as Deepak 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 five markets where our market share is more than 7%, around the same number of markets in the 5%-7%, and some markets in the less than 5%. Now, this offers a very multi-point strategy wherein our market share is on the higher side. Again, there is a very calibrated growth in these markets which are 5%-7%. We deepen our presence, and in the less than 5% market share are the geographies which we plan to expand. So together, if you look at the market share, it has moved from around 5% odd percentage three years back to around 7.6%.
This is a time when we had invested a lot in our processes and which actually scaled up our business in the last two years. Now, with this, we are well poised for a very calibrated growth in the geographies with deep presence, added other geographies where we have a lesser presence. Now, if you look at how this entire last two years, what all we have been building on has been explained in the previous presentations, through the lifecycle of the loan which comes in our system through the different checks that happen, ending in disbursement, data and technology are strongly embedded.
So 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 happens in 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 are any dummy applications coming in. So there are a lot of risk checks which happen over here.
What passes through the credit rule engine runs in real time to check the exposure association as well as the DPD norms. Within a minute, the credit sanction is given. 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 this with the bureau data, and we are rolling out pre-approved offers for these customers. That is one of the reasons why we have a very strong repeat franchise.
This is the inbuilt 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 drop checks validate 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.
A rural woman in the rural area who can access her 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. 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, that reduction from four to two and a half days, and 100% paperless disbursement. This is on the disbursement and productivity side. When we look at the collection and the portfolio side, this is what you see.
For collections monitoring, we have automated route maps which are developed so that we can track the entire collections team who is doing the collections on the field and monitor that. There's 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 multiple options to ensure that we do digital collections very methodically on the field. Portfolio tracking, this is extremely important because when we enter a PIN code, it's very important to know whether that PIN code is something 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. 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 customer's 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 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% collection efficiency as of September, which we have reached 99.6% in October, 96% zero DPD book, 96% full group collection, 2.6% 90+ DPD, and 35% digital collections. This is the impact that digital and tech has caused in our book.
Now, when you look at the journey forward and on some of the things that we are working on, as well as some of the projects which are taking shape, are these. So the one-on-one customer identification, real-time alerts, this is majorly for ensuring that our exclusive customer remains with us. 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 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 so that repeat customers can approach the company without waiting for a field agent 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 Deepak 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 to whom we are lending is the right customer.
This is also used to track group collections as well as provide a real-time helpline 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 was having more than three 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%. From 17%, it has now moved to 4% over the last five quarters.
Now it's just 4% of customers who have more than three associations, which is a very minimal portion of our portfolio at this point. If you look at how it has fared compared to industry, we are here. Here I have marked Q1 FY26 because I'm comparing industry with Q1 data. Q2 data is not yet available for the industry. This is how the other major players pan out. You can see the percentage with more than three lenders is around 7%-11%, whereas we are at 5%, which shows that our portfolio is under leverage. If you look at collection performance, as was explained in the first presentation, our collections have 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 zero DPD as well as 0- 90 DPD collections, and the result of this collection efficiency can be seen in the zero DPD book and the 90 DPD books. This is indexed across industry, so you see this industry representation of 100%. Our zero DPD book has been at 111%, climbing up to 126%, which has been increasing steadily, and the 90+ DPD book, when industry is at 100, we were at almost one-third of the industry, and today, we are at 12%, 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, a quick 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 2000 trajectory of business. And if you look at this business, how has it enabled AUM growth? Our AUM has moved from 25,795- 25,962 between Q1- Q2 in the last one 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. 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 two. We are expanding these five 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 FY24, the book has built to INR 843 crores and is well-positioned to test INR 1,000 crores by the end of this quarter. A quarterly business trend of 200-odd crores, which is around INR 770 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%, an LTV of 52%, regular collection efficiency of 99.9%, and portfolio yield of 18%-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 cost. Together with both of these businesses, we are ready for the future growth.
That's it, and I would like to invite Ashish Goyal, the Chief Executive of Farm Finance, to come on stage.
Good evening. You guys have heard a lot regarding the interventions that we've taken over the past one year. Ramesh and Deepak 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 four 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 happening, all of them are green. All of them tick off. This is very different from what it was two 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 efficiencies 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's 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 non-starter, and if you look at it, this is indexed against March of 2024.
