Why data is the bedrock of AI All About AI EP2

Every once in a while, humans invent a technology that puts human civilization on a completely unimagined trajectory. Internet, perhaps, was the last such technological innovation. AI, or artificial intelligence, is set to be the next, they say. Just like the Internet, it will change everything about our lives including the way we work and run businesses. The pace of user centric AI based tools like chat, GPD, Copilot has been stupendous. Tools like Einstein developed by sales force for businesses are changing the way. For instance, a fintech disperses loans or a shipping based company figures out the ideal route and load per container. Hello and welcome to the second episode of Mint's all new series on AI. Now, unless you intend to live under the rock in the future, you need to watch this entire series. And why do I say that? Well, I say that because AI is about to take our lives by * and we have put together a very interesting panel. The 1st guest is with us. I have Mr. Sachin Mangalat, CFO and CIO, Hero Fincorp to talk to us about AI and how it is going to change our lives. And Hero Fincorp. Before we start talking about, you know, what you're noticing on the ground, let me just get your initial thoughts on AI and how it is going to transform our lives. Thank you. Thank you Abhishek for inviting me on the show. Very excited to be here. Hero Fin Corp is a tech LED financial services firm. So when I look at how the scenario has changed over the past few years, we have been able to leverage much more data than what we used to do in the past. So first of all, there is a lot of data that retail customers generate. People use lot of digital stuff including mobile. They browse and they have a lot of information that they generate. For example, you would have noticed that there is much more penetration of banking recently. So that means that you have banking data. When you use wallets or UPI, you generate a lot more data. If you think about the merchants, SMEs in the country, lot of them are becoming formal now. They have GST data, for instance, right? So the first thing which comes to my mind is that we have much more data so that we understand the customer, whether it is a retail customer or whether it's a business merchant, much more data than what we had in the past. Now how we use that data is very important. So that's where AI comes in, Machine learning comes in. We have the tools now, but I would say the building block is that we have much more reliable, accurate data, which describes the customer now, right? And if you could give us a low down of data as a fossil fuel on which this entire massive engine of AI is supposed to run runs. And the more accurate the input, the better solutions you're going to get. So what you've seen, because the tech is in its recent stage, of course, there are amazing tools which are already out there. But if you could just sort of give us some examples of that. First, let me talk about somebody in our space who requires a loan for some purpose in their lives. So initially the customer will come searching for a bank for a NBFC seeking a loan. Now how we advertise, we would go to a newspaper or a television to put out an ad pass. Now while all those are still there, the AI technology now allows us to pinpoint the customer who is requiring such and such a loan and who has the ability to pay for such a loan. So if you can sort of, you know, give us a bit of a comparison after AI, what has happened on ground Because at the end of the day, Hero of Incorp is in the business of disbursing loans, small loans. And these are the people who might not have the kind of credit profile which makes them, you know, which makes you gives you the kind of surety that an organization like yours looks at. There is much more AYC data available, for example, in Aadhar, for instance, in credit bureaus. All of that now is a non intrusive way of seeking that information from the customer. You can provide, you can provide your consent. We can read the KYC data that you have now. After that, to be able to judge whether you have the repayment capability or what's the loan size that might be appropriate for you earlier, you have to provide bank statements to us. Now with technology like account aggregate or for instance, you can give us consent to read the banking data to the extent that is required for us to be able to underwrite your requirement. Now, once you have this data, how does it happen now? Now, earlier there was a subjective process of underwriting you, right? So somebody would look at your data, somebody would then decide whether it fits your risk policy, how much loan can this individual get? Now, what data in AI has done is that it is removed the subjectivity, it has made the very objective process. Now because of the kind of data is available, you're able to now bring in a lot more population into your ability to finance compared to what it used to be in the past. Now the the way the machine learning models work and a lot of new technology has come vendors like Salesforce, software, etcetera, now makes it very easy to run those technology platforms. You'd have seen the developments in AI which just happened, Gen. AI, open AI, you know, many of you would have heard of all that. But underlying factor is that the revolutions in computing now allows us to manage much more data and you are able to then decide what's the right product for the customer. You would say that the data is so accurate now that the number of bad loads, etc has come down. Have you noticed that what has now happened is because of the way the how much data is available and the objective way of processing the data, you get a much more accurate reading of the risk that every customer represents right now. What is the, what you need in financial services is your ability to predict the risk, predict the losses. Once you're able to