>> Live from Las Vegas, its theCUBE. Covering Informatica World 2019. Brought to you by Informatica. >> Welcome back everyone to the theCUBE's live coverage of Informatica World 2019 here in Sin City. I'm your host Rebecca Knight. We're here with Arun Varadarajan. He is the vice president of AI and anaylsitcs at Cognizant. Thank you so much for coming on theCUBE Arun. >> Wonderful its also great to meet you folks at theCUBE. >> You are Cube alarm. >> I am a Cube alarm. This is probably the third or fourth time that I'm on theCUBE. >> Excellent. Well for those viewers who have not seen your previous clips tell us a little bit about your role at Cognizant. >> My role at Cognizant is focused on two primary things. One is to really get our customers ready for AI and truly compete in the digital world. The second big focus for me is to get them there. To me it's all about the data. So many times we don't realize this that if you look at a lot of the FANG players. The digital natives are born digital who really have leveraged machine learning and AI to disrupt the market place. They do it with data. It's all about the data. So the big push that I'm working on these days is to help our clients create this new modern data platform that can truly help them leverage AI and disrupt the market where possible. >> So tell us what you've-- So we know that this journey is incredibly complex and there's a lot of layers, a lot of questions, hard questions that companies are wrestling with. >> Yes. >> Give us the lay of the land. What do you see as sort of the big dominant forces happening in AI and ML? >> I think the first place is companies are still trying to figure out where do they apply AI and ML. I think that is where they need to start because if it is not designed and the initiative is not purposed around any sort of specific business area or business focus or business outcome, it becomes an engineering project that really doesn't see light of day. If you remember back in the days when Hadoop was big. Hadoop was almost like a solution trying to find-- A problem or a solution trying to find a problem, whichever way it is. I think as opposed to taking a technology view which has been the traditional approach that most of the CIO organizations have used. In AI even more so, there needs to be significant participation for the business to decide where are the opportunities for me to drive business value. So I've always told my clients that the place to start is where can I apply AI and machine learning because at the end of the day it is just a technique right, and the technique has to be focused on delivering true business outcomes and business values. So that is where I think our clients need to start. If you go back in time and remember the ERP days when people were implementing SAP and Oracle there was this very strong focus on process optimization and process excellence. How do I get a straight through process organization? Really create that process orchestration layer that could execute at excellence. I think that needs to be brought back today but in a different light and the light is, now let me view my value chain, not just from a process orchestration standpoint but where are the opportunities for me to leverage machine learning and AI to create very different outcomes within that process layer? And I think-- Sorry. >> I definitely want to go back to that but I also want to remember that we are here at Informatica World and I want to make sure I ask you how you at Cognizant work with Informatica. >> Informatica is a strategic partner of ours and as I was saying, while you start with that outcome in mind and really say these are the areas I want to drive business outcomes it's very important you understand how data plays a role in delivering those outcomes. So that's where Informatica and our partnership really comes to fruition. You know that Informatica has been working very strong in the areas of metadata management, data governance, security. All of these are essential part of you knowing your data and knowing where your data's coming from, where is it going, who is using it, how is it being consumed, in what form and shape should it be delivered so that we can deliver business value is a key aspect of really leveraging AI and machine learning. In AI and machine learning the one thing that we have to be cognizant of, pun intended, is the fact that when you're going to get the machine to start making decisions for you, the quality of your data has to be significantly higher than just a report that is inaccurate, right. Report inaccuracy, yes you're going to get shouted at by the consumer of the report but that's the only problem you face but with AI and machine learning coming into play if your data is not truly representative of the decision area that the machine is working on then you're going to have a very bad outcome. >> This is a deep and philosophical issue because if the data is shoddy or biased there is a lot of problems that companies can get into. So where do you even start? How do you even work with a company to make sure that their data is the right data, is pure? What do you think? >> Interesting you ask that question. We've come up with this notion that even data has got IQ. We call it data IQ in Cognizant and it's a mathematical measure that we have come up with which allows us to score a