Workday Podcast: Tech Leadership in the Age of Generative AI

Jeremiah Barba: Generative AI tools like ChatGPT & Midjourney may be taking all the attention lately, but the implications of generative AI for business leaders are far beyond chatbots and image generators. Tech leaders in particular are paying close attention and working hard to find ways to harness the power of Gen AI for decision-making and efficiency. Today on the Workday podcast, it’s a conversation between Joe Wilson, CTO of Americas at Workday, and Rob Vatter, Executive Vice President of Enterprise Platform Services at Cognizant. Hope you enjoy the episode! 

Joe Wilson: All right, Rob, let's go, brother. This is going to be fun. So could you share a little bit about yourself, your background, and your work at Cognizant?

Rob Vatter: Yeah, thank you. Thanks for having me. I work at Cognizant. I run our enterprise platforms business. I've been there for about four years, been in the industry for over 30. So I bring a lot of experience across many aspects. On a personal note, I'm a single father of five, and live in the smallest state in the union, otherwise known as Rhode Island.

Wilson: Good grief.

Vatter: Yeah.

Wilson: A father of five, you must be busy.

Vatter: Fairly busy.

Wilson: All right.

Vatter: That's why I use generative AI.

Wilson: Oh, that's a great segue. So let's actually dive into a little bit about that. So I think it's fair to say that, unless you've been living under a rock, or maybe a single father of five, you probably have heard about generative AI and all the things that are happening in the market around it, lot of buzz, a lot of hype.

Vatter: Yeah.

Wilson: A lot of it's real. Some of it, who knows? So let's start here. Why is the hype so strong around generative AI specifically, across, I guess, all parts of industry or all markets?

Vatter: Listen, I think the major thing that has occurred---because AI has been around for quite some time.

Wilson: Yeah.

Vatter: Whether you want to call it machine learning or AI, it's probably been in the, the market for 15 to 20 years. The fundamental difference and what's changed was last November when there was a release of ChatGPT and the foundational model that went with it. Effectively, what it did is it democratized it. So it made it easy for people to use. And now people are seeing the dream, seeing the promise, the use cases of what this type of technology can do. And it's very different than coding, because of the nature of, of the prompts and the text. And so it's really made it human. And so people are actually starting to use it for things, around creating a resumé, or students using it to write their essay, or doing analytics and insights into, tell me the best place to go to surf on November, right? Just a lot of things that have made it real and made it really democratized is what's really changed the nature of the field.

Wilson: I think back to the previous hype cycles, whether it's RPA or blockchain, I personally don't know a lot of people that had a blockchain server sitting in their closet at the house with, you know, 30 fans pointing on it. But to your point, everybody can get in and play with these commercially available LLMs out there, right, ChatGPT or otherwise and do wild stuff. So what's the most wild thing that you've done from a prompt and an outcome perspective?

Vatter: Well, me personally, not a whole heck of a lot, other than a lot of analytics, but our teams have done some pretty interesting things---prediction models on where COVID would go to and how to get ahead of the spread of it. So even back two or three years, even though it wasn't ChatGPT, we were doing those type of models to do those forecasting. And those have real impact on society in terms of, not necessarily solving problems, but maybe preventing problems. I think the most interesting thing that I've seen from a ChatGPT, is really just some of the common stuff. We've been using it with clients on creating content for marketing. So think about images, think about advertising templates. It's really an interesting thing to do because it lets people like you and I play in that space a little bit and create our own, advertising, for lack of a better term. So those are the things that I think are out there. You'll see other stuff in terms of the pharmacy world, people using it on cancer analysis and so it's developing as we go. But those are some of the more interesting ones that I've seen so far.

Wilson: All right, how about on a personal level?

Vatter: A personal level, I wish I had one that was actually a good use case for that, but---

Wilson: No, no haikus?

Vatter: What's that?

Wilson: No generating haikus or poems about your dog? No?

Vatter: No, no, no. I think maybe solving personal problems would be one that I would like to do, but I'm afraid that if I put it in there, I might actually get an answer.

Wilson: Yeah.

Vatter: I haven't done it. I haven't done anything really on a personal level at this point.

Wilson: All right, Rob, be brave. You can do it.

Vatter: Yeah.

Wilson: My 78-year-old father actually wrote a prompt to take the ingredients in the fridge and make a recipe out of it.

Vatter: Oh, very nice.

Wilson: Yeah.

Vatter: And what was the recipe?

Wilson: I have no idea.

Vatter: Oh, there you go.

Wilson: I asked him if he liked it. He said, "No."

Vatter: All right.

Wilson: All right, we'll see where this goes.

Vatter: Well, that's creative.

