Buying AI: How to Invest in Artificial Intelligence

Artificial intelligence can make companies dramatically more efficient, but investing in the technology can come with risks and complications. Tiger Tyagarajan, CEO of Genpact, tells Michael Krigsman of CXOTALK about the best strategies for buying AI to improve your organization.

43:56

Jul 21, 2017
6,268 Views

Artificial intelligence can make companies dramatically more efficient, but investing in the technology can come with risks and complications. Tiger Tyagarajan, CEO of Genpact, tells Michael Krigsman of CXOTalk about the best strategies for buying AI to improve your business.

Tyagarajan says companies interested in AI sometimes adopt disparate solutions that aren’t properly connected across the entire organization. Genpact has launched Genpact Cora, which integrates automation, analytics, and AI engines into a single platform to tackle operational business challenges from beginning to end. Genpact Cora is already helping a pharma company refine drug safety, and assisting a financial services institution in cutting costs 75% from dealing with regulatory compliance issues.

As CEO of Genpact, Tiger has overseen the company’s transformation from a division of GE in India into a leading global professional services firm with $2.46 billion revenue. Based in New York, Tiger frequently writes and speaks about artificial intelligence, global talent issues, continuous skill development, and the importance of building a strong corporate culture. He is also passionate about diversity and is a member of the WSJ CEO Council.

Genpact is a global leader in digitally-powered business process management and services. For two decades, first as a General Electric division and later as an independent company, has been servicing hundreds of strategic clients, including approximately one-fifth of the Fortune Global 500. Today, Genpact employs over 75,000 people in 25 countries, with key offices in New York City.

Transcript

Michael Krigsman: Artificial intelligence is a big deal today. It’s so confusing. There’s so much jargon. If you work in the enterprise, how do you sift through the jargon? How do you sift through the hype, and how do you actually buy AI in a way that will actually make sense and is meaningful? Today, on Episode #246 of CxOTalk, we’re exploring this topic.

I’m Michael Krigsman, an industry analyst and the host of CxOTalk. And today, I have the pleasure of speaking with Tiger Tyagarajan, who is the CEO of Genpact. Hey, Tiger! How are you? Thank you for taking the time to be here today!

Tiger Tyagarajan: Great, Michael! Great to be here and thank you for inviting me to your show!

Michael Krigsman: Tiger, tell us about Genpact. But first, I just want to say "Thank you" to Livestream, who is our streaming partner because those guys are great. And, if you're doing live streaming, go to Livestream.com, and if you go to Livestream.com/CxOTalk, they will actually give you a discount on their plans.

So Tiger, tell us about Genpact.

Tiger Tyagarajan: Michael, Genpact is a global professional services company. Being around for about twenty years, we really deliver to over a thousand corporations two things: intelligent operations, which is one part of our business, where we deliver operations with digital tools and better than them; analytical capabilities and insights, and predictive analytical capabilities, and better than them; and run those operations for our clients. Intelligent operations.

For example, think about operations where insurance claims are processed and decisions are taken on what claims to be made, how they are to be made, etc. […] When does a part get delivered in order ot repair equipment out in the field, that then the field engineer ends up there and actually repairs it, that flow of information, etc. So, running intelligent operations.

And the second part of our business it taking a number of our digital and analytical solutions to our clients and changing the way they run their business. We call that digitally-led solutions and platforms. So, that's what we do as a business.

Michael Krigsman: Okay. So, you have been around for a long time and I know that a key focus of your business is working with clients who are undertaking various programs of digital transformation. And certainly, AI, artificial intelligence, is important with that. And so, to begin this conversation on how to buy AI, if you’re in the enterprise, maybe share with us what do you see your clients doing with AI at this stage of its development; the development of AI?

Tiger Tyagarajan: So, Michael, it think I loved your introduction to the topic right at the top of the show where you said it is very topical, the topic of artificial intelligence, for almost every one of our clients., And within the clients, every one of the C-suite is our clients. And, it's in the context of how do I bring AI into the company to become part of the way we run the company and become part of our digital transformation journey that is just described?

