TCS COIN™ podcast series Episode: 5
Unlocking data, the intelligent way
Want to know how to make sense of complex data? Listen in.
in this episode
What’s a structured way to tackle unstructured data? Find out in this episode of the Next Big Think! as our our host and TCS futurist Kevin Benedict and Amit Jnagal, CEO and Founder of TCS COIN™ Accelerator start-up Infrrd, take you through a journey of acronyms, automation, and much more.
Amit demystifies ML, IDP, and RPA and explains how they come together at Infrrd to automate the process of reading documents and extract value. Amit elucidates Infrrd’s technology with a mortgage industry use case.
A believer in ‘dreams are work in progress’, Amit maintains that we are moving towards a future in which intelligent processing and robotic capabilities come together to create RPA 2.0 or Intelligent Work. Time to tune in and connect the dots!
Amit Jnagal, CEO and Founder of Infrrd
Amit Jnagal is the CEO and Founder of Infrrd Inc, a new generation artificial intelligence/machine learning (AI/ML)-powered startup that enables intelligent work. He is passionate about solving complex business problems involving automation and analytics. An AI-enablement visionary and author, he is responsible for overall business strategy, growth, and culture at Infrrd. He has nearly 20 years of multi-market experience helping more than one hundred organizations achieve measurable top-line and bottom-line impact through the innovative application of technology.
Kevin Benedict: Welcome to The Next Big Think! In this podcast, we give a shout out to the future. I'm your host, Kevin Benedict, a Partner and Futurist here at TCS. And I want to thank each of you for listening. Our guest today is Amit Jnagal. He's the CEO and Founder of Infrrd, Amit thanks for joining us today.
Amit Jnagal: Thanks Kevin, thanks for having me.
Kevin Benedict: Your company is called Infrrd. INFRRD. No vowels. So, you gotta leave that up to our imagination. But how do you pronounce it? Now talk to us about as an entrepreneur with a series of startups under your belt. What's the process of naming a startup company? Walk us through kind of how that evolves.
Amit Jnagal: Right. So, when I started my first company, I remember the name that we pulled out was from a startup name generator, there's an app. For that we just put random birds together to see, you know what domain names are available. But over time, I've started getting a lot of deeper meaning to the name of the company and what it stands for. So, our name Infrrd, the root word is inference, a lot of work that we do is something pulling out insights that you just cannot pull out plainly, right, it's not like two plus two is always four, we look at, you know, a is equal to b and b is equal to c, then most likely a is equal to c in a simplified version. So, a lot of stuff that we do, totally depends on making inferences from large volume of data. And that's where the name comes from. But I've had, you know, some names, which were even better, but the domain names are not available. So that is also a big factor, usually. But dropping the vowels was again, one of the interesting things that one of our advisors told us to do at that time, Tumblr had a very big exit with Yahoo, I think, and they did not have any vowels. So, we dropped the vowels. But we've also it has also backfired in some cases where, you know, some of our prospects called call us the infrared instead of Infrrd. It's a matter of time, once you become a brand, you know, everybody will know how to pronounce Infrrd. But that's a story.
Kevin Benedict: Oh, that's so much fun. Thank you for sharing that, it's good. Everybody has a different story. Some people, you know, they try to rewrite history because they have some name somebody in marketing created, then they go back and try to create a legend around it. But this is just awesome, where you just walk us through how the process works, from beginning to end. So, thank you for doing that. So, when I looked at your bio, Amit, you're a true entrepreneur, you've started all kinds of different companies, what experiences and lessons did you learn in earlier startups that you think is gonna benefit Infrrd and Infrrd’s customers today?
Amit Jnagal: A lot of them over the years, right. So, I think the biggest one that I've learned is team is everything. If you have the right team, in the right place, that solves like 90% of your problems for you and your customers. Right, so some of the lessons that I've learned in my earlier companies, we've always dealt with large enterprise customers, even though we were small companies, our target has always been large firms. So just a way of how to get a startup to work with large companies, right, we figured that out to a large extent. So, our customers, when they work with us, they get the best of both worlds, right? They get the… the product and the knowledge that comes from really large companies, but the agility of a really small firm that knows how to move quickly and start showing benefits to our customers. We've also over the years, right I've realized that it pays off a lot if your customer-focused and investor-aware rather than the other way around. I have a lot of my friends have companies that are totally investor-focused and customer-aware where it can wreak havoc during the initial years right cause a lot of confusion. So that's one thing that we are very clear about your customer-focused and investor-aware rather than being the other way around. And that has worked well for us and for our customers too.