Wilson: Yeah, no doubt. So I mean, as, as we kind of move past this hype cycle, and I completely agree. It's really the accessibility, right?

Vatter: Yeah.

Wilson: This is the first time in a long time that the lay population's had access. But now organizations are starting to really look deeply at the opportunity, right? It's moved past the kind of general pop type experience.

Vatter: That's right.

Wilson: And now we're thinking, what can we do to leverage this to solve business problems? So when I think about that, what are some of those real benefits that you think are available to the enterprise by incorporating this tech today?

Vatter: I think they'll start with smaller use cases---but examples that I would use, let's take HR, for example, and Workday. One of the things that we always struggle with as enterprises is what are competitive salaries based on location, MSA, country, state, city? If you think about the data that sits in a Workday platform, getting underneath that analytics now is going to be a lot easier. So instead of having to go outside to one of the consulting companies that might provide you that data, it's going to all be available to you. So you'll be able to have that prompt. That saves an enormous amount of time. So that won't necessarily show up as hard dollar savings, but from a productivity standpoint, enormous savings. If you go into the world of sales ops and sales operations, the ability to forecast and predict closing and bookings on a quarterly or monthly basis. You know, today, a lot of people use systems to do it, but then they load it back into Excel, and then they apply their own sort of measures against it.

These large language models will allow us to start really being a lot more accurate and bring a lot more attributes in, in terms of the data sets. So our ability to forecast sales will become a lot more accurate. You go talk to any sales leader, they'll tell you, if we could get more information on that and be more accurate into our forecast process, whether it's for sales or even revenue, or in the case of creating, where do we actually go invest in terms of opening up a new delivery center by using some of the HR data, I think those are going to be the real use cases you're going to start seeing and they are rolling out already. People are trying it. Now, are they actually using it in terms of operationalizing it? In some cases, yes, but I think they're starting to actually get underneath it and use it. That's where I see this starting, and particularly because those use cases. The way I describe it to people is, since we're in the early innings of this, you want to use use cases where I affectionately say, people aren't going to get hurt, right? Bad things are not going to happen. Right? So what if you get wrong, you know, the sales forecast? Well, that happens anyway. I think when we start getting further into, say, in the medical world of diagnosis and treatment, now you're starting to get into an interesting area. These issues of bias and accuracy of the data sets, and, how real is it, become really important items to discuss.

Wilson: Yeah, no doubt. It's interesting because you talked about, like, we're in the safe space right now. We're just kind of getting started. I loved your examples of, like, content generation, of maybe summarization. Those are really impactful, they kind of lead towards this notion of automation. I've never met anybody in my entire professional life who says, "I got everything done today at work." So maybe there's an opportunity to kind of maybe have a copilot, right, to assist me as I move forward and do these things from kind of a human-centric, but also being able to capitalize on some of that tech to do cool stuff. So when you think about that then, and you're thinking about the power of automation, with IT typically owning that orchestration, what do you think it's going to require for us to lean on these technologies with a level of assurance or trust, especially as we're thinking about these, more sensitive use cases, in terms of business?

Vatter: One thing that generative AI is doing and forcing is a change in the way we work and therefore a change in the people who are making decisions and using it. So to your point around IT being the orchestrator, the reality is if you think about what's really happening, the types of people and the types of skill sets are changing rapidly, meaning that, I was with somebody last week who’s an expert in gen AI, and their comment to me was, and kind of funny for anybody who has an engineering degree, maybe not, that this is the revenge of the liberal arts education. In the fact that now when you have these prompts and developing these prompts, you have to understand what they do and what you're trying to accomplish. So to your point, these more sensitive business cases and how they get applied, become very interesting. Because you're right. these co-pilots fundamentally will take over some of the kind of repetitive tasks and things that you can, for lack of a better term, run in the background, kind of like a software program, and I can do other things. Are we going to lose jobs? There is no doubt there will be efficiency gains across the board in all areas relative to some of these administrative processes or handoffs in humans. However, if you think about it in the context of creating this efficiency and releasing that value back into the system, that value will show up somewhere. I don't think anybody's really nailed it yet in terms of what the monetization of that value will be. Will it show up in more acceleration, therefore maybe more products to be sold in the market? I'm not sure. And I think it'll depend on industry as well. But there is no doubt, I think, the way we describe it at times, it’s the creation of the new cotton gin. All of a sudden you've got another hyper growth in industrialization because of the acceleration and the release of productivity into the market. Will we have skills issues? I'm sure we will, right? But I don't think there hasn't been a time where it's caught up. But I would ask this question of everybody, if you look at any of these technologies, whether it's the internet, whether it's the cell phone, whether it's the cotton gin, there's usually a 20-year cycle before it actually gets to maturation and tipping point. It took a long time for the internet to really explode. Right? It was there for a long time. DARPA had created it. It was hanging. Then we had our first UIs, right? And then finally, Google came in and it just kind of blew the market right open, right? So I think, I think that's what's going to happen here.