However, AI is just one of the many new digital technologies that are becoming more and more leverageable in the world we are in. That is changing the way work gets done. That is changing the way decisions get taken in the industry. So, every client wants to talk about it. We talk about it. We bring it in. But, the way to describe AI in that digital transformation journey is at the edge of that journey. It's at the pinnacle of that journey. It's at the edge where most value will potentially get created in the future in that journey. And also, to think about it, not just as AI, but a set of tools and technologies, one of them being AI, and probably the most valuable but really, most valuable when it's strung together.

Michael Krigsman: Tiger, so when you talk about bringing AI into the digital transformation journey, can you elaborate on that? Or, you said bring AI and make it part of the digital transformation and part of the business? And, this seems to be a very key point.

Tiger Tyagarajan: It is, Michael! And, I said that as the first overarching theme. But, there’s then the second thing underneath that. When an enterprise thinks about leveraging AI, if they think about it on day one, as I want to leverage AI across my business, in its entirety, I think people run into one problem, which is it’s just too much of an ocean to boil. And, it’s too much of a problem to try and grapple with and solve. And, AI… That’s not necessarily the most effective, efficient way of bringing AI to bear onto a problem. Where we have found the most success in most client situations, is when a client says, “I want to bring AI in. I want to start leveraging AI,” my biggest problem that I want to attack and solve for, let’s say, is managing receivables in my enterprise. It’s managing working capital. And within that, I actually struggle often with the way we send our bills out, the way we actually then work with our clients to make sure they pay on time, and all the administrative aspects of doing all of that.

And, I wish I could solve that because I really want to drive customer satisfaction much higher than it is today. I want to obviously bring my working capital down. I don’t want to use labor to do this because it’s too painful, and therefore I want to use a machine. And, I want to get better, and better, and cleverer at doing this. And by the way, in all of that, I want to win in the marketplace. So, it’s actually a lot about competitive advantage. It’s about outcomes, customer satisfaction, growth, and working capital in this example and more cash.

And, when the AI gets brought in to bear on a specific problem, we see the best answers emerge. When AI is coming into the company and it attempts to solve the whole company and, "Let's deal with AI in the company," it's a much, much different journey and often, people get disappointed.

Michael Krigsman: So, […] is it because of the hype around AI that people tend to focus on AI as a kind of, I was going to say "magical solutions," but let's say disconnected from what they're trying to do with their business? Is that what often happens?

Tiger Tyagarajan: So, Michael, I would start by saying all of us are learning. AI has been around and various aspects of AI, whether it's computation, linguistics, or it's computer vision and using computer vision for intelligent decision making… I mean, a number of those technologies have been around for decades. What's changed is really the ability to leverage them with the speed that's available on networks with the ability to process data and the fact that gigabytes and petabytes of data can be processed, and the speed at which that can be done, all of that has made AI come alive, and come alive in both the […] space… And we are all users of gadgets like the iPhone and other smartphones, as well as in the enterprise space.

However, the actual use of those now are getting into use-cases that are beginning to come alive, but it's still early. So, that was a lot of learning going on. Anytime you have [some] learning going on, you have people exploring, you have people experimenting, you have people asking questions that need answers. And they may come to the conclusion after some time saying it was the wrong question to ask.

So, the question could be the board, and the CEO, and the CXO team saying, “We want to use AI in the company. Let’s just bring AI into the company.” Well, after some time, they realized actually the better way to approach it is artificial narrow intelligence, ANI, which is when we narrow down the problem to the most valuable one we can solve, and really attack it with AI. And, that’s often called “artificial narrow intelligence.” We find those to be highly successful.

Michael Krigsman: You know, it’s […] an interesting set of points that make a lot of sense. But, at the same time, in a way, if you go to your clients and say this, is there a little bit of a let-down because they’d want to do all of this stuff with AI and you’re saying, “Well, is this going to solve specific problems or efficiencies in your company?” and it kind of makes it lose the shiny, new object aspect of it?