predict, you know, what is the likelihood of a loss that every customer segment brings, you can have any appropriate product. For me right now, while that might have happened, the question of ethics comes in because at the end of the day, we're talking about customer data and that data is being collected from, as you were mentioning, UPI, Facebook, Google, YouTube usage, etc. The various government and their agencies have started to make available formal data which much beyond what it used to be in the past. For example, RBI has an Innovation Center which makes available the amount of milk that a farmer produces, which never used to be the case in the past. Lot of land records have now started getting digitized right, which is which? Which never was. The scan was never there in the past. The word these kind of things do is that you get formal dependable data now. There was a phase where people used to take data from phones and all that. Now much more formal reliable data is now available. Now I am sure most of you have heard of account aggregator. It's a consent based architect. You give me consent. I can see your formal banking there. So you would say more transparency is coming when it comes to the kind of data which is being collected from customers, that terms and agreements form which pops on your phone every time you download an app. Nobody necessarily sort of goes through the entire literature I would say. So the ethics of the situation, what would you say? Whatever organizations doing to ensure that the customers are sort of, you know, can sleep easy, not worried that something which is which, which they're not, which they would not want to consent to has not been agreed to. That is governed mostly by policy framework put out by the government as well as the RBI and the financial services industry. It tells you first of all, the customer has to very in a very transparent fashion know what loan they're taking and what data we're coming. They're trying to tell me that that forms going to become like even more succinct like in four or five points. And I can read them and agree both due to policy as well as due to competition, right. So I was just telling joking before the session. We have new application forms where we just ask for 11 fields of data. The remaining we read from various data sources without bothering the customer. But the policy now states that you have access to the data. We only manage the data till you allow us. And once the purpose of that data is over, whether your loan is over, we have to delete and destroy the data. By the way, because of fear mongering around the kind of data which is out there as far as the customer is concerned and how organizations are sort of, you know, sharing it amongst themselves, leveraging it, perhaps even selling it. I mean, so glad to know from you that when it comes to policy and when it comes to organization by Kios, the commitment that, you know, this is for a very specific purpose at the end of it, well, we're going to put it away, delete it. So that gives customer a great surety. So thank you so much for sharing that with us. And that's about as much time we had for this session. Thank you so much, Sachin. Thank you so much for talking to us. Thank you very much. So Next up on stage with me is Mr. Arun Kumar Parameswaran, SVP and MD, Salesforce India. Now. Welcome, Mr. Parameswaran, thank you so much for joining us. Thank you so much for having me here. So Salesforce, of course, is an organization which offers solutions to organization when we were talking to Sajan earlier and offers the kind of solutions and tools which they can use to process all this massive data which is coming in constantly. When it comes to that, we will discuss, you know, we'll break it down in multiple parts and discuss it. But before we start, just let us just get some thoughts of yours on AI because when it comes to AI, there is a new development every single day. I mean, I would say the technology is evolving every day instead of sort of, you know, incrementally improving. So your first few thoughts. First things first, you know we're a 25 year old company and we started transforming the way our customers engage with their customers when we bought CRM, our customer relationship management to the fore back in 1999. So if you think about our journey over the last 25 years, we've been evolving and we've been innovating with the latest technology trends. And now if you talk about AI, we've been doing AI for the last almost 10 years, right? That's been more on the predictive side. So things like when you're shopping and you get a next suggestion that typically is predictive AI At work, we make something like 2.5 trillion AI predictions a week for our customers and our platform today, right? And of course, the last 18 months, we've seen the whole focus shift from predictive to generative AI. And of course, now the scale at which this is operating and the promise that it holds, it clearly is going to disrupt and transform every industry, every function, everything that we do. So in that sense, I think it is easily one of the most disruptive technology trends that I've seen. And I think every customer is keen to figure out how they can jump on that bandwagon. So we are innovating fast and furious to help our customers in that journey, right? So it has been around for a while, of course, but ChatGPT, I mean, the first model was launched and within a few months of that, we had the next version. That is the rate at which technology is evolving. So in what ways do you think the technology that we have already is going to change the way businesses are dealing with customers? I think it's important you distinguish consumer AI from what I call business AI, right? Consumer AI chat, JPD is great. You want to write a love letter, you want to have an AI version of you. I think those things work fine, but the moment you start thinking about