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 MOB 90+, 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're 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 eNACH penetration stands at about 88%, and the clearance at about 72%. Again, this number is relevant because two 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 two out of three 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 zero 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 three things that we wanted to achieve. We wanted to have a differentiated offering. We wanted a better ATS, 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% 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 continued to be better for 91%. The LTVs moved from 67%- 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 tag 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 tag.
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% across the exception of 6%. And obviously, on the non-starter, a minuscule 0.2% against a 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 expected 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 launched the Partner Planet app. Let's see exactly what that brings to the table. So again, I want to go into the key features they've 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 retail disbursement with 10% more dealers taking trade advance from us. Let's look at a few collection numbers and from 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 break 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.
From there, it would move into the Breeze app and be available to the frontline once he 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 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 tried 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 a customer risk score, and one was a predictive EMI bounce score. Now, both of these 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, is the result that brings us that this is something that's working well, a 156 basis point betterment on collection efficiencies on the zero DPD book and the bounce rate 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 am going to do is take you through what we are doing on the home loans, two-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 two-wheelers in the ratio of 54 and 46, and disbursements in the ratio of 68 and 32%. Also, from an insurance perspective, there are four 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 you 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 three 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, etc., 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 two parts on the home loans and the two-wheeler side. While I start on the two-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 two-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 two-wheeler business are actually catering to all these touch points, leading on to some broad indicators of 28 months of tenure and a INR 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 these INR 13,800 crores, INR 8,700 crores is actually sitting in a Cyclops-driven credit engine.
That means close to about two-thirds 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 we look at our peak months, which are typically during the festive periods and when most people buy two-wheelers, we see actually productivity per employee go up from about 18 odd lakhs to about 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, on behavior on payments, etc., 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 basis points. 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%- 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% off 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 two 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's, 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 collections 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. 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 refer 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 two-wheeler segmentation of their base and how are we performing against that. We identified three segments out there: low risk, medium risk, high risk. And typically, with TransUnion, they have these kind of bad rates. Low risk has a 1.5%, and high risk has a greater than 12%. Now, what is defined as a bad rate? Bad rate is basically 90+ 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 their 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 indexed manner, has actually fallen down by close to 16% at a time when my book is actually growing. My buckets one, two, and three 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 two-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 velocity 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 two large OEMs. Hero and Honda, as you know, control more than 50% of the market today.
With two of them, we are having an integration which has been done, and we were live in September 2025 to ensure that we are able to cater to them in a much better way. During the festive season, two other large manufacturers, and I don't need to tell you the names of them because there are really only four 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. Ashish, just before me, spoke about some of the capabilities that are there.
We've enhanced some of these capabilities for our two-wheeler dealers because the requirements of a two-wheeler dealer are relatively further different from what is there. The next video will actually take you through this. I would like to extend my appreciation to the L&T Finance team for developing the Partner Planet app, a truly remarkable initiative. The platform seamlessly integrates key dealer requirements such as TA status, disbursement, delinquency, RC pendency, pre-approved TA limits, statement of account, and customer-wise case tracking all in one unified interface. Disbursement status availability is really helpful, wherein you can plan your business, and that helps in improving your finances and the stocking of the vehicle. 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, of Honda, etc., 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 logins 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 our 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 zero DPD collections 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 zero 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 two-wheeler business. I'll move on to the mortgage side. Mortgage, once again, a decade-old business with us, close to about 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 our average ticket size is very nicely pegged at about INR 67-90 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+ 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 loan 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 Neo 2 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 co-pilot 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 a credit approval memo. That's already improving efficiency. If we see today, 83% of our base is 730+ on the bureau. Our TAT that we have are anywhere between four and five days. Our lowest TAT across both is actually about one 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 all these are collection parameters, our collections efficiency has improved marginally, and if we look at our 30+ DPD, it has actually fallen.
Overall, 30+ and zero 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 five 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 pre-buying 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, Ganesh. My name is Manish Gupta, and I lead the unsecured business for L&T Finance.