Tiger Tyagarajan: So, Michael, you are right. Typically, that's what one sees in the world of technology when it's going through what is called a "hype cycle", and clearly AI is on that cycle. But here is where I think it does tend to ultimately resonate with clients. As long as we pick big enough problems that are either big enough pain points for our clients of value for them competitively, or big enough opportunities that they want to grab onto that can differentiate them. And as long as that value can be brought alive and it can be seen, and if the cycle-time to make it happen is short enough, if the cost of that experimentation is not high, and if at the end of that experiment, you can actually say that this has worked, let's scale it up. Or, if it has not worked, let's pull the plug and get on to the next problem.

I think people start to really… One, agreeing that it's a great approach, gravitating to that approach, and actually beginning to see the value. I think there are enough examples beginning to build up in the enterprise space, in different industries and different solutions, that is beginning to resonate. And as those get known in various industries, and obviously people like us allow that propagation of knowledge to go through, I think people are beginning to understand this. But you're right. I mean, there's still wonder. Why can't I bring it in altogether into many places and solve world hunger?

Michael Krigsman: So, there’s this practical aspect of AI. So, how do we make our accounts payable more efficient? How do we make our customer service more efficient, which is crucially important from a business standpoint, but there’s also the innovation dimension of this, looking forward. And you’re suggesting start with maybe smaller projects to explore that.

Tiger Tyagarajan: So, Michael, now I’m not suggesting that actually. Clearly, that is an option. But, I’m not suggesting that. Our view, and our belief and my belief is that when AI gets applied to the most crying problem, the biggest opportunity is where the value is the highest. You typically try and do that with a proof-of-concept so you… So, the problem is big, the opportunity is big, and if solved for, could be a huge benefit, not a small benefit. And you say, “Let’s test this in a small part of that problem.” So, create a proof-of-concept. And, that can be done. It’s not that difficult, and it’s worth doing it because these are short experiments. When that works, then you say, “Now, I’m going to scale it” in that still, same problem. And that’s a big solution.

So, let me give you one example of… So, think about, in a banking environment, in let's say, the small business-lending world, every bank and every financial institution before it decides to approve a lending, decides that based on reading financial statements of the business they are lending to and understanding it, doing cash flows, etc., that's a pretty significant expert process driven by knowledge of understanding financial statements, and then applying them to risk policies, etc. that the bank may have. Those financial statements, unfortunately, given the nature of small and medium businesses come in a variety of forms, have a variety of ways some of those get laid out. So, depreciation gets calculated different ways by different small businesses. Cash flow gets calculated differently, people lay out their financials differently; and then there are footnotes, and then there are caveats, all of that is a dollar sheet and P&L document.

Today, an expert opens the document, lays the financials out, and then says “I understand the cash flow of this company. By the way, this is the way we define it in the bank. Our risk policies are the following, and therefore, I’m going to approve this customer for half a million dollars to buy the equipment they want to buy.” You can now say, “Why don’t we bring in AI that takes all this data, ingests it, reads all the information, understands words, understands the meaning of depreciation in the context of the way this company has laid it out;” then reads the footnotes and says, “Oh! The footnote says this, so let’s change the definition of depreciation that this company is using to actually match the way we think it should be defined.” And then, it spits out that financial statement that otherwise an expert would have to write, and then compares that to the risk parameter that says, “This is the amount you should lend.” That whole thing now gets done by the expert artificial intelligence-based product.

In fact, we often have a discussion about a priority like that. We call that “live spread,” in a process called “live credit.” And by definition, the use of the word “live” basically means that what used to be a four or five-day process of underwriting to make a decision could now be a thirty-minute process. So all of a sudden, you’re not talking about efficiency. You are talking about winning in the marketplace because that bank that does this will win more; will get more lending out, will have a bigger book of lending out and by the way, will actually get the best customers. So, actually, longer-term, we’ll build a longer book that actually has a better quality than someone else.