putting AI into business, you know you basically have to solve the three problems, right? You've got to solve for a trust layer problem, which is how do you make sure that AI, you know in your organization is taking care of all the trust related parameters that you should be worried about. The underlying data model is going to be the key to how how accurate your AI predictions can be. You've seen the hallucinations on all the different consumer AI tools that people have talked about. And I think the third part is, you know, how we're going to drive efficiency and productivity for your organization. So I think the world of AI in business is very different from the world of AI with you as a consumer and as a ChatGPT, you know, playing around with that tool, so to speak. And so I think that's where we're starting to see, you know, the separation happen and a lot of our customers are keen on figuring out how they solve those problems, right? Data, for example, you know, we, we know that almost every organization has more than 70% of the data sitting in silos. Now, if you want to bring AI to work, your first order of, of task, if you will, is to how do you bring all that data together? So, you know, we've got a 5 steps to it, You know, AI enterprise vision that we are setting for our customers saying bring your data together, harmonize all your data, have a single view of your customer. We call it the customer 360, right? Your sales system should not have a different view of the customer from your service system, different from the, what the marketing guys use, right? So customer 360 is critical. AI is going to require, you know, organizations to breakthrough silos and work and collaborate. So how good are you from a collaboration perspective, right? And how, what tools are you using? Because I think old generation tools will not find a way to survive in the new world. I think the 4th part of this is going to be around how do you drive your systems to come together to bring intelligent analytics and insights? Because a is going to throw a lot of data at you. How do you bring that data to a simple visual interface that says, OK, if I'm looking at a gender type with an age group X in a city Y, what is my addressable market? Can I just, you know, in a very simple way show you that? And of course, the last one I talked about the trust layer. So these are the, we call it the five steps to the yeah, enterprise. And that's what we're working with all our customers on, right. So could you give examples of the kind of solutions which Salesforce have been developing for organizations, let's say like Eurofin Corp? Yeah, I mean, he was talking about the whole credit underwriting process, right? I mean, we've now within our platform today have created solutions where a whole range of underwriting processes that were very manual today is going to get completely automated by AI, right. He talked about all the different bureaus, a conductor, creative frameworks, all of this data is available, but we still went through a fairly manual process in which we were doing this. So today in the Salesforce platform, we have the ability to do all of this in a very simple way, right? I'll give you an example. I use in Salesforce use Slack as a collaboration tool, right? And recently, you know, we did some innovation. So when I'm going to meet a customer, typically my team will create a briefing document and that process could take up to two hours. You know, the manager reviewing it, her manager, his manager reviewing it. Today, we push a button right? In Salesforce, Slack basically goes in chromes through all of the data we have across all of our systems and literally generates a briefing document for me on that customer within 10 seconds, right? There's a classical example of how AI is going to foundationally revolutionize how we're doing. If you're a contact center person and you're taking a call from a customer, typically at the end of each call, that contact center Rep will have to summarize the call and put the actions in place, right? That typically could take eight to 10 minutes. You could have a call for four minutes, but spend 10 minutes summarizing it. Today, the push of a button in our service Cloud platform, you could pretty much summarize an entire call. You could read the transcripts and look at the tone of the conversation and you could make out that maybe this was not a very happy customer. And maybe then decide maybe we should nudge the customer with some offer to kind of make him feel a little better because we couldn't deal with the kind of. So it's all coming down to two things. How well do you know a customer, which comes down to how well your data is organized? How personalized are you going to be, like Sajan said, the right product for that customer at that point in time and how do you wow him? This is the battle that every customer of ours are trying to win with their customers like Sajan is trying to win with his customers. Our job is to enable him to be able to build those platforms, right? And when it comes to that, one of the key considerations is to ensure data integrity. Privacy and security are also the other important concerns. So in a shared data environment, especially while leveraging AI, what do you think organizations can do need to do to ensure that all of that is exactly as it needs to be from everybody's point of view? So our view is we'll take that complexity out of the equation for you, right? So when you come to Salesforce, AI starts with our trust layer. And that trust layer is all about a one looking at your data and making sure that the data is, is where it needs to be, that the data is not used. We don't own customers data. We don't store their data. We're very clear about our data use and policy. Like he said, we are very, very black and white in that sense. Then beyond that, once you start throwing the data into the large language models that