Now, we've heard Debarag, Sudipta, and Ramesh kind of talk about multiple initiatives on the tech and the data side. I think personal loan has actually been one of the biggest beneficiaries of that particular transformation that has happened in the last one year. 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 one year. We actually our whole idea is to actually embrace technology, embrace innovation, to actually build a very strong quality portfolio. We talked about two 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 RACE last year. 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'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 one or two 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%- 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+ CIBIL, which are actually very, very good, that number has moved from 54% in FY25 to about 69% in FY26, and what is also interesting is that basically our customers who we are sourcing and who have a secured tradeline. Now, what do we mean by secured dominant tradeline? The customers who have a home loan or a two-wheeler before, and they've shown a certain behavior on that, has actually increased from 21%- 37%. So we are getting quality customers across our book, and not just from a large partnership, but across all our partnerships that you see here. Now, if I go to the actual numbers, September 2024, we were actually disbursing about INR 476 crores of personal loans. Now, that number has gone up 146% to about 1170. 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 12 crores contribution to now 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 four big parameters that I'm talking about, bounce rates, our bounce rates have gone down from 14.1%- 8.2%, which is a decrease of almost 42%.
Self-care, which Ganesh 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 2024 to about 34%, and it's keeping on increasing, and our GNS and NNS, which is gross non-starter and the net non-starter, which is the customers who are coming in the first time, are they bouncing or not? That number has gone down from 5.2%- 2.9% and 0.53%- 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.
care. 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, 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 do there 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 build 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 microservices.
And that connects to both the lead generation partners, which are basically Google Pay, PhonePe, INDmoney, Amazon, CRED, etc., and our underwriting staff, which is our internal staff, 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 detailed 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-sale 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 microservices 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 two things that 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's 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 saw the demo as well. We are doing an in-app agent. I'll actually talk about that in agentic voice-based chat interface, which is there. Revamped consumer journey, which again, you'll see in the next few days, our journeys are actually better, but we're 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 pre-approved 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 , for example, there is a dedicated effort through agency maximization that we'll be doing. A mandatory banking, of course, in Nostradamus, you've seen that it actually requires a lot of banking information. You can monitor the customer, what he's doing, how his bank account information is changing, etc., with consent. We'll actually be putting mandatory account aggregator and, of course, automated contactability through a 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. So.
Hello Rahul Jain. Please help us in verifying our details. Is your full name Rahul Vijay Jain? Yes, it's right. That's great. Next, can you please confirm if your DOB is the 16th of September, 1992? Oh, this is not my DOB. Okay, Rahul, would you please give me your date of birth as per your PAN? Yes, it's 7th of September, 1989. Could you please confirm if your PAN number is ASMPV1234H? Yes. Please help in verifying your address. Is the above-mentioned correct? Yes, it is. Thanks for verifying your details, Rahul. Just a few more steps. What is the loan amount you are planning to take? INR 500,000.
For how long are you planning to take this loan? 48 months. Please give me a moment to calculate your monthly EMI. So your monthly EMI is INR 26,514. Would you like to go ahead? Yes, go ahead. Congratulations, Rahul. Your loan has been approved. Let's proceed to signing your key fact statement.
Okay. Yeah, 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- 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 four years because we are just above a four-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 in 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 which sector is slowing down or which segment customer cohort is slowing down and which one is growing, then the segment which is around 1-10 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 two years from around 41%-43%. This is the segment that we play because our average ticket size is around 29 odd lakhs, and the minimum ticket size is around 10 odd lakhs. That's up to from 10 lakhs-1 crore is the segment that we play. Hence, this becomes a close proxy. Another defining factor about this segment is that this segment, from a delinquency perspective, also has been doing pretty well.
And in the last two years, the number has dipped from 2.3% odd 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 the 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 the 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 2025, we were operating in around 110 odd locations.
In the next two 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 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 three cohorts of customers, which is around premium, value, and core, and then try to tailor our offering based on 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, we have been very, very focused on our collection efforts. We have been able to maintain our CDC at a 99.5% level over the last four years.
The intervention of AI has enabled the self-cure and call center fees 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 two cohorts, if we map it against the industry, I would like to draw your attention to the right top table. The industry is indexed to 5-10 lakhs being the core delinquency. Then the delinquency, as we start moving up the ticket size, starts declining with the sweet spot starting from 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 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 low risk. That's the customer segment that we play, 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 plus and super prime.
As an outcome of that, what happens is that if industry standard is indexed at 100, our standard book is at 109, which then subsequently in the buckets, our number starts becoming less. So that's the kind of play which is 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, our 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 %. We continue to be focused on this sub-one crore ticket segment by adding features so that our products become much more attractive to this particular segment.
