How Agentic AI Will Impact Business

In this episode of Data Analytics Chat, we welcome Sundip Gorai, Global Head of Data Science and AI at Global Payments, a Fortune 500 company. Sundip shares insights from his expansive career journey, which spans across various industries like banking and manufacturing, and discusses how agentic AI is revolutionising business. 

He offers insight into AI strategy, the challenges of adopting AI, and the importance of striking a balance between technology and ethics and social responsibility. Sundip also reflects on the future of work, the impact of AI on jobs, and the need for continuous learning in this rapidly evolving field.

00:00 Introduction and Initial Thoughts on Leadership
01:04 Welcome to Data Analytics Chat
02:18 Sundip Gorai's Career Journey
08:11 Defining Moments and Challenges in Career
17:17 Leadership Skills and Attributes
19:47 Meditation and Stress Management
24:10 Approaching AI Strategy
28:54 Choosing the Right Tools for AI Strategy
29:04 The Unchanging Layers of AI Strategy
30:01 The Obsession with Tools and Technologies
31:07 Understanding Agent AI
34:07 Industries Adopting Agent AI
39:01 Challenges in Deploying AI
43:15 Impact of AI on Jobs and Skills
49:10 Future of Business and Mankind with AI

Thank you for listening!

Transcript

sundip gorai: 0:00

I probably never wanted to be a leader. Let me just clarify that. What I wanted to do was do good work. Being a leader is not a very easy job in a space like data and data science and ai because the resources are sharp, intelligent, and people have their own desires, intense and very smart people. Get impatient, fast, and then you have to take care of them at every point in time. I think obsolescence is the only constant in this marketplace, right? We need to adopt and change to the changing times and figure out to survive and build ourselves in this new ecosystem of change because all these processes are going to be attacked by agents and ai. Great power comes great responsibility, right? It was, I think, said by Rose Walter and then Spiderman, right? And that is exactly the problem. We really need to see that just because something is powerful and it can do it, should we do it?

ben parker: 1:06

Welcome back to Data Analytics Chat, the podcast where we discuss the world of data, AI and the careers shaping it. Today I'm excited to welcome Sun Bit Gore, global head of data Science and AI at Global Payments fortune 500 Company. So in today's episode, we'll explore its exciting career journey and then discuss the data topic on how agentic AI will impact business Sunday. Welcome to the podcast.

sundip gorai: 1:37

Thank you. Thank you for inviting me, Ben. It's a pleasure.

ben parker: 1:41

Brilliant. And I'm looking forward to today's topic'cause it's making a lot of headlines.

sundip gorai: 1:45

Yeah, of course it is making a lot of headlines.

ben parker: 1:49

So obviously firstly, congratulations on your new role, and I hope that obviously starts out well for you. And obviously I, they're large in the transaction space. They, do I say trillions of transactions, is that correct?

sundip gorai: 2:04

I think they are one of the largest transaction processors. Global Payments is processes, transactions across 227 countries 4 trillion transactions annually.

ben parker: 2:15

Amazing. So it'd be interesting work for yourself. So do you wanna just give us a quick introduction to your new role?

sundip gorai: 2:22

Sure. And of course I will keep it at a very high level at this point in time, but it'll give a fair perspective of what. Possibly I'll be going to do in this new role. So essentially my role will be defining and driving the vision of AI and data science for global payments cutting across their various businesses. And as you can understand, a company that processes so many large transactions, it cuts across pretty much everything you can think of. Retail education, parking, banking, anything that cuts across all these transactions, making sense of these transactions, keeping the aspect of regulations in mind, keeping the aspect of social responsibility in mind. So that is one. So monetization of this data set and scaling up, because scaling is going to be a very big thing. Security is going to be a very big thing. So from the value generation, monetization. Use case articulation to build up of the teams, defining the AI data science, architecture tools, technology data sets, and aligning very closely to the data engineering vision. Because I think all this finally rests on the data and the data quality based on which AI and data science will happen.

ben parker: 3:38

So Sounds, yeah, you'll be busy for the rest of the

sundip gorai: 3:41

Yeah. Yeah.

ben parker: 3:43

Brilliant. And so obviously you've had a fascinating career. Obviously you've worked in multiple locations, you've done startup, you've done corporate. Do you wanna give a, give the listeners explanation of your career journey?

sundip gorai: 3:58

Sure. And by some strange quirk of it. I would say that I have been fortunate enough to basically embrace the data journey as it has evolved. More or less from, pretty much from the inception, very close to the inception the earliest data warehouses started coming up in around 1996 as we know it today, and I embarked on this journey around 1998. So pretty much I have been through the whole journey of evolution. So I in the recent past I was A-S-V-P-N-W-N-S, which is which is a company which was started by British Airways, which got, just got acquired and is in the acquisition processes around. With Capgemini, my role cut across multiple ver verticals working with CIO CDOs, defining, articulating their vision. And as you mentioned, prior to this, I had ventured into building a data company with a Israel based firm. And I think because of the Israel war I exited and then. Took up this role. Prior to that, I had done very similar roles of basically nurturing, incubating data practices in some of the large billion plus organizations, companies like Go Forge, companies like Genack, companies like l and t and Infotech Hexa, where all of these companies are typically range somewhere between a couple of billion to 20 odd billion dollars. So most of these companies I have started the data. Foundation, building the data teams and creating the offerings solutions, building the partnerships, laying the foundation of multiple architectures doing data science. And of course I also was with IBM and Oracle before that, with IBMI was one of the leading Players in that Watson space When Watson came out as of the first integrated data science platform, I know in Oracle I was part of the risk analytics team had a good fortune of working across 30 odd countries, doing lots and lots of work across banks, insurance, travel, transportation, hospitality, life sciences, advertising, oil and gas, and more.