you're building, you've got a whole bunch of issues you have to deal with, right? How am I dealing with toxicity? Am I dealing with bias, right? Am I saying that you know what I'm going to give a loan only to this particular person in a city, but I'm not going to give it to somebody in a town. I mean, you don't want that kind of biases coming in right? The dark side of profile exactly right. So you don't you AI should be for the good and therefore how do you solve toxicity? How do you solve bias? How do you solve hallucination right now interestingly, in India, you know was the customer and said, look, how are you dealing with hallucination? He said, look, my manual process I am assuming has 50% hallucination. So if AI is going to bring it down to 25, I'm still ahead of the of the map, so to speak, right? So I think these are all issues that eventually will will become much bigger at scale. Like today we don't have AI deployments at scale. People are still kicking the tires, right? The cost, the cost economics, particularly in our emerging market like India is not where it is compared to the developed world, right? Think about it, in the in India, a call center person probably earns ₹10,000 a month, that person probably earns $100,000 US, right? So the economics plays very differently. So I think you're going to see this play over a period of time, right? It was a great conversation. Thank you so much for sharing your thoughts with us. But before I let you go, there might be questions from the audience. So if anybody has a question, there's a mic right here please. So basically I understand that being a solution provider, the solutions are mainly focused on the business side of it. But as a user, if I ask like certain questions where the recent trends show that there are a lot of data threats and identity threats which are happening and these cyber frauds are increasing to a certain extent. So are there any measures which are like jointly taken by the organizations like sales force and government to implement the same or is there any measures that your organization is doing internally to secure that data theft would not be there? Trust and customer success are top two values of the company. You have to understand that customers like, you know of Incorp, trust their customer data with us, right? Because they run their entire business on our platform, which means, you know their customer data is sitting with us. This is not something we take very lightly. In fact, it's one of the biggest investments we make on behalf of our customers, right? Which is how do we make sure we protect their data, which is essentially your data? How do we make sure that when you're providing information to them, how do we make sure we encrypt it? How do we make sure that the data is only used for the purposes, like you said, it's used for the purposes meant to be? How do we make sure that data is not stored in a way that can potentially be compromised? And so all of this we've been doing just FII, right? We started the combination of software as a service. So we started this 25 years back when the concept didn't even exist, right? We're the first company to bring in a software as a service capability. And so we've been doing this for 25 years. And I think, you know, we've stayed always 2 steps ahead of where we need to be because we have to start predicting what those threats could be. So I think with us as a SAS platform, because of the way we've evolved over the last 25 years, I think most of our customers know that their data is safe, which means you should feel assured that your data is safe. Thank you so much. Thank you so much. Thank you, Thank you, thank you. All right, Next up with me is Mr. Kabul Mahajan, Global CITO, all cargo logistics. Thank you so much, Mr. Mahajan, for joining us. First question and before I open it up and talk specifically about the logistics industry, supply chain, etc, your view about how transformative AI is, OK, thank you for having me. It's not just transformative, it's hyper transformative with my view and it's already started impacting industries, including ours. Obviously financial sectors is majorly using the technology. So in terms of I think how it'll power the next 10 years, over a decade for different industries and verticals, I believe that it's going to augment every business process going forward. So it's going to bring a lot of cognition into the enterprise. By that what I mean is every business process without a user actually knowing, there will be an AI workhorse in the background that will be doing a lot of the heavy lifting or a lot of the repetitive tasks would get automated within the core systems of the enterprises. So that from from my point of view is a very, you know is a very big statement because we have been working through traditional Erps and how those Erps will now transform and you know, assuring new service lines, new product portfolios, new competitive edge. All of that is going to change because now the technology has moved away from Pocs to actually just breathing there on the ground, right? Remains, still remains a labor force heavy industry, right? And there are fears around the fact that how AI might take away a lot of these jobs, the augmentation that you were talking about, a lot of processes. So what have you noticed? Because I mean, it's been a while, as we were told earlier that you know, technology has been around for a while, solutions are being implemented by the organizations. But this one big fear. What have you noticed in your industry? Yeah, I think this fear, like the fellow spoke, speakers spoke about, I've also gone through decades of transformation. It's always a fear. Ask that question. And again and again is simply because men's audience happens to be in that 20 to 55 bracket, right. So this entire jobs question and all doomsday predictions around it do end up sort of, you know, making them a