As the next stage, what we want to build is to 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, 12-18 months duration timeframe. So that's the aim that we are chasing. In terms of collection strategy, we have a two-pronged approach. The first approach is that we understand the portfolio performance and weed out the segments that we do not want to entertain. So around that, various policy interventions keep coming, which is at the top of the funnel that the 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 applicably seen in data that our collection efficiency at the zero DPD level has been in the range of 99.5% for the last two years. That's the approach that we have given, and that's the outcome that we have. Now, a bit of a deep dive into Cyclops 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 need 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.
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. Hence, a unified view of the customer was not seen at a point of time that we used to be seen over a point of time, but not at the same point of time, resulting in higher turnaround time. What Cyclops solves is that it takes all the data points and buckets the customer into premium, value, and core upfront itself. 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 two 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 you want me to 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 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 one-fifth 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.
On this, we are further building on phase II and phase III of Helios, which is around other documents. Phase I 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 will help our partners to have a better lead sourcing and give a better customer experience. To understand what exactly were the challenges which this AI solved for, we have seen the video of it.
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 is available in that bureau report if it is made there. Most importantly, it allows the underwriter or it nudges the underwriter to ask certain specific PD questions based on the deviations or variations that it has seen in the bureau track record of that particular customer.
On 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 gold loan offering, which is through various feature addition, tapping into the new markets, riding onto the existing distribution of either gold or farm or two-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 four anchors and two partners which we have onboarded. So overall, just to sum it all, we'll continue to be focused on attracting a 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. Gold 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 secure product in the form of gold finance was found a suitable product. It was also necessary from the following key aspects.
As a whole, banks and NBFCs have lent close to 13 lakhs in the gold finance market. The NBFC share straddles between 20%-30% in that. We do have a physical presence in almost 2,200 locations in the country already. We believe that getting into a gold finance business would be the right fit for us. That is how we ended up scouting a boutique firm in Chandigarh who is present in north and west. We onboarded that particular business by way of a slump sale somewhere on 9th June exactly. We did the usual protocol of completing due diligence, both financial, technical, and legal. 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 transferred on our side. 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 a lakh customer with almost 1,300 crore of an AUM which we acquired. In a short period that is gone by, we have ensured that we maintained 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 is gone by, and we have seen close to 200 crore of a 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.
It was important because we were keen to have a seamless onboarding of all the customers on our ecosystem without disturbing. 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 the, in fact, an enriched experience. That happened because of seamless integration of IT systems we shared. We also had to give a proper orientation to all the employees which came to L&T Finance. 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 a 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 is 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 customer's 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 Association's 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 prowess of AI and digital which we have inherently developed in L&T Finance over the last 12- 18 months period, and that has been implemented with 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 five 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+ security protocols. And that is important given the fact that we understand that gold is quite close to customers' heart is being given the jewelry.
We want to ensure that the faith with which customers would be depositing 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 a bit of an archaic system, if I were to use that word. Nonetheless, it is a two-factor security protocol which is there. But Debarag, he 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 an 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 the 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 nine 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 two 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 crore of gold loans which they have availed from other financiers. And we want to make use of tapping those customers into the 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 base. We did not have that kind of a structure so far. We are moving towards there from the holy 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 serviced for our customer. Here is a quick video of what's there in that. 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. Good evening, everyone. This is the last presentation before the Q&A session and dinner. Please bear with me. Since we all know that Jasprit Bumrah is the brand ambassador for L&T Finance, so I'm going to begin with a message from him for all of you.
Good evening, everyone. How are you? A warm welcome to the L&T Finance Investor Digital Day. We are thrilled to have you join us today.
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. 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 him 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 Two Wheeler Loan campaign is, "अब लोन मिलेगा बुमराह की स्पीड पर," और हमारे L&T Finance बिजनेस लोन्स हैं आपके बिजनेस का गेम चेंजर। Zooming into our campaign, which is Just Zoom Two 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. 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. 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 leveraged the technology. So on a real-time basis, audio ads were placed in the metro based on the time of the day. So here is what, let's say, a traveler returning back from work would hear.
घर पहुंचने से पहले लोन मिलेगा बुमराह की स्पीड पे। L&T Finance presents L&T Finance Just Zoom Two-Wheeler Loans. Visit your nearest Two-Wheeler dealership now.
So as you can see, this is perfectly timed because of technology. So 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.
क्या बात है, ये तो मेरे जितना फास्ट है। L&T Finance Just Zoom Two-Wheeler Loans.