ben parker: 6:01

Brilliant. It's quite a journey and broad experience, which I think hopefully adds to your credentials.

sundip gorai: 6:08

My credentials means for education wise. I studied from IRT Kaur, which is one of the premium institutions in India, and I did my MBA from amba in marketing and. Finance, which is also in India and all these across, I have been the roles that have played has been the head of data science, head of data engineering, head of data engineering, data science, and ai.

ben parker: 6:34

Cool. And how have you found moving countries? Because obviously it's Yeah, obviously different. You've come from, yeah. You've gone from one side of the world to the other, so

sundip gorai: 6:44

Very interesting question. Yeah, very interesting question. Of course. If you see that in the early parts of the career, I was very much resident in India. That is around 20 odd years ago, and I used to travel a lot all across, and I had an enormous. Embracing of cultures. One time in Africa, one time in Middle East, one time in Latin America, one time in Europe. And it was a boiling cold run of knowing people, understanding people and embracing that. So when I, moving to the United States was not that hard that way because I had been constantly be in touch with the cultures, and of course there are, over time you learn the subtle nuances of a particular culture and space. So likewise has been. I am based out Atlanta, which is a lovely city, and that's where I have been for the last 20 odd years.

ben parker: 7:37

Brilliant. No, that's good. I think it's, I think it does add to your I think it adds to your life experience when you, I've done same move. Not as far as I moved from England to Spain. So obviously when you go to different cultures, I think it does add to your, you obviously you learn and you meet new people. I think it adds to your yourself.

sundip gorai: 7:56

very. You moved from England to Spain and I understand that's a very different but similar experience like me, right? You're embracing different cultures.

ben parker: 8:04

Yeah, exactly. Brilliant. Okay. And so then, what were, obviously you've had a great successful career today. What's been the sort of defining moments that have changed your career?

sundip gorai: 8:16

Defining, there are various points in time, and I would, again, as I said, I was fortunate when I finished my engineering, that was in 94. I did my industrial engineering and I started as a shop floor industrial engineer with Blow Plus Group, which is the group that I think they were part of Mattel Toys, the company that makes Barbies and others. So as a shop floor. Engineer. That was a good grounding in understanding manufacturing processes hands-on. And after my MBA, I started with the second largest bank in India, I-C-S-C-A bank, and pretty much everything I did running the branch operations, setting up a branch. So that was a very strong grounding in the services industry. So in early parts of the career, having been hands-on in a manufacturing industry. And having been hands-on in the services industry, that was a phenomenal, I think, founding stone for somebody who was going to do analytics in 17 to 20 industries in the rest of his career.'cause every time you could go back and feel it how it is. A services business work, how it is a manufacturing business. So that was a defining moment. The second defining moment came in when I was in Oracle, came through acquisition called Rebellious ifl. We were a risk analytics firm, and that is when I got to know, got to see a hundred odd banks and financial institutions across the globe. And till that time I had been. Working more on the technology side of AI and analytics, but first time I started understanding the business manifestation of analytics and that was a complete change in my perspective. That it is all about business. It is all about actionables. It is all about insights and technology, tools, algorithms. All these are means to an end. That was the second big defining moment. These were the two major I would say, and. After that it has more been scaling. These were the main defining moments, and from a challenges perspective, COVID was a very different experience altogether of handling everything.

ben parker: 10:10

Brilliant. Yeah, no, I think it's obviously ki obviously you've come from a technical background and then you've had to, obviously you've gotta learn the business side and then also, like obviously you've been yeah, fortunate to progress into leadership. You've gotta learn the soft skills there as well, haven't you? So it's, I guess it's, it is a massive learning curve for people that obviously want to progress in a similar situation like you.

sundip gorai: 10:32

Yeah. Yeah.

ben parker: 10:34

So what what's been the big challenges you faced along the way and Them?

sundip gorai: 10:40

Yeah, so challenges, I think there were different challenges. If you look at, let's look at the kind of projects which I have done. I'll talk through those challenges first. So one of the biggest challenges when you are in the data space right from the beginning is first is the stakeholder buy-in. Okay? That is the most important thing. Whatever projects which you are going to do for your. Stakeholders. They're understanding their pains, understanding those gains that they're trying to do, and more importantly, unifying them and arriving at a singular vision, which everybody agrees to. That is one of the biggest challenges. So understanding the stakeholder intent, bringing them to the common understanding. The second big challenge, which happens in most of these. Project is the data itself. Now, when I say data itself, one is the data science related data problem, and second is the data engineering related data problem. So let me explain what is the difference in a data science related data engineering problem? Because end of the day, whatever you are doing, if your data science algorithms are running or if you are. If you are deep learning and AI algorithms are running, they require relevant data sets. And without getting too technical or what is called the feature vectors and the relevant feature vectors need to be understood, identified. So that is a very critical challenge and finding that, creating the right feature vectors for data science project and getting it at the right place, right time with right volume, right velocity. That is one kind of an. Big challenge, which happened in data science projects. The second challenge, which happens in data science project is the model drifts, which happen over time, right? So let me give a analogy of this. For example you are served one kind of n pizza and you like it and you like it and you like it every day. But one fine day, you don't start liking the taste of that. And similarly in a data science model, what happens is that the end intention and the outcome of the data science model may change over time. So that drift has to be understood. Secondly, sometimes what happens is. One fine day. You are liking certain toppings on the pi and after a few days, you want some new toppings in the pi. Similarly, in a data science project, what happens is that the data inputs which are coming in, they may drift over time. So you may have to then essentially revalidate or recalibrate your models from that. Particular perspective. So the third thing is the velocity of the data. Sometimes you have been building models and static data. Now there may be more dynamic models required Dynamic pricing required, dynamic forecasting required so that, again, you may have to re-architect the whole model and the architecture. Secondly, on the data engineering side, the bigger challenges are common. As we say, the it is a volume related problems. Volume has been increasing over time. The architecture changes over time. It's a velocity related problem that you are having. More. I think now you have more streaming data and you need use cases around streaming data. Variety is a problem earlier, a simple structured data. Now your structured and unstructured data. And likewise, from data warehouses, we have moved to data Lake House to Delta Lakes, and all kinds of new architectures are coming in. So those were some of the challenges which we had to handle. And as the space has been evolving, the. Demands in the marketplace have been evolving. The tools in the marketplace likewise have been evolving, and all this therefore means then the people like me and others who have been associated in the space have to essentially realign our thinking, design vision for people, process and technology to the changing times.