little concern. Yeah. So, so very relevant. And I think that every generation has gone through that because there have been leaps in technologies, starting with the, you know, how a desktop computer when it came out, people thought jobs will go away. You know, we won't have accountants anymore, but we still have I think more accountants and we had decades back. So coming back to this particular, I think technology or piece of technology is, in my view, I don't have exact statistics, but I think the prediction says that you would have additional 15 to 20 million jobs that'll get created. Apart from AI and big data, machine learning, etc, the other big theme in our lives, I mean, you know, wherever we go, where people are very, very conscious of their carbon footprint, etc. And do you think AI can play some role when it comes to organizations and industry like yours, Logistics, some role there? Yeah. Yeah. So I'll give you an example where it's playing a role for us. So we run three different lines of businesses. There's a business called Gathi that does express business in India. So that's surface in air and then we have the shipping, right. So for Express in particular, because it's more relevant, more easy to correlate, what we do is that we run all this network network. So all these vehicles are part of our network. So few thousand vehicles that run on specific routes, right? And they're scheduled like an airline runs from a flight to Delhi, Bombay goes at 5:00 or 8:00 and 10:00 from Vista and sometimes say Air India, whatever. Similarly, we run the network like that and that's a traditional way of running the network. Ah, so if the way the truck is loaded, not loaded, it has to move because maybe the load from origin is low, but the destination volumes are higher. So the truck has to reach to pick the material and come back to the to the destination. Now. So there we are using, in fact last year we did a project, we were able to optimize the route by about 20%. We did the entire route optimization. It saved us few $1,000,000 where we used AI models to predict what are how do we design a network. So we redesigned the network using the AI model, right. There's a lot that went behind it. I'll spare it for another conversation, but but so that that's one. So by doing that, you're doing lesser trips, you are running trucks that are running more loads. So instead of running three partially empty trucks, you can run with two. So your carbon footprint, numbers of your mission, your cost, everything comes down. And similarly for vessels, we have now kind of lined up with the lot of the companies that give us the carbon emission, you know, for the different vessels. So and that is being showed at the part of the court because every company now has PSG, you know, that's functional and they want to see how are they reducing the carbon footprint. So a customer can, you know, decide more proactively to choose a route that kind of emits lesser carbon. So, and that is where we're using AI to prompt them to pick up those routes, although they are a little more expensive to sail on. So thank you so much for your time. And now I will again go back to the audience for some questions that they might have for you, please. Yes, hi Sir. So I really wanted to circle back to the fear of jobs question that we had. And I feel like there are two main conflicting points that I have with it. On the one hand, we say that, you know, AI is going to create a lot more jobs now. But then on the other hand, we always boast of that. Oh, earlier this job used to be done by ABCDE and now this job can only be done with a with the help of AI. So we did eliminate the BCDE out of the equation. So that is one point. And the second point is what do we classify as intelligent jobs? If you look at the job profile 10 years back and look at the jobs JD's that are there today, they do very different. Forget the traditional jobs of banking and all of that, right? There are Youtubers and there are, I don't know, GNZ, there's all kind of stuff, right? There are a lot of jobs within AI. Now to run this AI, this platforms is not easy. You require a different kind of a mindset and a different kind of IQ to run it, right? So we would need people to do that. There is a positive talent right now. Lot of the companies I can't scale because there is no talent. This whole plethora of jobs that are lying out there, but we don't have people to fill them up, right? And similarly, there are a lot of other jobs that will get created. But you're right that the ABCD jobs got eliminated and that E&F&G will also start getting eliminated. But you will have something else getting added to the horizon. And that's how I think. I think this debate is endless. And we've been having ever since horsemen were replaced by cars. But then what? People don't realize that those horsemen ended up becoming drivers and newer and newer jobs kept evolving. And of course, so this debate will go on. There is no end to it. We'll find out what AI ends up. I think we sorry, I think the Hollywood makes it worse, right, Because they'll show that doomsday. Yeah, taking more so. So we all have seen the dark side of and and the tech billionaires also sort of keep adding fuel to the fire. I mean, Musk will say something today which is entirely apocalyptic and everybody would sort of, you know, start worrying about those things. But well, that is the nature of technology that we're talking about, the kind of impact it can have. So all I'll say in the end, it was it was a spectacular sort of, you know, set of panelists we had here. We ended up having a lot of insightful conversation. So over the next few episodes, we'll try and address all questions you might have related to AI. So huge round of applause audience for the stellar cellar speaker we had here. And to everybody watching this on YouTube, on men's website, thank you so much for watching. Thank you.