data's ability to perform in a given area or function. So it could be in the area of sales effectiveness. Look we have a large retail company that is really trying to figure out how can they improve their store level information so that they can execute more sales orders with their customers. Their assumption is that they're working with a data set that can help them drive that outcome. How do they know that? Well there's one way to find out, which is for you to experiment, test, and learn and test and learn but that's an arduous process. Which is why a lot of the data science work that is happening today is, I would say, probably seventy to 80% of the data science effort goes waste because there are experiments that fail. This was-- >> But is that a waste? So it failed, but you tried and you maybe had some learnings from it, right? >> So a lot of people keep saying that failure is a great teacher of-- >> That's the Silicon Valley mantra right now. >> Well you can be smart about where you fail. >> True. >> Right. >> Good point. >> If there are opportunities for you to prevent that failure why wouldn't you? >> Okay. All right. >> That's what we're looking at. So what I'm saying is that before you go into doing any data science experiment, what if I came back and told you that the data that you're working on is not going to be sufficient for you to deliver that outcome. Would it not be interesting? >> Exactly, so it's making sure that you at least are maximizing your chance of success by having the right data to begin with. It is a failure for failure's sake if you're not even starting with the right data. >> Absolutely and you know the other thing that people don't realize is is if you go and ask-- If you just do it, I'm going back to my industrial engineering days, if you go and do a simple time and motion study of data science, data scientists, I can guarantee you that 80% or 90% of their time is spent on just prepping the data and only less than 10% or 15% on truly driving business value. So my question is you're spending big dollars on data science experiments where eighty to 90% of the time the data scientists are prepping. Looking at the data, is it the right skew, has it the right features, do I need to do some feature engineering, do you denormalize it? There are a whole bunch of data prep work that they do. My question is, what if we take that pain away from them? That's what I call as data science freedom and this is what we are promoting to our clients saying what can you do with your data so that your data is ready for the data science folks? Today it's data science folks, tomorrow it's going to be hopefully machine learning algorithms that can self model because a lot of people are talking about auto ML which is the new buzz-word, which is AI doing AI and that's an area that we're heavily invested in. Where you really want to make sure that the data going in is of the veracity and the complexity and the texture required for that outcome area. So that's where I think things like data IQ as a concept would really help our clients to know that hey the data I'm working with has got the intrinsic intelligence in that outcome area for me to drive that particular business outcome that I'm working on. That's where I think the magic lies. >> That's where they'll see the value. >> That's where they'll see the value. >> So talk a little about the AI journey because that is, it's all intertwined but so many companies are coming to you, to Cognizant and saying we know we need to do more of this, we want to make it real, how do we get there? So what do you say? What's your advice? >> So, I think I mentioned this right up-front when we started the conversation. It all has to start with purpose. Without purpose no AI project really succeeds. You'll end up creating a few bots. In fact when I look out there in the world and look at the kind of work that is happening in machine Learning and AI, many of the so-called AI projects, if you double click on them, are just bots. So we are doing some level of maybe process automation, we're trying to reduce labor content, bringing in bots, but are we truly driving change? I'm not saying that that's not a change. There is definitely a change but it's more of an incremental change. It is not the kind of disruptive change that some of the FANG leaders that are showing right. If you take Facebook, Amazon, the whole gamut of digital natives, they are truly disrupting the market place. Some of them are even able to do a million predictions a second to match demand, supply, and price. Now that is how they are using it. Now the question I think for our clients, for our enterprise clients is to say that's a great goal to have but where do I start and how do I start? It starts with, in my opinion, two or three big notions. One is, honestly ask yourself, how much of a change are you willing to make, because if you have to compete and really leverage AI and machine learning the way it has been designed to do so you have to be willing to press the reset button. You have to be willing to destroy what you have today and there is, I think Bill Baker back in the days, he was a SQL server guy. He was talking about this whole concept of what is known as scale up and scale out and he was talking about it from the angle of managing a pet versus managing cattle. So when you're managing a pet, a