ben parker: 14:16

Brilliant. And yeah, and I've, I'm glad you also opened up any challenges.'cause I think it's, I think there's so much happening in the field, isn't there? The, you've got to, you've gotta con continuously adapt, haven't you? There's, and move, you've gotta keep learning. Like it's just the way it is now.

sundip gorai: 14:32

Yes, absolutely.

ben parker: 14:35

Just on yourself. So do you, do you, have you always wanted to be a leader? Say, obviously after your fresh out of your studies university is, do you always wanna be a leader or is that sort of over time you've gained confidence and experience and it's just the way it's gone?

sundip gorai: 14:54

Okay, so a small correction. I probably never wanted to be a leader. Let me just clarify that. What I wanted to do was do good work. Okay? That has been the driving factor of my journey wherever I have gained, where, wherever I have changed jobs, it is. Throughout, I think out 80% of the time, it is the content and the nature of the job that has magnetized to me to change a job. Not any other primary reason, but one of the main reasons I want to do something new in this space. When I wanted to build a data startup, that was the entire, when I joined IBM, because it was one of the forefront of the big data when I joined each of the companies that I have worked to build a data practice. When I joined Oracle, it was to learn banking, analytics, and merit. So I would say my driving factor. Has been learning. Always driving Factor has been to learn new things in the evolving space and that way I think Destiny has been fortunate that I have grown along the process. That said, leadership. As you move over time and as you take new responsibilities, you learn the skills of leadership. You breed the skills of leadership and being a leader is not a very easy job in a space like data and data science and ai because the resources are sharp, intelligent, and people have their own desires, intense and very smart people. Get impatient, fast, and then you have to take care of them at every point in time. Like in IBM, when my team was like around, if I remember around 40 odd people, many of them were PhDs and some of them are from Stanford, MIT. So you need to take care of people to ensure that they feel valued, respected, and essentially are doing good things. So leadership comes with its own set of I think greatness, joy, and challenges as well.

ben parker: 16:42

Yeah, no, definitely. Leadership is. You've come from, you're being a technical expert, and then you've gotta deal with people. And e everyone is different. Everyone's got, everyone's in their own little bubble. Everyone's going through their journey. So it's, and obviously you've got, obviously managing that is, yeah, it's obviously there's lots to do, isn't there? And also you've got to continue this keep up with the. The learning that's especially in AI field now, it's just, yeah, every new week there seems like something new happening.

sundip gorai: 17:15

Absolutely. Absolutely.

ben parker: 17:17

So what, what skills or attributes do you think has helped progress you into like leadership?

sundip gorai: 17:24

So I would define the skills primarily. There is one is the different layers I would call it. One layer is first on the pure core content of the job is of course having strong business skills, understanding business acumen of whatever I have been doing that has been my strongest differentiator even today in the marketplace. If you see that there are just a handful of people who straddle both the business and the technology side of data science, data engineering, together very few people who understand both very well, you will have people better than me. In understanding business, you'll have people better than me in understanding technology and data science, but there are very few people in the market who have the combination of skills with business and technology, so that is one of the driving forces of leadership. Second driving force of leadership is of course you need to hone your communication skills, storytelling skills very important that building great data science models. But if you're not able to communicate in a simple layman's language, then the adoption is not going to happen. Senior leadership is not going to buy in what you're going to say. The third leadership skill is of course, being courageous, straightforward, ethical. These are some of the very important aspects of life. And most of that value system, I believe, has come from my meditation practice and a deep meditator. And that has really helped me I, by those aspects of the practice. And last but most important is decision making. With, in in the timeframe of ambiguity, right? Again and again, you'll be coming across crossroads where you need to take a decision and the one of the simplest metrics which I use to take a decision is that speed over quality or quality over speed. Okay? So many times I think the situation, you will come again and again in, in projects and you're multitasking and then something that is. Time pressure for deliverable. And that is the conscious thing that one needs to think that MI is quality is more important than quantity at this point in time. Or quantity is more important than quality, given what the stakeholder wants, given all other parameters. And that is one of the thing which comes with practice over time. That what do you need to choose quality or quantity or quality?

ben parker: 19:46

Brilliant. And then I do like where you mentioned about meditation, because I think it's, I think it's key like adding that to your daily routine, like whether it's. Doing exercise, reading, meditation, going for life. Just I think it does add to your focus.'cause obviously we are in crazy world where things a hundred miles per hour. I think sometimes you just do need to slow down and just have a step away.