Kevin Benedict: That's… that's really interesting. Here at TCS, we have a big emphasis today on stakeholder value. So, it's not just about serving the investors or the shareholders. It's about taking care of your employees. It's about taking care of your customers and shareholders as part of those stakeholder ecosystem as well, but not trying to over emphasize our investors, because that again can steer you wrong, so great appreciation for that. So, Amit you're a CEO and a founder, you've done this a number of times, tell us from your experience, what's the best and worst parts about being a CEO and founder?
Amit Jnagal: It's a very long list, Kevin, for both sides. So, when anybody starts a company that is my first-hand experience, as well. And I think there is a lot of literature that is written on this, you start a company, because you're good at something, right? You start a furniture shop, because you're really good at making furniture. But as you become an entrepreneur, you realize that the job that you love was the last thing that the business will let you do, you now have to evolve into businessman and actually hire somebody else to do the job that you really love. Right? So, I started my entrepreneurial journey because I was a really good coder. Right? I love solving problems that others said, cannot be solved. Right? So, there has been a, it's been a fun ride, being a CEO and founder, right? So, you get to have a lot of accountability at the end of the day, right? So, your performance, how your year was, how did you do your job, as you know, very clearly in front of you, the company has grown, it means you did something right. If you are struggling, which means you are you have not done a good job as a CEO, right. That's very different than being in a job where you depend on somebody else's opinion to tell you how well you have done in a year. This is, you know, like, very brutal, it's like looking in the mirror, right? The scorecard is right in front of you. There is also a lot of freedom, that's the other side of the coin. So, you get to take decisions, but you also get to live with them. Right? If they work out great, if they don't work out, they can blow up in your face. Right. So that's the part, the not-so-great part about being a CEO is, it's a fairly lonely job, right? You have challenges, you have a lot of stuff that you want to discuss with somebody. And for the longest time, you know, I tried to discuss it with my family, they had nothing but empathy for me, they didn't understand my world, I tried to discuss it with my friends, they also had no appreciation of my world, right, they are not in a similar job elsewhere. And eventually, you know, I formed a very solid board of advisors around me, who I used to when to take guidance, you know, how to express how I really feel on certain days. So that has been very, very helpful. Without that, I think it would have been 10 times more stressful than doing this totally on your own. The other thing that I really saw a difference from being an employee to an entrepreneur, right was when you're in a job you're responsible for… usually responsible for one function. Right? If you are running a product, then you don't worry about where the people are going to come from you don't worry about where the money is gonna come from you don't worry about how people will find us, right. But as a CEO, everything is your responsibility, right? And you have to be a stay on top of everything, which is a big stepping stone for me when I moved from my corporate job to an entrepreneur’s shoes.
Kevin Benedict: You know, a man with your experiences there, mirror my own. I spent five years as the CEO of a startup software company, and same very similar experiences, both on the good and the bad. It's a tough job. So yeah, so that kind of leads into my next question here, which is what motivates you to go through all that?
Amit Jnagal: Well, it's my life's dream, right? This is something that I wanted to, this is something that I think the world needs, right? And so if you ever received an email from me, my email always has this signature line at the bottom—dreams are work in progress—which I got from a discussion with one of my friends who told me that when you want to do something, the first step is to dream, right, that you are going to do it and then everything else, you start working on making that dream a reality. So, and we've got an amazing team, right? Look, seeing them go through a lot of hard work, trying to, you know, make a dent in the universe, is what keeps me going even during the toughest days, right. And we've got a beautiful product at the end of, you know, five, six years that you've invested in it. So that's what keeps you going. You see the delight on customers’ faces when they see the kind of value that you deliver for them the kind of difference that you've made in their world. There's no bigger fuel than seeing a happy customer and a team that is proud of what they've achieved.
Kevin Benedict: So, there's a phrase, you know, what comes first the chicken or the egg? Or that's a question, I guess what comes first the chicken or the egg? So, what came first, in a similar line? What came first, the recognition of a particular problem that you could solve, or your dream to just start up a technology company, and then find the problem later? Which one of those happened first?