pet is a very unique component like your mail server So Bob the mail server, if the mail server goes down then all hell breaks loose and hopefully you have another alternate to Bob to manage the mail server. So it's more like a scale up model where you are looking at, hey how do I manage high availability as opposed to today's world where you have the opportunity to really look at things in a far more expansive manner. So if you have to do that you can't be saying I have this on-prem data warehouse right, which is running on X Y Z, and I want to take that on-prem data warehouse and move it to the Cloud and expect magic to happen, because all you're doing is you're shifting your mess from your data center to somebody else's data center which is called the Cloud. >> Right. >> Right? So I think the big thing for clients to really understand is how much are they invested in this change. How are they willing to drive this change? I'll tell you it's not about the technology. There are so many technology options today and we have got some really smart engineers who know how to engineer things. The question is, what are you doing this for? Are you willing, if you want to compete in that paradigm, are you willing to let go of what you have tody? That is a big question. That I would start with. >> An important question but I want to sneak in one more question and that is about the skills gap because this is something that we hear so much about. So many companies facing a, there is a dearth of qualified candidates who can do these jobs in data science and AI and ML. What are you seeing at Cognizant and what are you doing to remedy the problem? >> So I think it's definitely a challenge for the industry at large and what we are starting to see is two things emerging. One is the new workforce coming into the market is better equipped because of the way the school systems have changed in the last few years and I would say this is a global phenomenon not just in North America or in Europe or in China or India. It's a global phenomenon. We're starting to see that undergrad students who come out of school today are better equipped to learn the new capabilities. That's number one. Which is very heartening for us right, in the whole talent space. What I've always believed in, and this is my personal view on this, what I've always believed in is that these skills will come into fashion and go out of fashion in months and days. It's about the kind of engineering approach you have that stays constant, right. If you look at any of the new technologies today, they all are based on some core standard principles. Yes the semantics will change, the structure will change, but some of the engineering principles remain the same. So what we've been doing in Cognizant is really investing in our engineering talent. So we call it data engineering and to us data engineering means that if you're a data engineer you can't tell me I will only work with A, B, or C technology. You should be in a position to work with all of these technologies and you should be in a position to approach it from an engineering mindset as opposed to a skill or a tool based mindset and that's the change that we need with fads coming in and out of Vogue. I think it's super important for all consultants in this space to be grounded on some core engineering principles. That's what we are investing in very heavily. >> Well it sounds like a sound investment. Well thank you so much for coming on the show Arun. I appreciate it. >> Thank you so much. It was a pleasure. >> I'm Rebecca Knight for theCUBE. You are watching theCUBE at Informatica World 2019. Stay tuned. (lighthearted music)

>> Voiceover: Live from Las Vegas, it's theCUBE. Covering Informatica World 2018, brought to you by Informatica. >> Hey, welcome back everyone, we're here live at the Venetian, we're at the Sands Convention Center, Venetian, the Palazzo, for Informatica World 2018. I'm John Furrier, with Peter Burris, my co-host with you. Our next guest, Arun Varadarajan, who's the VP of AI and Analytics at Cognizant. Great to see you. It's been awhile. Thanks for coming on. >> Thank you. Thank you John, it's wonderful meeting you again. >> So, last time you were on was 2015 in the queue. We were at the San Francisco, where the event was. You kind of nailed the real time piece; also, the disruption of data. Look ing forward, right now, we're kind of right at the spot you were talking about there. What's different? What's new for you? ASI data's at the center of the value preposition. >> Arun: Yep. People are now realizing, I need to have strategic data plan, not just store it, and go do analytics on it. GDPR is a signal; obviously we're seeing that. What's new? >> So, I think a couple of things, John. One is, I think the customers have realized that there is a need to have a very deliberate approach. Last time, when we spoke, we spoke about digital transformation; it was a cool thing. It had this nice feel to it. But I think what has happened in the last couple of years is that we've been able to help our clients understand what exactly is digital transformation, apart from it being a very simple comparative tactic to deal with the fact that digital natives are, you know, barking down your path. It also is an opportunity for you to really reimagine your business architecture. So, what we're telling our