sundip gorai: 20:16

Yeah. I think, see that is something, of course, I wouldn't say that I'm a pro. I have been trying for now, last 20 years doing various kinds of meditation. Now I follow a specific form of meditation technique known as ana. You can check the website. mar.org, which is there in 200 countries, a Buddhist form of meditation, very simple breath observation, meditation, but for in general, anybody else, once hour, if you can take a 32nd break and just observe your breath, the extremities of your breath as it goes in and goes out without changing it, that will do an enormous. Change in the stress levels of people. I've seen it myself. Last thing, you go before going to sleep. One minute you observe your breath without changing it. First thing you do is when you wake up one minute you observe your breath, you will see the change in yourself. And stress management becomes. Very easy, and this is a whole topic in itself, the intersection and understanding of AI and meditation. Maybe some other day I will talk and have a detailed conversation. How is AI related to meditation? How meditation. I've done a lot of thinking on this piece, and I think a whole POD podcast is needed. We'll cover that sometime

ben parker: 21:21

Yeah. No it's interesting you mentioned about F Ana'cause I, obviously there's, they do the 10 day retreat. But again, that's obviously we, that's a different topic altogether, but I think it's fascinating.

sundip gorai: 21:33

I have done five, five times of the 10 day retreat, so it is really

ben parker: 21:37

Oh, amazing. Well done. And then obviously, yeah. Cool. Alright. Getting this sidetracked there, but, and also I liked how you mentioned your, you've got the balance of business and the technical I'll call it the glue.'Cause I think businesses do struggle to, especially leaders, having that blend of skills. I think that is a massive strength to have in today's market.

sundip gorai: 22:00

Yeah. Yeah, I agree. I agree on that.

ben parker: 22:03

Okay. Brilliant. What advice would you give to someone, aspiring to become a top leader like yourself?

sundip gorai: 22:10

See first and foremost in the data space. That one, I personally believe that one has to be a voracious learner. If in this space, the space is changing so fast, right? I think obsolescence is the only constant in this marketplace, right? So I think somebody who was knowing something in the data space 10 years ago and has not invited anything new. Is a dinosaur in this space today. Somebody who had IBI even something three years ago and has stopped learning is also a dinosaur today because everything that came out in the last three years, generative ai, agent T, ai, and all the bunch of stuff which has been coming, the large language models was not there three to four years ago. So I think learning is an essential thing, but at the same time, I think we are. Information is out to get us. That is the way I put it. Philip k Dick is one of the greatest science fiction writers of all times who has written movies like Blade Runner Next Minority Report, and many more things. He mentioned this word Kipp. I think that was maybe some 60 odd years ago. KIPP is all the useless junk, which is trying to probably get into our head, and it is not, we trying to get information in, information is trying to get us, so we have to be very careful of what we want to take in and what we don't want to take in. Okay. What is the grain from the shaft? That clarity should be very much there because so much of information bombardment is there. So navigating through the space and distilling the grain from the shaft, that is a skill in itself.

ben parker: 23:47

Yeah, no, definitely. I think it's so much dis even things like I, for myself, I just, when I'm doing my deep work, I just, I close my inbox.'cause if you're constantly getting an email and looking at it, it's like a stop start. I think you just need like focus and find ways that can help you. Obviously just get ahead and. okay, cool. So the data topic we're gonna discuss today then is around like how a agen AI will impact business. So I guess to start off with, how would you go approaching AI strategy?