Michael Krigsman: So, you're talking about innovation that is focused on improving the business, as opposed to, "Oh, well let's adopt AI in vague terms."

Tiger Tyagarajan: Yes, yes. Yes, Michael. And obviously, this becomes a debate. Often, industry suite, where should we apply it first? It becomes a debate in our kind of business with the client back and forth. Obviously, the more experience either we or other people who do this have, they can bring those to life. The advantage of things like AI and other digital technologies is that once you do it, twice you do it, three times you do it, you can actually then repeat it more easily and more often and at a better speed. And, obviously, by definition AI, so it learns and gets better and better. And typically, if you apply it to a problem in an industry that I just described as an example, and four people use it, twenty other people jump in… Actually, the four who originally started using it get the benefit of the fact that now, twenty people have jumped in.

So, one of the core benefits of AI longer-term, not today, but longer-term that’s going to happen is the network effect similar to other technologies that have had huge benefits to everyone in the industry because of network effects. We think AI is going to bring network effect benefits that far exceed anything that we’ve even seen today.

Michael Krigsman: I want to remind everybody that we’re speaking with Tiger Tyagarajan, who is the CEO of Genpact. And right now, there’s a tweet chat going on using the hashtag #cxotalk and you can jump in and add your comments, and ask your questions to Tiger.

Tiger, you acquired an AI company. Maybe, tell us about that, please?

Tiger Tyagarajan: Yes, Michael. We acquired a company called Rage Frameworks, a company that has an AI-based platform. And, the way to think about that company is it's in one of the AI spaces. There are a couple of more spaces that one could talk about in the context of AI. The space that Rage Frameworks and the company we bought based out of Boston straddles is called "using computational linguistics." And, in very simple terms, what the platform does and what the company does, and what they've built over eight, nine years is the ability to ingest and read information. The information could be in the form of documents. Let's talk about 500-page contracts, or it could be in the form of a set of emails where there's a conversation about that 500-page contract. It could be in the form of transcripts of customer service conversations. Actually, it could be in the form of actual customer-service conversations that have been recorded that then…

All of that can be ingested by the platform and converted into a specific language-based understanding of the 500-page contract with the emails, with the text conversion from the voice calls that then gets ingested. And all of that, put together, becomes the basis of them saying, "Here's exactly what the contract is saying, here's what the customer is saying about the invoice they sent you against that contract, or they received against that contract, and here are the two problems that invoice has that a human being would have taken forever to discover, and maybe the human being would never have discovered it because it's too expensive to apply a human being to discover that, and would have taken too long."

The credit underwriting decision example that I just gave, again, is another rich frameworks product build on that same platform that is offered to banks and so on. The contract discussion that I just teed-up is something that could apply to a range of industries. So, that's a company we acquired, and it's our way of bringing AI into our services that we offer our clients to help really disruptively solve for that competitive advantage in their marketplace.

Michael Krigsman: We have an interesting question from Twitter, from Arsalan Khan, who makes the point that combining AI with enterprise architecture for customer satisfaction and optimizing business processes is key. But, he says there is so much jargon to overcome. So, how do you do that; combine AI with enterprise architecture? That seems like a really interesting question to me.

Tiger Tyagarajan: It is a great question. And, part of the problem that we are all grappling with, Michael, is jargon that we all use, that our clients use; that our clients actually have to grapple with often that the industry uses. And part of the nature of business conversations is to de-jargonize that and make it real, and the best way we find, often, to make it real is to actually convert them into real examples that come alive.

So, to answer the question, let's go back to the example about underwriting that we just used. That live spread tool, the AI-based, that actually helps that underwriting decision is the AI-based tool and the point that is being made is that, alone, does not drive as much value as if you then combine it with a workflow that allows the tool to be embedded in that workflow that receives the application. […] The customer that I described is making a loan that then pushes it out ultimately to an underwriter because remember, the machine is able to let's say, take a decision, only for 80% of the cases. In 20% of the cases, the machine says, “I don’t know about the decision that I’m trying to recommend. Here is a recommendation, but I really want the next person to look at it.”