clients is that when you're thinking about digital transformation, think of it from a 3-layer standpoint, the first layer being your business model itself, right? Because, if you're a traditional taxi service, and you're dealing with the Uber war, you better reimagine your business model. It starts there. And then, if your business model has to change to compete in the digital world, your operating model has to be extremely aligned to that new business model paradigm that you've defined. And, to that, if you don't have a technology model that is adapting to that change, none of this is going to happen. So, we're telling our clients, when you think about digital transformation, think of it from these three dimensions. >> It's interesting, because back in the old days, your technology model dictated what you could do. It's almost flipped around, where the business model is dictating the direction. So, business model, operating model, technology model. Is that because technology is more versatile? Or, as Peter says, processes are known, and you can manage it? It used to be, hey, let's pick a technology decision. Which database, and we're off to the races. Now it seems to be flipped around. >> There are two reasons for that. One is, I think, technology itself has proliferated so much that there are so many choices to be made. And if you start looking at technology first, you get kind of burdened by the choices you need to make. Because, at the end of the day, the choice you make on technology has to have a very strong alignment and impact to business. So, what we're telling our clients is, choices are there; there are plenty of choices. There are compute strategies available that are out there. There's new analytical capabilities. There's a whole lot of that. But if you do not purpose and engineer your technology model to a specific business objective, it's lost. So, when we think about business architecture, and really competing in the digital space, it's really about you saying, how do I make sure that my business model is such that I can thwart the competition that is likely to come from digital natives? You saw Amazon the other day, right? They bought an insurance company. Who knows what they're going to buy next? My view is that Uber may buy one of the auto companies, and completely change the car industry. So, what does Ford do? What does General Motors do? And, if they're going to go about this in a very incremental fashion, my view is that they may not exist. >> So, we have been in our research arguing that digital transformation does mean something. We think that it's the difference between a business and a digital business is the role that data plays in a digital 6business, and whether or not a business treats data as an asset. Now, in every business, in every business strategy, the most simple, straightforward, bottom-line thing you can acknowledge is that businesses organize work around assets. >> John: Yep. >> So, does it comport with your observation that, to many respects, what we're talking about here is, how are we reinstitutionalizing work around data, and what impact does that have on our business model, our operating model, and our technology selection? Does that line up for you? >> Totally, totally. So, if you think about business model change, to me, it starts by re-imagining your engagement process with your customers. Re-imagining customer experience. Now, how are you going to be able to re-imagine customer experience and customer engagement if you don't know your customer? Right? So, the first building block in my mind is, do you have customer intelligence? So, when you're talking about data as an asset, to me, the asset is intelligence, right? So, customer intelligence, to me, is the first analytical building block for you to start re-imagining your business model. The second block, very clearly, is fantastic. I've re-imagined customer experience. I've re-imagined how I am going to engage with my customer. Is your product, and service, intelligent enough to develop that experience? Because, experience has to change with customers wanting new things. You know, today I was okay with buying that item online, and getting the shipment done to me in 4 days. But, that may change; I may need overnight shipping. How do you know that, right? Are you really aware of my preferences, and how quickly is your product and service aligning to that change? And, to your point, if I have customer intelligence, and product intelligence sorted out, I better make sure that my business processes are equally capable of institutionalizing intelligence. Right? So, my process orchestration, whether it's my supply chain, whether it's my auto management, whether it's my, you know, let's say fulfillment process; all of these must be equally intelligent. So, in my mind, these are three intelligent blocks: there's customer intelligence, product intelligence, and operations intelligence. If you have these three building blocks in place, then I think you can start thinking about what should your new data foundation look like. >> I want to take that and overlay kind of like, what's going on in the landscape of the industry. You have infrastructure world, which you buy some rack and stack the servers; clouds now on the scene, so there's overlapping there. We used to have a big data category. You know, ADO; but, that's now