sundip gorai: 24:24

Yeah, very good question. I think see today, first of all, AI means different things to different people. In my mind, the simplest explanation of understanding of AI is effectively anything that does sensorial acceleration. When you say sensorial acceleration, what I mean is if you see, can you see, use a technology device to see better? Can you hear? Can you use a technology device to hear better? If you speak, can you do a technology device to speak better, smell better, touch better, or translate one sensorial activity to another, let's say converting speech to text or some other form to some, another form. That is the primary driver of the AI technologies. Now, how they enable various business processes or day-to-day living processes. That is what we call that. AI adoption. Now in the business world, in the enterprise world, if you see the, there is always a mixture. Earlier, I think data science was seen as a different world. AI was seen as a different world today. What was data science is called ai. What is AI is called data science. So the way I put it is there is those use cases which require pure AI skills like speech, text, vision, mainly deep learning related kind of skills. That is one kind of an. Business ai. The other side of AI is the technical ai, which is things like AI ops or infrastructure observability or ML ops and stuff like that. Then there are business use cases or the data science use cases, which are also today getting called as the ai. And of course, later on comes the generative AI and agent ai, which are new things in this space. Now, as far as. AI adoption is concerned. My approach has been very simple, and it is so simple. Anybody, it may appear as sheer common sense, but the whole space of AI boils down to this. There are people, there are personas and companies who have pains who want certain gains. That's point number one. Those personas that need outcomes. Point number two, outcomes are of two types, right? Let's call it 2.1, which is people want to increase revenue, reduce cost, or increase profits, or 2.2 outcomes could be related to speed to market, innovations, regulations, more qualitative in nature. So you have personas, you have outcomes. Then you have the business itself, which is three. That business could be a bank, looks at credit, a bank looks at finance, a bank, look at customer acquisition. A bank looks at something similarly airlines looks at cargo. Airlines, looks at passengers, airlines looks at bookings. Airlines looks at flights. Okay, so on and so forth. So the third of personas. Outcomes and the business in which that AI or analytics will be applied. Then comes fourth is how will you solve it or what? I simply use the word analytics. Now, what is analytics here? It could be just paper and a pencil. You have a problem. The person has a pain. It is related to a business. You have a paper and a pencil and you solve it, or you need an Excel sheet and you solve it. Or you have slightly higher descriptive analytics like Power bi, tableau, and other tools you are using to solve it. Then you have predictive analytics like machine learning, data science, which you're using it. Then you have cognitive analytics which you're using deep learning, neural networks and all that to solve it. Then you have more generative new stuff in nature law, language models to solve it. Okay? Then you have agentic, which is more agentic solutions where collaboration and all that is happening to solve it. Then you have diagnostic, right? More streaming data to solve it, right? And then finally you have prescriptive, which is more. Using advanced topics like operations, research and simulation, digital twins. So coming, I started with persona, then I went to outcome, then I went to business, then I went to analytics. Now after this fourth layer, the fifth layer is architecture. Now, how are you going to solve this? That architecture could be, again, very simple. Database, like an Excel sheet. It could be other databases, it could be a data warehouse, it could be a data lake, and a host of problems that comes with it. Data quality, master data, metadata and so on and so forth. Then from this layer, fifth layer architecture, the sixth layer is con tools. So far you see, I'm not talking of talked of any tools. I've talked about persona. I've talked about outcomes. I've talked about business. I have talked about architecture. I've talked about analytics, but no tools. Then this is where you decide what tools will do it. Do I need Google, do I need Microsoft? Do I need this and that to solve it? And the seventh layer is essentially the people and the execution team. So your whole AI strategy is just this. There is nothing outside this. The world may change in 20, 25 years I've been doing this. This has never changed. This is a constant. Okay, so AI strategy thinking, and the Lego blocks of AI strategy, one Lego block may increase or augment, but the layers are going, seven layers are going to remain. Each layer may go through a little bit of a change here and there, but even after 20 years, if you and I are having this podcast, we'll have a similar conversation.

ben parker: 29:33

Yeah, no. Fascinating to hear what you say.'cause it's funny how. People or businesses, they look at tools early in the early in the strategy when it's, yeah, again, look like you touched it on the head. You need to look at the business outcomes. And also, also you need it like ai, you need to embed within the business, isn't it?'cause it is so many different components to the business. You need to have that embedded before you

sundip gorai: 30:01

Yeah I think the sad aspect of the marketplace and the social media is the obsession with tools, technologies, architecture means, but why are we doing it? What is the need of it? Who will get value out of it? Those fundamental questions are not being asked, or sometimes I really feel that why are these not being asked? Early on in the stages, I have seen very mature and senior leaders, maybe people have, are so obsessed with what they have been doing. They want to start with what they have been doing all their life, and that is why I have seen this obsession with technology methods, maths algorithms, all that is important. But they are part, they're cogs in the wheel. They're not the wheels itself.

ben parker: 30:45

Yeah. Again, it's, yeah, like I said, it's, the fascination with the new shiny toys on the market. Wanna be the talk of town. But yeah, I guess again, you need to just end of day business is business. You need to have the fundamentals, business basics what you trying to, what's your problem, and then go from there, really. Okay. Cool. Cool, cool. So what does Agen AI mean for business today?

sundip gorai: 31:13

Yeah, see Agent ai again, agent ai also, again, a lot of confusion. Is there, what exactly is agent ai? Different people have different meanings for me. The way I put it is that agent AI is like decisions are being taken where a series of actors are in play. The series aspect is very important. It's like a relay race. One guy handing the bat to the next guy and next guy and so forth. So where the problem. Is elevates to multiplayers then that essentially is a proper agent. Take AI use case, not that a single agent can't be a agent use case. The most important aspect of the agent is one is autonomy, independence to take a decision. Second is cognition. Okay? Ability to use various kinds of data science or cognitive techniques to take a decision. Third is from an agent perspective, there has to be a goal orientation, either of a single agent or multi-agent together, they need to be solving for a particular goal. Okay? The fourth thing is that individually and collectively, there has to be learning and whatever they're doing, they learn from the mistakes in the next cycle, they're able to retain that learning, have a memory. And improve upon that mistake in the next cycle. Now, there have been workflow tools before this. If you see the low-code, no-code tools were there, but the difference was they didn't have those cognitive elements. They didn't have the large language models embedded in them. They didn't have search in them. So these newer aspects, which has come in as changing the whole manifestation of the use case completely from the past. Okay. So I think this is going to be a big thing. And if you see, I keep writing about agent care use cases. My LinkedIn, and I know time, we cannot get into details, but there are use cases and insurance use cases in banking, use cases in hr. Every business aspect is going to get touched with agent ai. And the way we are going to see it is that there's a, there is. The cost over the long run is going to come down, productivity is going to go up. And then essentially there are the core processes, agents, which a business is running, and the non-core processes that could be related to HR or finance, which are vital to the organization. Survival, but not core of the business. Airlines has a core business and a bank has a core business, but common functions like hr. Number two is finance. Number three is procurement. Number four is contact center. Number five is sales and marketing. These five functions are, they're in almost all organizations and these functions are today the way, traditionally they're BPO companies and BPM companies. They're outsourced and all that. So all that is going through a massive. Change and upheaval and how essentially all this is going to function because, like it or not, that's coming. And we need to adopt and change to the changing times and figure out to survive and build ourselves in this new ecosystem of change because all these processes are going to be attacked by agents and ai.