Wilson: I think about the opportunity to automate, what might seem like a very simple task, but it’s a time-robbing task. So within the last year and a half, I had to write another job description for a job that did not exist within our catalog. It would have been so epic just to press the prompt. Right? And in fact, now that I know that we can do this, I'd probably rather chew on broken glass than actually write another job description. Right?

Vatter: Yeah, listen, the whole job description thing, for anybody who's involved in that space, right, you know, somebody takes the first draft, then somebody takes the second draft, then somebody takes the third draft, then somebody says, "Wait, this draft is completely ridiculous. Let's go and create a new draft." Right? So I agree. Those type of things, and again, you'll be able to write the job description very personalized to an area, to a region, to a particular state or country. I think this becomes really important and really efficient in terms of what we need to do.

Wilson: Yeah, thinking about the macro, and if I look at the time that I spent doing it, and I'm one person, and I amplify that across the organization, and then across all organizations, how much time can I return back? It's going to be phenomenal.

Vatter: Well, yeah. To be a little facetious, too, how much time can you get back from procrastinating before you actually do it?

Wilson: Oh, now, now you're blaming me. You know me too well already. So you alluded to this concept of the robots taking your jobs. But to your point, I don't think there's been a cycle, whether you want to call it hype or otherwise, that hasn't offered the opportunity for people to upskill, or do something different or more meaningful. So when we think about this, this notion of, like, maybe calming the nerves of a general population who thinks, I'm losing my job because of AI, because it's just not true, how would you answer them and then provide them a level of assurance that things will actually become better through the introduction of these technologies?

Vatter: Yeah, there are really two parts to that, right? And I think you hit on the communication aspect of it, is to make it real. You just used the use case of creating a job description. Across all the spectrum of jobs, everybody's got elements of their job that they wish they didn't have to do, and would love for somebody to take it over so that they could actually go on and do something different, right? And I think that's the first step is to get people to recognize that it's an efficiency gain that doesn't necessarily take away from them, but will allow them to grow. And by the way, that's not the easiest of communications, right?

As you can imagine, some of the examples I give that are not necessarily gen AI, but, you know, at one point, somebody dug a ditch. Right? Then at some point, somebody said, "Hey, why don't we get a horse and put this thing on the back of it and let it dig the ditch?" And then all of a sudden, engines and hydraulics, and now ditches are done in a much more rapid time. Yes, did the ditch digging stopped for the individual in terms of a shovel in hand? Yeah, the answer is yes. Did they go on to other jobs like carpentry or electrical? Probably yes. So while they got displaced at that moment, they moved up to the next higher order. But let's not kid ourselves. In that moment, it's an emotional issue for anybody who's being impacted. And I think people just need to be empathetic about that and understand it and communicate it. And then secondly, we do have to set up training. I think the, the one area that we need to be conscious about the skills people are going to need in order to operate in this new economy, having co-pilots. Communication is going to be a huge part of it. I go back to the prompt. It's a communication prompt, right? What you tell it and what you're asking it will determine where it's going to go and how it's going to search. So how you write matters. That aspect of changing how you think about things, I think it's going to be enormous. This notion of robots taking over, I'd put it in the category, let's make your life easier so that you can do the things that you want to do.

Wilson: Okay. I just want to make sure that everybody listening knows the robots are not taking your job. It's going to be OK.

Vatter: It's going to be okay. And, and candidly, if you're in the right role, your job actually might get a lot easier, you know. And--

Wilson: Yeah, I like that. That's glass half full.

Vatter: Yeah, and you won't be frustrated with things. We're half full, not half empty.

Wilson: Yeah, right on. That's great. You touched on the concept of organizational design actually changing as a result of this, right---not just in terms of skill displacement or opportunities to generate more, but maybe just more seismic shifts in terms of how we structure today. Any insight on where you think that might have the most impact?