AI and machine learning, and data ware. It's kind of its own category, call it AI. And then, you have kind of emerging tech, whether you call, block chain, these kind of... confluence of all these things. But there's a data component that sits in the center of all these things. Security, data, IOT, traverse infrastructure, cloud, the classic data industry, analytics, AI, and emerging. You need data that traverses all these new environments. How does someone set up their architecture so that, because now I say, okay, I got a dat big data analytics package over here. I'm doing some analytics, next gen analytics. But, now I got to move data around for its cloud services, or for an application. So, you're seeing data as to being architected to be addressable across multiple industries. >> Great point John. In fact, that leads logically to the next thing that me and my team are working on. So we are calling it the Adaptive Data Foundation. Right? The reason why we chose the word adaptive is because in my mind it's all about adapting to change. I think Chal Salvan, or somebody said that the survival of the fittest is not, the survival is not of the survival of the fittest or the survival of the species that is intelligent, but it's the survival of those who can adapt to change, right? To me, your data foundation has to be super adaptive. So what we've done is, in fact, my notion, and I keep throwing this at you every time I meet you, in my opinion, big data is legacy. >> John: Yeah, I would agree with that. >> And its coming.. >> John: The debate. >> It's pretty much legacy in my mind. Today it's all about scale-out, responsive, compute. The data world. Now, if you looked at most of the architectures of the past of the data world, it was all about store and forward. Right? I would, it's a left to right architecture. To me it's become a multi-directional architecture. Therefore what we have done is, and this is where I think the industry is still struggling, and so are our customers. I understand I need to have a new modern data foundation, but what does that look like? What does it feel like? So with the Adaptive Data Foundation... >> They've never seen it before by the way. >> They have not seen it. >> This is new. >> They are not able to envision it. >> It is net new. >> Exactly. They're not able to envision it. So what I tell my clients is, if you really want to reimagine, just as you're reimagining your business model, your operating model, you better reimagine your data model. Is your data model capable of high velocity resolutions? Whether it's identity resolution of a client who's calling in. Whether it's the resolution of the right product and service to deliver to the client. Whether it's your process orchestration, they're able to quickly resolve that this data, this distribution center is better capable of servicing their customer need. You better have that kind of environment, right? So, somebody told me the other day that Amazon can identify an analytical opportunity and deliver a new experience and productionize it in 11.56 seconds. Today my customers, on average, the enterprise customers, barely get to have a reasonable release on a monthly basis. Forget about 11.56 seconds. So if they have to move at that kind of velocity, and that kind of responsiveness, they need to reimagine their data foundation. What we have done is, we have tried to break it down into three broad components. The first component that they're saying is that you need a highly responsive architecture. The question that you asked. And a highly responsive architecture, we've defined, we've got about seven to eight attributes that defines what a responsive architecture is. And in my mind, you'll hear a lot of, I've been hearing a lot of this that a friend, even in today's conference, people are saying, 'Oh, its going to be a hybrid world. There's going to be Onprim, there's going to be cloud, there's going to be multicloud. My view is, if you're going to have all of that mess, you're going to die, right? So I know I'm being a little harsh on this subject, but my view is you got to move to a very simplified responsive architecture right up front. >> Well you'd be prepared for any architecture. >> I've always said, we've debated this many times, I think it's a cloud world, public cloud, everything. Where the data center on premise is a huge edge. Right, so? If you think of the data center as an edge, you can say okay, it's a large edge. It's a big fat edge. >> Our fundamentalists, I don't think it exists. Our fundamental position is data increasingly, the physical realities of data, the legal realities of data, the intellectual property control realities of data, the cost realities of data are going to dictate where the processing actually takes place. There's going to be a tendency to try to move the activity as close to the data as possible so you don't have to move the data. It's not in opposition, but we think increasingly people are going to not move the data to the cloud, but move the cloud to the data. That's how we think. >> That's an interesting notion. My view is that the data has to be really close to the source of position and execution, right? >> Peter: Yeah. Data has got to be close to the activity. >> It has to be very close to the activity. >> The locality matters. >> Exactly, exactly, and my view is, if you can, I