ben parker: 34:07

And are you seeing any particular industries that are really have seen a lot of the biggest adoption of AI or Gentech ai? Sorry.

sundip gorai: 34:16

Yeah, see the wherever there is a cross collaboration, wherever there is. Ambiguity of decision making those places will go through a massive disruption. Okay, if, let's take healthcare and life sciences, right? If you see and especially healthcare in Americas, if you see healthcare there are so many actors, so many players, so many decision making. So many traditional technologies. Even if you go to a doctor's office, people are still relying on faxes today. Okay? And email is not getting used. There are so many new machines getting adopted. Then there is integration of the payers. Who are the insurance companies? They're the providers, which are the hospitals. Then there is the pharma, which is providing there is the patient. Okay? So the actors are various, the needs are various. The decision making is various. So this is just one example. Likewise, in the banking and the payment industry, for example, the payment industry, so many new players are there. Traditional big players have been there like Visa, MasterCard, on one side, global payments on another side. And then there are newer players like Stripe, and then there are clovers, scrap, they're coming in. So the way that. Transaction is getting processed, that itself is going through a massive change. And the whole, if the value chain is changing, which naturally means that every point where the value chain operating model of the industry is changing, there is an opportunity for that to be disrupted by AI and agent AI and data science.

ben parker: 35:45

Cool. And then from your experience, are you. People that adopted adopting, yeah. Gen Gent. KI Are they, is it making impacts?'cause obviously it's a massive change, isn't it, for business? Are they making, or is it still like a learning curve for these companies?

sundip gorai: 36:03

Yes, first and foremost is majority of, I think when the generative ai hype started, I would say somewhere around 2023. Now we are in 20 25, 2 and a half years, three years at best I can pull it, so three years, the generative AI hype was there, and then I think one year odd onwards, the agent one, one and a half years. The agentic thing has taken shape, right? I would say one year, not more than one year. The agentic noise and the generative AI noise has been there for three years. And before it got replaced by the Agentic AI noise, the generative AI started. Again, people, large language models, they can answer questions. Then there were fear interpretations coming in. Hey, the answers are not accurate. Explainability is a problem. It is a black box and this and that. So bottom line is lot of experimentations happened. Okay? Again, people bought, and I'll tell you from my experience in 2000. When the big data ecosystems coming in, everybody jumped big. Servers were bought this data, that data, but people did not know what to do because the business use cases were not there. Eventually those big boxes gathered dust. Similar aspects happened with generative ai. Large number of investments happened. People have not got value out of it. Similar way the agent AI will happen. Okay? That large money will flow into this, okay? But hopefully people will be cautious because at this time I'm seeing more experimentations happening, more pointed use cases happening. I'm yet to see a large scale multi-agent ecosystem getting really adopted, which is changing an industry at least I am not aware of. Probably it is going to happen because there are so many moving parts, so many uncertainties, whether the large language model, the number of technologies hitting the market, the lack of resources, who know the technology well, lack of credible partners because their tools are evolving. So I think we are at the way I call it the Cambrian explosion, right? There was a point in time in our evolution, 5,000 years, ago, when suddenly all the species. Seemed to have emerged overnight, right from monkey to man, and in between there were suddenly millions and millions of species, which emerged overnight. We are at that Cambrian explosion of technologies where hundreds of new tools, technologies, everybody's tried crowding and jostling for the. Space where it'll be a survival of the fittest like evolution. Only a few will survey, most will go away. The bigger chunk of the market will still be controlled by the hyperscalers, like Google, Microsoft, and Amazon, and one or two big players that could be a Salesforce in between. And now there are other players, Databricks snowflake. These players are. Very important today, but probably, and I don't know, I'm saying probably they could be consumed in the bigger ecosystem of bigger players. Eventually there will be large players are going to have major say in this market, and there will be niche players who bring a clear differentiated mode for themselves. They will survive.

ben parker: 39:00

Okay. And then what, what would you say is the big risks and changes in deploying AI gen?

sundip gorai: 39:07

Yeah. So if you see the biggest, there are four or five things. Okay. The first risk is of course, the I would say one of the areas is the. Accuracy, I'll come to the others. First is accuracy, right? Is the model what the model is saying? Is it accurate? You ask Chad JA question, you get a answer. Okay? It looks correct, but can you measure and say, is it a hundred percent correct? Is it 90%? Correct. There are various means, techniques, metrics to do that, but it is still not confirmed and a lot is going on in that space. Okay. Second is explainability. Okay? Today, whatever is coming out. Suppose it's not the way I, a magician puts a handkerchief in a black hat and pulls out a rabbit. What happened in between? Nobody knows, right? It is that black box, right? Most of the deep learning models, you give input, you give output. And after appointed time, you give new output, it matches and gives you an answer like the whole output. But what happened in between, it's like a secret sauce. It's like an alchemist. It's like a saucer. It's like black magic. It's, that's something. Is explainability. There are models coming in to explain why things are happening the way things are happening, or why a neural network is working in a certain way. So that's the second important challenge of adoption. The third important challenge is ethics and social responsibility, because many models, I think if the. Great power comes great responsibility, right? It was, I think, said by Rose Walter and then Spiderman, right? And that is exactly the problem. We really need to see that just because something is powerful and it can do it, should we do it? Will it do for the greater good or not? That is the third important aspect is but a social responsibility. Then social responsibility and ethics. Fourth important aspect is scalability, right? The model may work in an experimentation layer, but when you go ahead and deploy it in a large scale, okay, with the terabyte, petabyte, terabyte scale of data, is the model going to perform the same way? Okay. And just as if you see that, if you draw a. Tiny straight line. And if you keep extending it, you could make it into a circle. But when you are looking at a circle is a circle. When you look at a small segment of that circle, it's a flat line. Just like we on, we are standing on this earth, which is flat, but actually the earth is round. What I'm trying to say that when you change the scale from small to the big. Then the whole perspective changes. And that is why what Newton had said about physics and what Einstein said about physics. When you start looking at the universe on a large scale, the whole laws of Newtonian physics probably changed or got extended, if I may say it exactly. When the scale changes, then the problems are of a totally different kind. And similarly in the data science space, when you solve a problem on Excel sheet, then you solve a problem on. Python on a desktop, and then you solve the problem on a reasonable size server, and then you are using GPU, TPU and all that to scale up in multi-country, multi-location, large number of customers. The game changes completely. So that's the other part of the problem. Scale. So to summarize, I mentioned about explainability, I mentioned about accuracy, I mentioned about ethics, I mentioned about responsibility, I mentioned about scalability. And last but not least, security. Security of systems and compliance with regulations. That is also brings a challenge just because something can do well and it may not be meeting the needs of the regulators.