That has to then land on the desk of an expert, which means it has to go through a workflow that lands up there. The expert then makes a decision, and if the decision is not fed back to the AI platform, then actually, you're missing the final closing of the loop on the feedback so that the machine continues to learn. Again, you need that dynamic workflow, typically cloud-based, that has to feed that back into the tool.

All of that ultimately you want to capture as long-term data in an enterprise architecture that they can do a long-term analysis of how these things perform over the long-term because this could be a seven-year lease on that equipment. So, I think the point is… But, here's the advantage of today's technologies. They allow for APIs to be built that trade information with each other as well as with enterprise architectures. Those APIs are often open architecture. If you modularly build these, whether it's done on a workflow, it's the AI that I talked about; it's, let's say, presenting these, using great visualization tools that are available in the marketplace, again cloud-based, and presenting them to people to make a decision. If all of those are orchestrated in modular fashion that connect back to each other through APIs that connect the enterprise architecture, and by the way, that is capable of being changed as technology evolves…

So, let’s say there’s a Beta AI mousetrap that gets built, two years later, you can dismantle this one and plug back a new one; and if that can be done easily, which actually can be done because it’s API and open architecture, then you have a truly enterprise-ready, industrialized, capable of being rolled-out, scaled, with all the investments you’ve done being protected for the future. Because, you really don’t want to keep investing and saying, “Oh, two years later, technology has changed. I’m going to reinvest.” That’s a bad decision.

Michael Krigsman: So, let’s say that a company is looking… has a digital transformation effort underway, and they are looking at their business model. So, it’s not just about optimizing efficiency or making small, incremental innovation improvements. Is there a role for AI in this, or does AI at that point just become a distraction or red herring?

Tiger Tyagarajan: There is an incredible role for AI in actually creating and standing up new business models. I mean, to call them innovative business models, it often is an understatement because they completely disrupt the space that they could be stood up in. And most often, these highly-disruptive, innovative new business models have, as one of its core [components], using new technologies like AI.

Let me give you one interesting example. So, let’s pivot this to a different industry. Let’s take insurance, and let’s go back to the conversation around insurance claims and let’s take one that is very personal to all of us. We all are obviously big users of automobiles in this country, and we all have… love our cars and we obviously have to make sure they are well-insured. Every once in a while, we have an insurance claim that we have to file. The good way of doing that in these days, that a lot of companies do very well… I have a minor accident, let’s call it a “minor accident,” I stop on the highway, I make my 1-800 call, I then get instructions to what I’m going to do, wait there; pick up truck comes in; someone says, “We’ll get back to you with an assessment of what this is going to cost you.” We’ll also then make sure that the insurance company knows what this is going to cost you, and then there’s this whole process that kicks in in terms of coming back with, “I think the insurance company is going to foot $800.00 for this damage from your insurance claim.”

That whole process takes time, takes effort, takes multiple follow-ups. It’s not great customer service in spite of great efforts by most insurance companies. And, there’s angst in the whole process. You can have a new business model that I think many insurance companies are trying, and we are in the middle of experimenting this with some of them where, same accident, you open up your smartphone, you press the app, you open up the app, and then you just press the fact that you had an accident. Automatically, it triggers someone, let’s call it an “Uber driver,” who lands up at your spot in about five to seven minutes, takes out his or her iPhone, and starts clicking photographs of everything that he or she can see at that moment, in that car; inside, outside… And, there’s a set of instructions that that Uber driver has to do what he or she has to do.

Instantaneously, as those photographs are being taken, it lands up in some expert person sitting in Boston, who actually looks at it, and… Actually, before it lands up there, it lands up in an AI tool that looks at those images, applies computer vision based on more than ten million images that it already has, and says, “Based on the machine’s assessment, this is the kind of damage… These are the kind of repair costs that will be applicable, therefore, this insurance claim should be paid to the extent of $747.00.”