Amit Jnagal: Yeah, so the latter part happened first dream of starting something happened first for me, and Infrrd is a like a pivot from one of my earlier experiments. So, we did not, when we started, we did not start out to become an ITP company, we were actually a retail analytics company. And we basically pulled out insights from unstructured data for retailers, primarily from product reviews and stuff. And that finally, morphed into Infrrd, where we started making sense of a lot of complex documents for large enterprises. But we did not set out to build Infrrd, we set out to do something really awesome with unstructured data, which was very difficult to tame in the beginning. So, in our case, the… we had a team who wanted to work together and make a dent in the universe, we started with one market and one approach, that didn't pan out, as we expected it to, so we morphed that into Infrrd. So that's how we landed here.
Kevin Benedict: Interesting. And before we get deeper into kind of the problem space that you guys have jumped into let's do a kind of an acronym review. So, we're gonna talk about RPA, AI, ML, IDP. So, there's four of those at least, why don't you walk us through what each of those means to you and how they all work together inside Infrrd?
Amit Jnagal: Sure. So let me take the AI and ML part first, because they I think, belong together in a separate space, and then RPA, IDP maybe. So, machine learning is a totally new way of having algorithms learned from large amount of data, right, and then make predictions based on that data. So, in one sense, it's like glorified, glorified mathematics, probability, and all of that good stuff that goes into building models. But machine learning is a fundamental unit that offers algorithms that can learn from data. Now, AI is much broader than machine learning, right? AI is intelligence, right? It's a, so if you, Kevin have a game on your mobile that you play, from time to time, right? If I were to give it to a true AI algorithm, it'll learn if you just tell it the rules, it will figure out how to play that game and keep getting better every time it plays. In order to do that, it has to figure out what to learn from what not to learn from, what old lessons to discard what new lessons to learn. So, all of that is AI, right. And AI uses ML in a lot of different variations for it to do its job. So that's one part of the world. Now the second part RPA is a technology has been around for a while. But it's so… RPA is more about automating stuff that traditionally needed human beings to do something. So, you know, take data from someplace, click here, do this, read this and send an email to somebody and stuff like that. So that's a basic automation of something that needed very low-level cognitive skills in the past. IDP is a new dimension on top of RPA, which is about understanding that. So, a more simplified way to say that, as you know, IDP is the brain and RPA is the arms and legs that make the automation work, right. So intelligent document processing, the field that input plays in is about getting insights from documents, give us a document, we try to answer questions around that document. Once you have the answers, then you figure out what you need to do next with those answers. You know, I process an insurance claim the amount is less than $100. What do I do with it? Right? The amount is more than $5,000. What do I do with that? Right? So, that's the whole equation where RPA comes in, right? Once you get the data, or the whole workflow and orchestration of how to move that data and what to do based on that data comes from RPA.
Kevin Benedict: Thank you, man. That was one of the most concise explanations I could have imagined. So, good job, and thank you for doing that. So, I also when I'm going through your website and reading all about your content, you talk about machine learning and deep learning. For those of us who are uninitiated, can you help us understand the difference between machine learning and deep learning?
Amit Jnagal: Yep, sure. So deep learning is a very specific branch of machine learning, which uses multiple layers of neural networks to extract insights from data, right, deep learning typically needs a lot more data than a machine learning needs, right? So, on a simplified version, if you have visited a website that an e-commerce website which where you bought a product, and they make a recommendation to you and say, most people who bought this also bought this, right that doesn't need very deep learning. It's usually based on statistical models. And that's something that machine learning technologies can handle. But well, the more you deal with the data that is very visual in nature, or need deeper insights that when you go, that's when you go to a deep-learning-based ML approach, which has multiple layers, right. So, in our case, for example, if you look at a document, and there are a lot of letters written on the document, ABCD, right. Normally, if I use a machine learning approach, I will have to collect image of all the A's, right and target and say this is A, give it to a machine learning algorithm for it to understand that this is an A, right. But for a deep learning algorithm, I can collect, you know millions of images of ABCD, all the letters of the alphabet and give it to a deep learning algorithm. Depending on what features it can extract out of it right on the first layer, it might separate all the pointy ones in one bucket, right, and say A and E and you know, something else looks similar. They're like a triangle. And D, and O and D have a big oval in them. So, they look similar, right? And then we'll keep doing that extracting features out of that and in different layers and make it finer and finer. So that works beautifully when the data is huge. The amount of insights that you want to extract are huge, right? But deep learning is just a very specialized branch of machine learning. We use both of them for different aspects of document understanding and input.
Kevin Benedict: So, let's talk about use cases here. What are some typical use cases that you guys get brought in to solve?