ben parker: 42:45

Yeah, no, it's good. And I, yeah, I love the challenges you mentioned.'cause yeah, like for anything, as soon as you change a process or business, whatever you're doing, obviously that. There will be positives, but also then you're changing the goalpost. So there's gonna be more challenges come up, isn't there? It's that constant evolution. So it's, I guess like you cha constantly chasing your tail, aren't you really? Like things change, you've gotta keep adapting.

sundip gorai: 43:11

yeah. Absolutely.

ben parker: 43:15

So we've gen K then how mean, how's it reshaping work and employee roles?

sundip gorai: 43:22

Very profound question. Yeah. Employee roles, jobs, nobody has the answer to this, and there is fear. There is trepidation whether all the jobs will go away. And I have a slightly different take and I'll give you two, three very interesting examples here now, and people have to understand because I have always seen history as a place where all the answers are hidden. Almost every answer that you're looking for in life is hidden in history. I have a deep passion for understanding the history of inventions and that there lies most of the answers. A few examples I'll give you, let's look at aluminum, right? Aluminum today is all over ubiquitous, all over the place. A common a uten common element and a utensil of aluminum is cheap. And in poorer countries, beggars use bolts for aluminum bolts. You'll be surprised to hear that around 200 years ago during the Louis, the 14th and 15th and aluminum used to be very precious. Silver used to be cheaper, okay? Because it is very hard to make aluminum. Okay, now one fine day two gentleman found out the electrolysis process which aluminum could be created dirt cheap. Okay. And the game changed. The whole industry changed and now aluminum. Instead of picking rich people's utensils, it became a poor person's utensils. There were large number of people who are manufacturing aluminum, and the specialists who are actually building aluminum earlier went out of business because now anybody could make it very fast, very cheap. But what happened as a result? Was that many new things started coming up today. Wright brothers use aluminum in their plane. Then there were soda cans getting made of aluminum every aspect of life. Got infiltrated with aluminum and thereby the number of jobs exploded in that particular space. AI will go through a similar revolution, right? First, I think it is a bunch of small people building geeky models and all that. Today that is getting democratized with large language models. Today, coding is getting democratized with large language models. You can probably write plain, simple English and you can, you'll be able to build the models and who knows? Tomorrow all the large contact centers operating out of countries like Philippines and others. Who they may be outsourcing all technical work because coding is becomes a plain English skill. So a large number of democratization would happen. The people who are doing it and the people who will be doing it in the future are going to be different. Okay. And the number of jobs are going to increase. That is 0.1. Similarly, if you look where there were hots. Thronging the cities of New York before, I think 1928 and five years later, from horse cards, everything moved to I think, model Ty cards after Ford's assembly line came in. So imagine the people who are building those horse cards. They had one kind of skills. Now to build a car require different kind of skills. Okay? Imagine the people who are essentially probably selling peanuts or other accessories to the, at the stations where the horse card stopped. And imagine the people who are essentially selling newspapers and other, so what I'm trying to say. That skills will change as the underlying need fulfilling invention is going to change. And if that need fulfilling invention is cheaper, faster, and better, thereby all the jobs associated it with it is also going to change. Now you may ask the question, Hey, all that is good and great. But what if earlier 50 people are doing the job and now a machine is doing the job and only one person is doing one machine is doing the job. What about the 49 people who lost their job? They will never get a job. Valid question. So one way of looking at it, and the assumption is that, okay, the job market is going to expand, but what if the job market shrinks in totality? Now here I have a. Point of view, and I have not seen anything written on the internet on this yet. I have written a couple of articles on LinkedIn, is that end of the day, if everybody loses all the jobs. Then who is there to buy? There is nobody there to buy. And if there is nobody there to buy, then a robot has no incentive for opt automating anything. So there will always be a balance. There will be job losses, there will be jobs created, and the job losses cannot be so massive and so emphatic that it destroyed the consumer base such that the robots. Have no incentive of producing anything. So it'll be it is going to be rough. It is going to be very different and new economic models, probably new government styles. Who knows? Probably there could be basic economic income and thing like that. I'm not a ser but I think a lot of changes are going to happen in the short term. It is going to be very uncertain and probably let's hope that in the long term it is going to be very promising for all of us.