In some cases, 20-30% of the cases, it kicks off to an expert and the expert then takes a decision. Almost in fifteen minutes, the message comes back on the same smartphone, telling the customer that your insurance claim is approved, $747.00, go get your stuff repaired, and that's it." Think about the customer satisfaction, think about the retention of that customer. Next car that the customer buys, it's the same insurance company. Think about the value to the insurance company. I mean, you can go on and on. That is a disruptive business model. Any insurance company that finds a way to grapple with this, solve it, is going to disrupt the industry. It's going to change how that industry works.

Michael Krigsman: So, at this point, what you’re describing, and correct me if I’m wrong, is from a business model standpoint, the AI is technology that enables the company to do things that it couldn’t otherwise do and that becomes the platform or the jumping-off point for the business model. So, it’s not the technology in itself, it’s the technology in the context, “Okay, how are we now going to use it to solve a particular problem?”, which gets back to what you were talking about at the very beginning of this conversation?

Tiger Tyagarajan: So, Michael, I couldn't have described it better than you just did. It is exactly the way you described it, and the way we think about it is AI is amazing. However, it really becomes amazing only if two other conditions are satisfied. Condition number one, you must apply AI to the context of the business and the problem, and you must immerse it in the domain of that problem; in the domain of the industry. In that example that I described, do you really know what happens when an accident like this, that I just described, happens on an interstate highway in the state of Michigan, and the car is, let's call it a Honda Accord, and here's what happens to a typical Honda Accord in the state of Michigan, in that kind of an interstate highway, etc., etc., etc. And if it happens at twelve, noon, here's what to expect, etc. Here's what it means when an Uber driver lands up and hears how customers react.

So, in all of that, do you really understand the insurance business? Do you really understand auto insurance in the US, in Michigan? Do you really understand how a customer interacts during or filing of an auto claim? Do you understand how frauds happen in the claims process, in auto claims? Because frauds do happen. So, it's really the depth of domain; makes a real difference.

And, the second thing that makes a difference is, do you have enough data? Do you have enough data of past claims? Do you have enough data of all the images and photographs that are being taken in the past? Do you have enough data to actually create the intelligence that actually can be used? So, without the context of the domain, and applying it to data that has to be applied to that AI, AI doesn't come alive and doesn't create value. So, we are big believers in AI, plus domain, plus data. Is true value AI? Without domain and data, there is actually no value.

Michael Krigsman: Alright. So that, then begs the question, let’s say a company has identified a business, the kind of business problem that makes sense. They’ve looked at a big problem inside the organization as you’ve described earlier. Now, they’re doing proofs of concept. How do they ensure that they’re going down the right path, that they have the domain knowledge, that they’ve chosen the right domain, and that they have access to the data? How do they do this? This seems to me to be a very complicated set of interrelated problems or challenges.

Tiger Tyagarajan: So again, Michael. You hit the nail on the head! It is complicated, it is interrelated, and it does take time to solve all of these in a full-industrialized, fully-running way. One could call that “the bad news” because it’s big, and it’s hairy, and it’s complex. The good news is that you can actually break it down into its pieces and its components. You can hone it down to, “Can we start at the proof of concept?” And in that proof of concept can we actually test this quickly? And, by the way, I’m okay if I don’t get the full value on day one. That’s the good news. The second good news is that the value is so big when it’s fully done that it’s worth it, which is why it’s important to pick the problem, right?

So, pick a problem that’s worth it, hone it down to a proof of concept that you can experiment and then be prepared to iterate your way through a series of experiments. So then, it’s okay to not solve all the problems that you described on day one. You can solve parts of the problem, knowing fully well that you might get only 20-30% of the decisions that are actually taken right by the AI technology. But, that’s 20-30%, versus zero. And then, as you improve all the things you described, it moves on, and on, and on. And, in its tenth iteration, it’s already got to 85%.