Amit Jnagal: Yeah, so we work with primarily financial services companies that have large team of people who read paper, a paper can be in physical form, or a PDF file and email, right that they need to process to get information out of that and into a CRM system and an ERP system for that system to work. So, if you're a mortgage company, for example, and you deal with lot of mortgage application document, and people give you their W2 forms, their pay stubs, their bank statement, account holdings, and you need to now extract data about, you know, what is the wealth of this person? How much does he earn? Typically, this… all this will go to a large team of people who would just read these documents and try to, you know, extract these values out of these documents, right, we try to automate all of that process from getting the data, chopping it into smaller pieces, and then extracting data from each of these documents also understanding what document type they are. And most of the cases where we have read the document clearly, and we are fairly confident that we read it right, we just pass it on to the downstream system to stay through processing, where we are not sure we need a pair of eyes to look at what we have extracted and tell us if we are right or wrong, we put it in a queue that goes to a human being like a team of people to look at that data. And they can either accept the prediction or reject it. Every time they rejected, we learn from that and try to see how to not make that same mistake again, automatically. That's where all of our machine learning comes in. So, over time, we try to make sure that the human environment is lesser and lesser, and more documents go through straight-through processing, and visitors save a lot of time doing that.
Kevin Benedict: So, are the motivations of your customers primarily cost reduction? Or, you know, I can think of maybe an improved accuracy, improved speed of processing? Are there more reasons that a company will come to you?
Amit Jnagal: Actually, so that's what was our going in understanding into this market that this is going to be a cost play, right? How much money can you save me, but surprisingly, most of the companies that we deal with, their challenges scale, and not necessarily the cost, scale, and SLA, right, they want to be faster. But primarily they want to handle a lot more data. So, some of our customers, you know, they've had such a large data processing team that it has become really inefficient for them to add more people on top right. We had this one customer for whom we processed like around two and a half million variations of documents, right. And they told us that when they get a new person in the door, it takes them nine months for that person to see all the documents in the system and come to a place where their data cannot be… not be reviewed. Right? It's good to be processed after they're done. And that's a very long, long lead time for them if they get a new very new large customer, right? Instead of celebrating, they start worrying out… worrying about how to scale, how to get these people up to speed really fast. So, what we do for them, as you know the servers don’t sleep, right, and you can scale them very easily. So, we do most of this data processing automatically and give them this ability to scale their business. The second thing is, for most of the customers that we work with time on how fast they can turn around documents, is also extremely critical for them. Like in the mortgage business, the SLA, right on how fast can you turn around the mortgage application? Has a big influence on whether you'll be able to get that customer or not. Right? And that you can only do through automation. Human processes still take a long time. But yeah, surprisingly, the motivation is not cost saving, as much as it is on how to scale the business and how to become more competitive by being faster.
Kevin Benedict: That's so interesting. Let me let me extend this question. Have you seen revenue? So, all of these things, so scaling, speed, probably cost savings accuracy, in the mix somewhere? Have you ever seen revenue-generating opportunities from what you guys do?
Amit Jnagal: All the time, all the time. So, we have, you know, some customers who had so much data to process that they had felt those projects, right, the math just did not work. We had a customer for whom we processed like a huge volume of mortgage documents to extract some insights. And they, when they did the math before, with an absence of a solution like ours, it did not make any sense to offer that service to their customers, they would be spending a lot more. And they knew that the customers appetite was not there to spend as much to get that service. Now, with automation, that project is all of a sudden become a possibility that you can extract data from those documents at a very less cost. And because you can do that you can now offer a new service to your customers. You know, that is a new source of revenue for you. So, we see it in both ways, right so a lot of companies who use our platform as a competitive differentiator, they end up getting more customers that they did not have access to before. So that's also, you know, cause of getting more revenue for them. But yes, we've seen a lot of options of revenue generation too.
Kevin Benedict: Ah, that's brilliant, that is brilliant, when you can play both sides of that equation, that it just makes it so much easier for your sales teams. And for you guys just to market your products. Very interesting. So, what do you think are going to be your biggest challenges over the next three years? You don't work in a vacuum, there's a lot of competition. There's a lot of innovations, a lot of development around AI and machine learning? What's going to be your company's inference, big challenges over the next three years?