Vatter: it's been coming for some time that, if you think about the traditional enterprise, it's structured in silos, functionally, right? So you've got HR, you've got finance, you got product, you got delivery, you got sales, you got marketing. And it's been slowly eroding. And now this is kind of accelerating that erosion because these type of models and then the applications that'll be on top of them, they don't care about these divisions that are siloed divisions. It's going to go straight across to solve a problem. If you take a use case like claims adjudication, if your prompt is how do I improve claims adjudication, it's going to look and design a process that starts from the person going to do something with a doctor or whatever it may be, and then the claim being adjudicated. So its horizontal journey is going to touch every aspect of what I'll just refer to as the supply chain and come up with a solution. And it's going to tell you that, hey, I don't need these other things in here. So you're going to start to see these organizational barriers come down. Now, will there still always be things like finance and HR? There will be, but it's going to force decision-making at a different level where the business owners will become more and more involved in making these changes, regardless of these underneath organizations like IT. How fast that goes, hard to say. I think some companies will move rather rapidly. I think some companies will move rather slow. But it's going to happen. There's no doubt about it.

Wilson: Yeah, so a flattening of the organization, but also maybe an interdisciplinary approach in terms of how we collaborate.

Vatter: Yeah, we've had the conversation, and this is still so early-innings, is any point of view wrong? Probably not. But if you think about how organizations are structured, not just from a functional standpoint, but from a hierarchy standpoint, everybody looks for that pyramid structure. We've had conversation, does that pyramid model change? Does it stay a pyramid, but is it a shorter pyramid? Right? Because maybe the bottom is done by ChatGPT. Is it a diamond, start at the top, but you also come down to the bottom, so you're bringing in people slowly to get to the middle of the organization? I think the notion of talent strategy and how you recruit and the types of people you recruit is going to be probably one of the larger challenges/opportunities that everybody's going to have in terms of how do you, do you run an enterprise. I don't think there's a single answer, but I do think it's going to help us solve some of the issues that we've been dealing with forever as an enterprise as we've evolved.

Wilson: Right on. Yeah, it's, it's interesting, right? Because you have the opportunity to be both futurist and weatherman here. Right?

Vatter: Yeah, exactly.

Wilson: It's going to be great.

Vatter: Well, it's partly cloudy.

Wilson: With a chance of meatballs.

Vatter: Yeah.

Wilson: All right. So we've talked about a lot of the opportunities, and they are a-plenty. And I'm just as big on the hype as anybody else. I think there's a real opportunity to do meaningful things across the entirety of work when it comes to these types of tools. Let's talk about risk. What are some of the things that you're thinking about there?

Vatter: The risk, back to my earlier point, is when you start moving these use cases into impactful spaces that impact people. So let me give you a simple example, if you use a ChatGPT or a model on supply chain, and let's say the data sets that were there weren't necessarily correct and maybe a little biased. And let's say that meant that insulin shipments didn't get to Finland for a month. You're actually having a big impact on society. So those type of risks that get introduced into the system---not maliciously, by the way---to understand the consequence of what might happen and all the downstream effects is enormous. If there's anything we know in the world we live in, is there's tremendous amounts of data out there. By definition, some of that data is biased, just because. Again, not that anybody's doing anything maliciously. It's just that the data's got some bias in it. So you--

Wilson: Or just plain wrong.

Vatter: Or plain wrong, right? It got collected the wrong way. So that piece of cleaning up the data sets, if we go back to point about risk, that's a big part of it. But it's when you apply it into specific use cases and trying to understand the outcome is where it gets really real, because there'll be unintended outcomes. The world is so networked together now that if you make a change in Sri Lanka, you may not think it's going to have an impact in Podunk, Iowa, but it actually may have an impact in Podunk, Iowa. And so those are the types of things where you have to be able to connect all the dots, and see what those use cases possibly could end up being. Not an easy thing to do.

Wilson: Yeah, especially the folks in Podunk, Iowa.

Vatter: Yeah, those poor guys, I guess.

Wilson: Well, I've heard it's a lovely place this time of year. It's interesting because we know that risk is a-plenty. We know it's going to be there. We're going to remediate or mitigate against it. What are some of the things organizations should be thinking about today to really hedge against that risk as they consider these tools?

Vatter: The first thing they need to be thinking about is the governance process, how it gets implemented and where it gets implemented---and where the data sets live. Does it live in production or non-production? You got to get the governance model set up. Because it's still nascent. And as you've been on the hype, I think my favorite one is hallucinations. Talk about scaring the world, of all the things to call something is a hallucination, and it's because of the data sets make stuff up. It's a great marketing gimmick, but it also scares people. So I think governance is the number one thing, and who gets to use those data sets, and then how do you test for it? Who is responsible? No different than compliance in some of the industries that we have today. How do you set up a set of compliance and regulatory activities to really govern the use of that? It's going to be staged too. It's going to be phased, and it's going to have different cohorts. Back to the use case, if you're doing creation of marketing content versus analytics on genetic information to create the next vaccine, the governance models don't necessarily need to be the same on them. So that's the first piece. People need to recognize that there is the potential that, when these models get created and they start to go into production, that there could be outcomes. So that's why they have to put the protections around them.