To experience this, you should visit the Bowl Like Bumrah in the demo zone, which is to my right. Moving on, of course, these were the results of the T20 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. We took up advertising through TV media on news channels and outdoor hoardings, as well as digital video. We generated high levels of engagement with a very close business community. How did we do that?
We created for the very first time an AI-led microsite, which enabled these business owners to simply go 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 their picture, and the AI model would create a business poster exclusively for them. 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. This generated us great business results. 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%- 31%. And the consideration for L&T Finance business loans improved from 8%-2 0%. 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 GenAI, we create topical and festive videos all around the year, which have high frequency. And here's what it looks like.
इस दिवाली, रोशन हो बस घर नहीं, आपके सपने भी। अब हर अरमान को मिले नई रोशनी। L&T Finance Personal Loan के साथ, चाहे घर सजाना हो या नई शुरुआत करनी हो, कम डॉक्यूमेंट्स, आसान डिजिटल प्रक्रिया, और तुरंत अप्रूवल के साथ आपका त्योहार बने और भी खास। तो इस दिवाली, रोशनी सिर्फ दिए की नहीं, आपके सपनों की भी जगमग हो। L&T Finance Personal Loan. हर खुशी अब आपके करीब। हैप्पी दिवाली।
Who would believe this is all GenAI generated? Not just this, GenAI also helps us to create personalized messages in the customer journey, each one at an individual level. Let's see how.
Great news. Your L&T Finance Personal Loan has been successfully sanctioned. You're just a few steps away from receiving the funds. To ensure a smooth disbursement, please complete the digitized application at the earliest. Please note that valid address proof will be required during the video verification process. Click on Continue Application now and unlock your funds easily.
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 GenAI. So that's exactly what we did for our Diwali campaign, an entire campaign generated through GenAI, both from visuals to sound to graphics, all of that done using GenAI. Let's have a look.
महफ़िल पे रंग लाएगी, रंग लाएगी, बेगिस्ता, क्योंकि अब पूरा होगा हर ख़्वाब, सब होंगे कामयाब, सब होंगे कामयाब, सब होंगे कामयाब।
So this generated high views on Instagram and engagement. So here's wishing from all of us, इस दिवाली सब होंगे कामयाब L&T Finance के साथ। 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. We request audience to restrict their questions to one for maximum participation. 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. Yes. Sir. Kindly give me your name and organization before you ask the question.
Sir, this is
Ramesh Bhojwani from Mehta Vakil & Co. First and foremost, many heartiest congratulations on such a wonderful presentation, encapsulating 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 masterpieces communicated. One is Cyclops and the 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.
These are like one of the things 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 origin of 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 say, Nostradamus was a French philosopher and a clairvoyant who saw the future.
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? That's the process. But it also has. Nostradamus also has a very large cross-language along with it. Nostradamus also identifies customers' cross-sell potential much earlier than others do. 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 two to three months, three years down the line depending upon his income growth. Nostradamus helps us to identify those also. Together, these systems are supposed to operate in conjunction. Cyclops is obviously the origination layer.
Nostradamus is the 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 12-1 8 months to build where the signals from Nostradamus will go and automatically tighten parameters in Cyclops. To that extent that, you know, really the credit managers can go on holiday and sit on a beach while the system operates on its own, auto-correcting itself. Thank you and all the best.
My name is Nitin Naik. Hello. My name is Nitin Naik, director of Naik Consulting. And I would like to ask the question that, you know, with the last division which you have, that is the Gold Finance division. The gentleman said that, you know, when it was started, the very first day itself, you could, on day one, your customer could get the feeling of, you know, paperless office. So does it mean that while the negotiations were going on for the takeover, at the same time, you were 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 on that, 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 three-month integration process that we had, we built everything.
Thank you. Thank you so much. Good evening, sir. 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. Thanks.
See, it's very early days, right? So Cyclops, for example, two-wheelers is about 14, 15 months in operation. You know, in the tractor business, it's about six to eight months in operation, and in the SME business, it's only two months in operation, but whatever I can say, gleaning the learnings from the 14 months of two-wheeler operations, you see, two-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 two-wheeler portfolio industry bounce rates are anywhere between 20%-22%. This is the average bounce rate. Our Cyclops underwritten two-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 two-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 finely dissecting customers who 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%-60% is what my gut feel is. But actually, we would need to run Cyclops continuously for three 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, you know, 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 know, you're looking at cyclicality, but ultimately, you know, end result has to be reduction in, you know, either collection cost or credit cost. And as Sudipta mentioned, the first impact of this will be felt in fourth quarter as far as two-wheelers is concerned because there has to be some seasoning involved. And two-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 have been, you know, have 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 FY27 will be the year when you will start seeing significant part of the books have, you know, routed or being underwritten through Cyclops.