Amit Jnagal: Yeah. So, to be fair, we are in a market that is just started to peak. It's a fairly young market IDP, compared to the rest of the markets. So, for us, the single biggest factor that we keep an eye on is our speed of execution. As we scale, you know, how do we get more customers in the door and make sure we still maintain the quality of customer success that we've had over the years, right. So, speed of execution is the number one thing that's on our mind. The second part is, you know, as you work on in IDP, there are a lot of new possibilities that come to you as potential opportunities. A lot of them at times are distractions, which look like opportunity prima facie. So, you need to figure out distractions from opportunities, right, and not lose focus on what you're doing. That's also something that we are very aware of. And the last thing that you mentioned is the whole AI ML field is moving very fast. And it if you don't keep on your toes, it comes very easy for you to be replaced by somebody whose AI is more advanced than yours. So, we... the way we have tackled the challenges, we've got a dedicated research team that does not work on our product. They just work on problems that are of paramount importance to our customers. But no solution exists for them today. Right? They are also on top of all of the new advancements on AI and machine learning algorithm. In fact, our team just won a competition last week, on understanding content from data from documents. So that's something that we invest in quite a lot to make sure our AI is not outdated. But yeah, these… these three challenges are the biggest one in my mind, Kevin.
Kevin Benedict: So, have you seen this? So, we just talked about kind of what you see are challenges for Infrrd, let's shift over and ask you to put on your future’s hat if we can, and look forward five years, what's going to be different? How is this solution space going to have evolved over the next five years? What's your best prediction?
Amit Jnagal: Yeah. So, before I look into the future, let me take a look like 20 years ago, 30 years ago, okay. When I was a kid, right, I had some of my family members who used to work at banks. And when they went about their job, there was no computer. They everything they did, they did it with a ledger and a pen and, you know, cash counting with their hand. Today, all the banks have computers, right? If you were to, if I were to go back in time and tell my uncle's right, that 10-20 years from now, most of the work that you guys do, is going to be overtaken by a computer, but you will still be working with computers, right? They will not believe me. Right? They wouldn't have taken me seriously. We are going through a similar phase with the AI and automation today. Right? So, a lot of the stuff that is low level cognitive jobs, right, going to be going to get automated. It's a... you know, you can imagine the time that you I'm sure, it's very fresh in my mind. Right time and no banks at ATMs. Right. But now ATMs have become like a de facto thing if you're starting a bank or have to have any team automation is going through the same stuff, right? Where we are focused right now is document understanding, like… like you said, it's the brain of cognitive thinking, right? But RP companies are focused on the arms and legs, right? Once you understand what this content means and what do you do with it? I think the future in the next three to five years is combination of both of these technologies, how they come together, you can call it RPA 2.0 or 3.0. We call it intelligent work, right? When intelligent document processing and the robotic capabilities come together, you actually have some semblance of intelligent work, right? So, some work that is done. For every vertical from what they processing, they do get a document is not only able to understand those documents, also figure out what is missing and take action based on the missing data and talk to your customers automatically and figure out what they need to do and what they've missed. So, I think that's where the world is going. We're very excited to be on that path as well. But the combination of intelligence and actions is where intelligent work will come from.
Kevin Benedict: Yeah, that's a fascinating, both reveal of the past and where we're going to be in the future. When you're talking about, you know, a number of a couple decades ago in banking, that brought up this vivid memory I had, we took a vacation down to Costa Rica, we walked into a bank, and we were doing a transaction and the gentleman behind the counter had this giant book. Yeah. And he was doing bookkeeping. When I asked for cash. He wrote it down with a pen or a pencil, I guess it would be on this giant book that was like three foot by two foot. And just he was doing it all by hand. And the other memory I have is he had a like a western style Colt 45 pistol on his hip. Yeah. In the giant book with a pencil, and that's how he did transaction now that was 25 years ago. So, things have all changed, I'm sure. But thank you for bringing up that memory as well.
Amit Jnagal: Yeah. So, I, Kevin, had written a book a couple of years ago. And the way I opened it… it was about how AI is going to change how businesses work. So, I started with a story about my son, who was, I think was six years old at that time. And like most boys, right, he was fascinated with planes and wanted to become a pilot. And I had written in my book that I'm not sure by the time he grows up, will we actually really need a pilot? Or will the planes fly? Totally on their own right without the pilot?
Kevin Benedict: Oh, yeah, the world is moving fast. Hey, man, I want to thank you so much for taking time out of your busy and entrepreneurial day and life to just share your insights with us today. So, thank you so much.
Amit Jnagal: Thanks, Kevin. Pleasure is all mine.
Kevin Benedict: So, guys, if you're looking for any of these things we talked about Infrrd - INFRRD. Thanks again Amit.