Wilson: No, it makes complete sense. You know, I think too, around this idea that as people approach new technologies, sometimes they go into lockout mode---because they just don't know what to do. Or a few things actually kind of escaped, and so they say, it's over. So I'm not going to name the name of the company, but we heard about the big IP loss as people were interacting with this commercially available large language model. And you think about the money that was lost in terms of intellectual property leaving the organization. But two, I think the governance models that are going to be available to be implemented and to be considered maybe predicated by domain, because they're not going to be all the same.

Vatter: Yeah, and they shouldn't be.

Wilson: 100%. But it's going to afford the opportunity to look at these things and then do them with enough mitigation in mind that you can feel at least safe as you adopt these things to do good things into the future.

Vatter: Yeah, a very fair point.

Wilson: Which brings me to this then. Where should IT leaders start, like, today, if they want to harness the power of generative AI? And what do you envision the future of work looking like through that adoption today?

Vatter: From an IT leadership standpoint, I think you got to start with small use cases. I think you need to prove it will produce what you set it out to produce. It could be small use cases like we were talking. It could be job descriptions, right? Create the small use case, get people to see that it's actually productive and can work, and then start building around it. And I think IT leaders need to look at it more from a use case and adoption than necessarily a technology standpoint, because we in the technology industry always rush to put the new technology in. But the reason the technology there, is there is really only from an enablement standpoint. Right? We're just using it to still get to the same outcome. I need to get a job description done, all right? Old days, I used to write it on paper. Then I do it with a computer. Now I'm going to do it with ChatGPT. But at the end, it's still a job description. Right? I got it done. Instead of taking two days, it took me two minutes. It's that type of thing that, that IT needs to get organized around and set up and then use those principles for moving forward. And then they can incrementally build on it. So as they get more comfortable with a use case, say, around job description, maybe the next one is supply chain prediction, right? Maybe the next one is next best action for a consumer. And so you can start building on these, and then you get some more of the difficult cases, but you'll have built up a catalog of applications that you've built that give you the confidence, and you'll learn from it. If you make a mistake, it's not going to be that dreadful mistake. if there was one takeaway I would have for people is, think about it in that context of, as you learn, stay away from the dreadful mistakes early on, and be focused on proving it into the organization, and then learn your own governance model, because each company has a different culture.

Wilson: Sure. And so it doesn't have to be a big bang. You can let the small wins generate to big wins over time.

Vatter: We have customers who are running 500 proof of concepts, right? They're small proof of concepts. Some of them are as simple as natural language program talking so that, a CFO can ask, "What is my current inventory of sneakers in India?" And ChatGPT will come back with it, right, and then do it in language. It's those little types of things where, if you think about things that we perform across our different functions, those are, those are natural areas that we can progress into.

Wilson: Do you think this is an opportunity for maybe borrowing on the axiom of business-led IT enabled to actually show real fruit?

Vatter: Yes, I do. I think that is a very truthful statement. I think it's one the IT industry in particular tends to forget sometimes, the way we talk about it all the time is business outcomes have remained the same. The only thing that's changed is how we do it. And if you think about ChatGPT, it's another tool in the quiver to still accomplish what we're set out to do, whether it's a job description, creation of an invoice, a collage of pictures that's beautiful to give to your grandmother. Each one of these are things we've always done, but how we do them is what's changing over time. And that's really what ChatGPT is doing is it's helping us, to your point earlier about RPA, RPA was automation. Automation for what? To automate tasks so that you could actually still get to the same outcome. I think if people can put things in that context of, what is it that you're really trying to solve for, like claims adjudication, if I can take a claims adjudication from 30 days to three minutes, what does that really mean? Customer experience goes up, cost to serve goes down, and all these different types of things that contribute to the expense of medical care in the US, which is at 19% of GDP. If you can fix some of those administration processes, it will take a lot of cost out of the system. And what does that mean? It means better care for people in the long run, because that money can be reinvested in other areas as opposed to administrative tasks. Those are the things that I think, are really the exciting things that are in front of us with this. But that's just my opinion at the moment.

Wilson: No, Rob, you're glass-half-full, I think it's important. We need to all stay that way.

Vatter: Yeah.

Wilson: That's great.

Barba: You’ve been listening to a conversation about generative AI and tech leadership with Joe Wilson from Workday and Rob Vatter from Cognizant. If you enjoyed what you heard today, be sure to follow us wherever you listen to your favorite podcasts. And remember you can find our entire catalog at workday.com/podcasts. Have a great workday!