ben parker: 48:11

Yeah, no, definitely. I agree. I think there's gonna change, but I think one example, I remember I spoke to one leader and like when Excel come in, everyone thought the account accountancy jobs would be gone. But look here, 20 plus years still there's pretty more accountants nowadays than ever. It gives, I think roles will get more diverse. Skill sets will get. Just tweets and roles will change. It's just constant evolution, isn't it? Things just will change and obviously people will adapt, learn new skills and yeah, I think it's obviously gonna be fascinating. Obviously it would be a rocky part, but like I said, there's always gonna be, it's, you can't just, all jobs go. It's just, I just can't see that it's just gonna be an evolution. And different ways of working in the future.

sundip gorai: 49:00

Absolutely fully agree with you. I think nobody knows how, what is going to happen. I think we can just do a guessing game at this point in time.

ben parker: 49:10

What would be your, what would be your view on the future of business then with AI gen and I guess even the effects on mankind.

sundip gorai: 49:20

Yeah, so effects on mankind is, I see something as I said, that we will probably sometime discuss about meditation at certain point in time. But if you see this whole process, what differentiate us from a robot is our sensorial system, right? We, every aspect of our life, every aspect, even the smallest aspect of our life, is driven by two fundamental instincts of survival and procreation, right? So if you really see that when any species which is out there, okay, what are the two drivers that gets up in the morning that I need to survive? And thereby I need to procreate and I need to nourish myself and thereby I need to eat. Okay. Or in a very raw sense, you can say that food and sex are the primary drivers for the survival of any, from the end to the dinosaur, to human beings are two of the primary drivers of its existence. And what is. Coded where this codification is there in that animal inside, probably its DNA, that you need to get up in the morning and you need to nourish yourself and therefore you need food and you need to ensure that you survive. Or if you can't survive, you procreate so that your next generation survives. So food and sex now as time have passed. As human being and evolution has evolved, it has taken a different form, right? This has taken the shape of greed and fear. So again, the survival, two aspects of every aspect of life, therefore gets driven. But underlying construct is the same, right? That I want to either procreate or I want to nourish myself. So greed and fear are the drivers. Again, who is driving the grid and the fear, your sensorial system of that animal. Okay, now come a little further. How does this grid and fear get fool through your senses, right? You hear something, you get angry. You see something and you want it. So your life is a constant series of this fear and greed or what we call a craving and aversion. Craving and aversion. I want this, or I don't want this, I don't want this. And which. Eventually at a higher level gets manifested as happiness and sadness. Okay? Now why am I saying all this? All throughout whatever I have spoken, if you see the underlying thing in the animal that is making that happen, it sits bundle of sensorial nerves. Now look at a robot. It doesn't have these sensorial nerves. Okay, it is making a decision making. I need to do this. I need to run this algorithm, run this, do that, whatever. As it is programmed to be done, whatever has master has programmed it to do. And why this becomes very important is that. In the long run, the robots are probably going to serve us and who knows? Probably in the long run many, there could be means and techniques maybe already. Elon Musk is experimenting with neural links and trying to embed electrodes into the brain and others probably there could be regulators, like a thermostat at every given point in time would control your grade. And fear could be, there could be a algorithms possibly running inside our head and. Could be doing the coming effects today, what is been getting provided by drugs and medicines and alcohol and other kinds of stimulants and others. Probably there could be AI algorithms which could understand, and that's a way, way far off because today all that we are doing in the AI data science space has. Is I would say that in the enterprise at least trying to do things, of course there are things happening on the consumer behavior and things happening on the neuroscience space for healthcare and others. When these two merge, I think we will have a totally new earth, totally new way of looking at mankind, and we just hope. That it becomes really grand and a beautiful place because if it goes the other side, probably the algorithmic forces are taken over by forces of war, taken over by forces of criminal. So we will have that choice and we will have, as in matrix, if you see the, are we choosing the red pill or are we choosing the blue pill? That is going to be the defining. Moment of the future, the responsibility lies in our hand, the choices that we make, not just for ourselves, our future generations and mankind as a whole. And probably I think as always, the. There is a regulatory system, a bigger force in the universe, which creates the equilibrium and creates the planets and create the solar system. Probably some bigger force of ai. Even many agents are working. Millions and billions of agents are working throughout the universe. Probably they will ge generate more cosmic level of consciousness and intelligence to control that. Humanity as a whole doesn't come to harm, and I think Asimov. Isaac Azimo, one of the greatest science fiction writers, and I think I would end it on that note that he created the three laws of robotics right as to robot will solve a mankind and robot will not hum harm the man. And finally, whenever in doubt, look at the first rule and second rule, and not being able to quote it properly, but something which is rarely spoken is the zero at loss of robotics, which Azimo had coined. And in that he said that while robots will optimize, do algorithms find outcome, do this, do that for man, for betterment, but the zero labor, if all of this. Put together harms humanity as a whole, then that will take precedence over anything else, and probably with a very promising envisioned future. We hope that the zero love of robotics will continue to be respected by the robots of today and the robots that are going to come to tomorrow.

ben parker: 55:14

Brilliant. I love your, yeah, I love your folks around that and yeah, no, it's been great having you on the podcast Sunday. But yeah I really like your learning approach first I think that's good to see I think it's obviously yeah, needed in the industry as well to keep learning. And like I said you've had an amazing career. So far and Yeah. I'll see all the best in your new role. And Yeah. thanks for joining on the podcast.

sundip gorai: 55:42

Thank you so much, and thank you for inviting me to this podcast and look forward to speaking with you again on some more specific topics around industry problems and others, and all the best to you and keep doing the good work that you have started.