Accordingly, H2 of FY27 will really be, I would say, the full impact of credit cost, where Sudipta also mentioned in his presentation that what we are targeting is ultimately to move from about 2.75%- 3% down to about 2%. So directionally, you have seen that barring, you know, microloans, the challenge which we went through over the last about four 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. You know, the fact is that, you know, 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 a festive month. And typically, what happens in festive month, your volumes go up. And normal common sense logic says, as your volumes go up, normally a large number of bad customers might sleep in, right? Because what happens is in a large manual process, if you run a sort of credit administration system that is largely manual, you know, as volumes go up, generally, credit discipline tends to fray on the edges. In the month of October, we processed about 65,000-70,000 two-wheeler loans. Sorry, in the month of September. In the month of October, we processed, we actually disbursed about 130,000 two-wheeler loans.
So actually, almost a double-size jump in volumes, right? So obviously, our feel was, and my worry was that, you know, we will see the gross non-starters 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 an aha moment for us, right? So as I said, you know, 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...
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.
Hello. Hi. Right here, all the way with the mic. My name is Parikshith. I'm from PKD Advisors. I have a two-part question. If I understand correctly, Cyclops is a fully automated underwriting system. And if that is the case, then why Helios even exists? Because that's, it's both targeting the SME loans. And if Cyclops is well on its way, then Helios would be redundant.
Very good question. The fact is that Cyclops is fully automated. It's very good, very good, you know, 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, two-wheeler, average loan ticket size is INR 1.1 lakhs, INR 1.2 lakhs, right? So we are comfortable with the 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 have to understand that in machines, you can have bugs, right? When we built 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 files 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, you know, we would like a human to look at it till we are confident enough after a history has built of about 12-18 months to say the human is not needed anymore. Cyclops can do it, so then, you know, Cyclops will do it, but we need to travel that distance, so that's why, you know, the Helios Copilot is currently helping our underwriters to supplement the work that Cyclops is doing.
Thank you, and if I don't mind, I'll just ask another question. A lot of the data that you showed shows a downward trend. But how have you controlled for the overall macros situation? Because even the macros have improved, right? So how should we look at your data from that angle?
See, macros have improved in pockets. Macros, you know, macros again, you know, move in cycles, right? One of the things which you saw in the Nostradamus tool was that the Nostradamus also consumes macro scorec ards, right? There is a set of macro score cards which have been built for Nostradamus. Those score cards 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 that area, the macro score card will overlay on that and sort of predict the sort of forward-looking delinquency in that area. In that way, we sort of try to address the macros through Nostradamus right now.
As of now, for example, you know, yes, business growth, you know, there is a GS2 impact on tractors. There's good rain. So generally, that has a positive impact. So macros give you your positive tailwinds. Macros give you headwinds as well, right? So, you know, you cannot plan for macros, unfortunately. So that is why in part of Nostradamus, the tool is that at least, you know, as and when macros keep on happening, at least it gives you an advanced warning as to what might happen.
This continuation of the same question, if Helios can do it, then what is the difference between Cyclops and Helios? Like, you know, if you're going to...
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 gives you the PD questions, right? Similarly, Helios has got many parts. Helios has got a bureau analyzer. Helios has got a banking statement analyzer. Helios will have a GST analyzer. Helios will have, you know, legal technical as they talked about it. So all these modules will come and build, right? Now, for example, Helios right now we have done for SME. But Helios will really be liberating for the mortgage underwriter. 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 you implement it for mortgage, will crash that number by about one-tenth, right? So a file that takes INR 20,000 to INR 20,000 to process should be done at INR 1,000.
Thank you so much.
Hi, good evening. It's Kedar 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 are 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 speaks in unison, 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, 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 in unison, when the entire organization is focused on one key outcome goal, and when you are trying to granularly distill these 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, you know, 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 nooks and corners of the country. And that is what we have been trying to do.
You know, our people have literally been on the road almost like, you know, two to three 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.