The second problem that you've got to deal with is making sure that people buy into the fact that they're going to go through a series of experiments. And then finally, there's a third problem as you get through the series of experiments, and let's say it's successful, and you want to scale up? And this is where I think AI could fail within its application into solutions and in an enterprise. And then, by the way, this applies to almost all digital technologies. One of the failure points could be, and we have seen this in a few cases, where when you scale up, you need a governance layer that watches all of these technologies.

So, here's the way to describe it. If you have a thousand people whose job it is to do things, and they follow a certain set of procedures to do whatever they have to do, you normally have managers, and then you have people who manage the managers, and then you have a hierarchy. You have leaders who watch everything. There are measurements and metrics that people watch, and if some of those are not going well, then you say, "Time out. Can we all get together? We discussed the problem; we maybe retrained people. We make people understand what mistakes they are making," and then they say, “Go back and continue what you are doing because you just learned a better way to do it. I just taught you that." And that works! Unfortunately, it takes time, it's human, it's effort, etc. but it works.

Now, convert that into now you have an extra thousand people. […] You have people who are real experts, but you also have robots, and you have machines, and you have AI bots and tools. Fantastic! Something changes in the business, you have a new compliance policy. You have a new procedure. You have a new risk-based tool and you want to implement it. And you’ve changed something this morning. If someone doesn’t bring all the robots kind of back, and then go back into those robots and the tools and change things. If someone doesn’t watch all of that and govern all of that, you could have AI gone wild. You could have robots gone wild. And that’s not very good. In fact, that’s pretty dangerous.

So, I would say, one of the dangers of the paths that AI and digital technologies could take is an assumption that it doesn't require governance. It doesn't require it to be watched. It's all well-coded and done, and it will work forever. The fact is, business models change, policies change, regulations change, lots of change, change, change. And all of that has means that you've got to keep watching all the technologies that you laid out. That requires a governance layer. That requires almost a platform to watch all of that, and then when something changes, you go in and make the changes that are needed. We think that's very, very important and a lot of our journeys are around that topic.

Michael Krigsman: Wow. We're covering a lot of ground here, and we only have less than ten minutes left and there is a bunch of things that we still need to talk about that are important. And so, let's shift gears for a moment. I feel like we shouldn't shift gears because we could go on. This topic you were just describing really matters a lot, but let's talk about the workforce implications because it's a lot of fear around AI. And so, maybe please, Tiger, share your thoughts on that.

Tiger Tyagarajan: So, Michael, I think the fears are valid, real, and I think we should all think hard, debate, and that applies to enterprises, that applies to leaders in enterprises; that applies to governments, institutions, academia, scientists, research etc. It's a very important topic. It's an important topic every time there is a significant shift driven by new technology. So, I think it's important, concerning, and it's a valid concern.

But I'm going to start by saying we must have true optimism in the way technologies have worked long-term in the world. And we've had hundreds of years of new technologies coming in. The agricultural to the industrial revolution. There was a time when this country had, what is it 45% of its labor force in agriculture, today it is 1-2%, and that's been a transition. That transition meant a number of people transitioning into manufacturing and then subsequently into services, etc. Was that transition painful? It was. But did we manage that transition and ultimately, did the overall economy and people and quality of life improve? Yes, it did. And how did that happen? A lot of new jobs got created. A lot of new work got created. A lot of that, people didn't even know what those would be as that journey started.

So, the fundamental optimism that I think is needed, particularly from people who are in leadership positions in academia, in governments, etc., is that new work will get created as these new technologies come in. And, the overall well-being of humankind will improve. But here's the other problem that we must grapple with. Those changes took time to ripple through, and therefore, that time allowed that adjustment and transition to happen probably better than this transition may allow. This may be a faster transition and when you have fast transitions of skills and workforce that are needed, you have the same generation that has to transition into the new workforce. When you have a next-generation transitioning into the new workforce and new skills, that's an easier problem than the same generation transitioning.