know it's tough, but a lot of our clients are struggling with that, I'm pushing them to move their data to the cloud, only for one purpose. It gives them that accessibility to a wide ranging of computer and analytical options. >> And also microservices. >> Oh yeah. >> We had a customer on earlier who's moved to the cloud. This is what we're saying about the edge being data centered. Hybrid cloud just means you're running cloud operations. Which just means you got to have a data architecture that supports cloud operations. Which means orchestration, not having siloed systems, but essentially having these kind of, data traversal, but workload management, and I think that seems to be the consistency there. This plays right into what you're saying. That adaptive platform has to enable that. >> Exactly. >> If it forecloses it, then you're missing an opportunity. I guess, how do you... Okay tell me about a customer where you had the opportunity to do the adaptive platform, and they say no, I want a silo inside my network. I got the cloud for that. I got the proprietary system here. Which is eventually foreclosing their future revenue. How do you handle that scenario? >> So the way we handle that scenario, is again, focusing on what the end objective, that the client has, from an analytical opportunity, respectfully. What I mean by that is that semi-customer says I need to be significantly more responsive in my service management, right? So if he says I want to get that achieved, then what we start thinking about is, what is that responsive data architecture that can tell us a better outcome because like you said, and you said, there's stuff on the data center, there's stuff all over the place, it's going to be difficult to take that all away. But can I create a purpose for change? Many times you need a purpose for change. So the purpose being if I can get to a much more intelligent service management framework, I will be able to either take cost out or I can increase my revenue through services. It has to be tied to an outcome. So then the conversation becomes very easy because you're building a business case for investing in change, resulting in a measurable, business outcome. So that engineer to purpose is the way I'm finding it easier to have that conversation. And I'm telling the plan, keep what you have so you've got all the speckety messes somebody said, right? You've got all of the speckety mess out there. Let us focus on, if there are 15 data sets, that we think are relevant for us to deliver service management intelligence, let's focus on those 15 data sets. Let's get that into a new scalable, hyper responsive modern architecture. Then it becomes easier. Then I can tell the customer, now we have created an equal system where we can truly get to the 11.56 seconds analytical opportunity getting productionized. Move to an experiment as a service. That's another concept. So all of that, in my opinion John, is if he can put a purpose around it, as opposed to saying let's rip and replay, let's do this large scale transformation program, those things cost a lot of money. >> Well the good news is containers and Cubernetties is stowing away to get those projects moving cloud natives as fast as possible. Love the architecture vision. Love to fault with you on that. Great conversation. I think that's a path, in my opinion. Now short-term, the house in on fire in many areas. I want to get your thoughts on this final question. GDPR, the house is on fire, it's kind of critical, it's kind of tactical. People don't like freaking out. Saying okay, saying what does this mean? Okay, it's a signal, it is important. I think it's a technical mess. I mean where's the data? What schema? John Furrier, am I J Furrier, or Furrier, John? There's data on me everywhere inside the company. It's hard. >> Arun: It is. >> So, how are you guys helping customers and navigate the landscape of GDPR? >> GDPR is a whole, it's actually a much bigger problem than we all thought it was. It is securing things at the source system because there's volatibilities of source system. Forget about it entering into any sort of mastering or data barrels. They're securing its source, that is so critical. Then, as you said, the same John Furrier, who was probably exposed to GDPR is defined in ten different ways. How do I make sure that those ten definitions are managed? >> Tells you, you need an adaptive data platform to understands. >> So right now most of our work, is just doing that impactive analysis, right? Whether it's at a source system level, it has data coverance issues, it has data security issues, it has mastering issues. So it's a fairly complex problem. I think customers are still grappling with it. They're barely, in my opinion, getting to the point of having that plan because May 18, 2018 May, was supposed to, for you to show evidence of a plan. So I think there... >> The plan is we have no plan. >> Right, the plan of the plan, I guess is what they're going to show. It may, as opposed to the plan. >> Well I'm sure it's keeping you guys super busy. I know it's on everyone's mind. We've been talking a lot about it. Great to have you on again. Great to see you. Live here at Informatica World. Day one of two days of coverage at theCUBE here. In Las Vegas, I'm John here with Peter Burris with more coverage after this short break. (techno music)