Thank you. Thanks. Good evening. Hi, good evening. 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, the 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 the management team?
Well, there are many things we worry about. I'll be very, very candid about it. We are present in some cyclical businesses. We are present in businesses where there are event risks. For example, JLG Lending is a business that has event risk. We saw, you know, policy decisions in one particular state suddenly tank collection efficiencies, you know, and our team struggling to keep afloat. 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 one thing which keeps us awake is, you know, the specter of losing our best people.
You know, over the last couple of quarters, we have become a magnet for technology talent because of some of the work that we have been doing. And trust me, word spreads very fast in the community that, you know, this place you do get to do good work, right? And people have been coming and joining us. You know, in fact, for that matter, you know, this year, we have actually got very good people from all the top varsities. And we really didn't struggle to hire them because we felt that, you know, all of them looked at the Race 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 asked them why, what pushed you, they said Race 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, you know, you know, 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 dragged from our backbook, which is underwritten by legacy algorithms. And then when that clears, 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.
Yes, Himanshu.
Thank you very much and all the best to the entire team. Thank you.
Hi, sir. Hi, sir. This is Himanshu here from Aditya Birla Sun Life. So just firstly, congratulations on the great investment study that you have held 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+ lender, at the start of the fiasco, we are also having somewhere between 50%-20% of the book. But the outcomes, the way you have delivered has been far better. Can you just help us by giving some prudent measures in your underwriting, which ultimately results in better outcomes?
I think, you know, if you go back to some of our presentations, you know, you will get that. But, you know, 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, you know, greater than two lakhs, you know, threshold, you know, should not be breached, you know, not more than three 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 was 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 January 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 said that we need to bring down the, because we had a sense that something was coming, we said that we need to bring down our accounts per collector to about 450-480 levels. So we actually pushed in 1,000 more people into collections, you know, before even the crisis hit to bring down our accounts per collector from 580- 480 or 450 odds, depending upon the markets, right?
So, and plus, you know, at the end of the day, the JLG business is about brutal discipline, right? The fact is that, you know, unless you're able to maintain the discipline of the workforce and a large amount of workforce on the ground, and unless you digitize to the last tier 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 have an extremely hardworking management team, and our every process in the JLG business is digitized, right? Through all this and obviously through proactive portfolio management, you know, 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, you know, 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. So second is the Project Cyclops and some of the rule engines which you have developed in the last, basically sometimes the last one to two years. I think, is it like, yeah, 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, very, I'll be very, very emphatic about this. Cyclops is a ground-up, natively designed machine. What exists in Cyclops does not, to 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 those scorecards using streaming data is something which has been probably built in the country for the first time. And to pull data from seven, eight disparate sources, make sense of it, you know, 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, you know, it doesn't exist in any large organization.
Sure. Thanks.
Yeah. I think we have room for just one question, if at all there is any.
It is zero downtime.
That's great. Thanks.
That does not qualify the question, but so maybe take one last question, one last one or two questions if there are. Gentlemen there.
My name is Vipul Shah. I'm an individual investor. I just want to know what does it cost to build all these platforms and what will be the cost going forward every year?
Yeah, the cost of building, I mean, the amount that we have spent till date on Cyclops and Nostradamus put together is about INR 60 crore. But it's still a work in progress. So Nostradamus is still in the making. We are 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 crore. Sorry. You had something more? Next year, see, out of the 60, about INR 35-40 crore have been spent on Cyclops. The balance INR 20 crore 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.
The cost of, you know, the employee cost, all these engineers are working around the clock. It depends on the time frame required to actually go through that implementation, is what will. Significant man hours.
Significant man hours would have gone into this. So we just don't count that.
So it took flat four months to implement Cyclops, and it was done through the shift. So it was not that it was a nine to five 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 it would have taken more.
Is there no question?
Yeah. Thank you, sir. With this, thank you, sir, 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, you know, the tag that you're wearing, you know, the access code that you have is valid for RACE 2025 tomorrow also. So please do visit us. There will be more stalls here coming by for RACE 2025. There'll be interesting speakers coming here for RACE 2025 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, you know, Bowl Like Bumrah in that side. So in case any one of you have time, I'd urge you to spend some time there. You'll get a hands-on feel of, you know, what some of the solutions are. Our teams are waiting out on the demo area. The demo area will also be available tomorrow as well for RACE. 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.