So, you are really talking about how do you find ways to re-skill people? How do people actually get motivated to re-skill themselves? Where is the participation from the government, from universities, from community colleges; from enterprises, from corporations, from enterprises such as ourselves to allow people to train themselves more and re-skill themselves to provide the programs, to provide the time, and the investment dollars. And I think we all have to grapple with it and do it.

The reason for it […], the reason for optimism around something like AI, and the fact that we don't even know what it's going to create, is the following. One of the ways to think about AI is that AI is going to make predictions become so, so inexpensive that it will do two things. Any job today that requires some element of prediction, a prediction that this customer is going to be a good customer so, therefore, we must lend this customer half a million dollars. That is a prediction based on financial analysis, etc. AI makes the cost of prediction lower and lower and lower, almost to zero. But, there are a million things today that we wish we could use prediction, but we don't because it's just too expensive. All of a sudden, cost of prediction is going to be so small, that we're going to use prediction in almost everything we do. Everything mankind does. And AI will be used for that.

That will require a lot of people to actually do what I just described: Capture data, build domain, and… All of that will improve our lives, because think about being able to predict lots of things, whether it’s health, the impact of disease, drugs, genetic… I mean, the whole world is there. So, that’s the optimism. There is a reskilling problem that we all have to deal with.

Michael Krigsman: So, we have literally about three minutes left, and there are two questions still that I just have to ask you. One is from me, and one's coming from Twitter. So, if you can answer each of these in a minute, and good luck with that, but let's see. But number one, as you were describing your optimism, but the problem is that doesn't help. The rising tide of technology lifts all boats, yes, over time, but that doesn't help the person in the Midwest who is feeling the effects of globalization, and he or she can't find a job. And so, that's the problem with the "rising tide" theory […]. Any thought on that, in thirty seconds? [Laughter]

Tiger Tyagarajan: That is the problem That is the problem, and I was talking about optimism in the long-run, a real problem that you just described in the short-run and the medium-run. And the final statement that I’ll make is we all have to deal with it, and I don’t think collectively, we are spending enough time on it. The fact that technology is the disruptor, nothing else is the disruptor, technology is the disruptor that we have to solve for in the population that is going to get disrupted. Irrespective of optimism, they are going to undergo disruption. We must solve for that. We must solve for it through re-skilling programs. It is the fundamental problem that the world has to deal with.

Michael Krigsman: It is the fundamental... one of the key fundamental ethical issues around AI.

Tiger Tyagarajan: Yes.

Michael Krigsman: And then we have, last but not least, and I’ll ask you to answer this very quickly, a question from Twitter. Gus Bekdash is asking, “What insights do you have on integrating analytics and AI?” So a more practical question, and very quickly, please.

Tiger Tyagarajan: No, it is actually… By definition, AI is completely integrated. Analytics with the process that you're trying to solve for with the answer that you want, with the outcomes that you're trying to predict, and I said, "AI is all about using data and domain to predict," and it is predicting it real-time. So, the best way to think about AI is that it's real-time predicting. What is the right answer to take, every moment, at every second?

I think one of the best examples of that is [..] every step of the way, it’s saying, “Take this because this is better traffic...”, no, no, no, “Take the left turn because that is better traffic.”

Michael Krigsman: Okay. We are unfortunately out of time. This has been a very, very fast forty-five minutes. Tiger Tyagarajan, thank you so much for being here with us!

Tiger Tyagarajan: Thank you, Michael! Thank you. Really enjoyed it!

Michael Krigsman: We have been speaking with Tiger Tyagarajan, who is the CEO of Genpact, which is an almost $3 billion professional services firm in technology. You've been watching Episode #246 of CxOTalk and next Friday; we're speaking with the Chief Data and Cognitive Officer at Royal Caribbean Cruises, so that's going to be a really interesting one, as well. So please, join us. Thanks, everybody and have a great weekend! Bye-bye!

Published Date: Jul 21, 2017

Author: Michael Krigsman

Episode ID: 449