Why Most AI Strategies Fail Before They Even Start ft. Zoher Karu

Hello, we are Bringing Data and AI to Life, a podcast by Informatica. I'm Amy Horowitz, our VP for solution Sales for data integration and data governance. And I'm Nick Dobbins, worldwide VP and Field CTO here at Informatica. If you've ever felt lost in the chaos of data and AI, you've come to the right place. We'll be conversing with industry experts who are here to shed light on the challenges all rocked up within these arenas. We're here to bring clarity to the chaos myth, busting the confusing parts and providing insights and guidance complex data problems by delivering trusted data for analytics and AI. Welcome to today's episode of Bringing Data and AI to Life. So today we are joined by one of the most experienced and respected leaders in enterprise data and AI. So Harry Carew has spent his entire career helping some of the world's largest organizations unlock the power of data at scale. So he's led global AI and analytics teams across industries, from building real time fraud detection and digital banking at its Citi to transforming healthcare as the chief Data and Analytics Officer at Blue Shield of California. So with a PhD from MIT in recognition as one of the top 100 data IQ leaders worldwide. So Hare brings a rare combination of that deep technical expertise along with the strategic business insight. So he's known for making complex ideas really simple and showing how AI can drive real impact across the industry. So here, welcome to the show. Really happy to have you on. Oh, thanks Nick, for having me. Excited to be here. Yeah. So, you know, we'll start off with a softball for you as we go through, you know, hop right into the big key topics. You know, you've led data and AI transformations across the industries. We talked, right. Finance, healthcare, retail, and now startups as we go like. So as you take that look back, what's that defining moment where you realize that there's just this immense power in data in AI that really could drive some dramatic business outcomes? You're right. Data is sort of the foundation of everything. And there is a lot of power in data if people can think about how to capture it, organize it, use it. I think that, I mean, there's a continued explosion of data in the world. And some of my even past jobs were things like analyzing walking patterns of people shopping inside of retail environments or analyzing conversations occurring in call centers. And all of these things are data. But you don't realize it's data unless you capture it right. Do most people turn right when they walk in or left when they walk in or how long they spend in aisle four, and all of these kinds of things. Nobody was really capturing that data. But there's data all around us, and the explosion in sensors and everything else, and some of them you're even wearing on your wrist are just exploding in terms of data that's being captured. And so the real question is, how are you going to harness all of this information and take advantage of it? And AI in its current form, with the large language models, et cetera, has burst onto the scene only in the last couple of years. But it's really just another tool in the toolbox of trying to take advantage of all of this data. And the conversation has accelerated, I think, in the last couple years, because, well, the way I describe it is the Internet existed for a long time before the web browser showed up, but then suddenly it accelerated the usage. AI has also existed for a long time before, like ChatGPT or your other favorite showed up, but now it's become more accessible, just like the web browser made the Internet more accessible, AI or generative AI more accessible. And so now everybody's trying to take advantage of this new tool, but at the end of the day, powering all of this is data and the way it's captured, organized, and used. And so the aha moment is really the opportunity to take advantage of larger and larger volumes of data that were harder to do before. I'll give you one analogy. Like if I say, take out a sheet of paper and pretend you're driving a car, and I want you to write down the rule of when to press the brake pedal break down, okay, traffic light is red coming up or stop sign, or there's a curve in the road, or it started to rain or the sun was in my eyes, or the whatever. The whatever. And there pretty soon you start to realize there's a lot of possibilities of when to press the brake pedal. And you will find it's very hard to write down the rule of when to press it and when not to press it. But yet everybody who's driving a car knows how to do it. How do they know how to do it? And that's because they are processing a tremendous amount of data, more than you can consciously think about. And you're processing all of these inputs simultaneously, and you're reaching this decision. So AI is a tool and capability to process large volumes of information from lots of different sources at the same time, more than kind of traditional 20 variable predictive models or something have, right? And so that is what is kind of driving this newfound power, largely, it's actually a really simple analogy to think of. And yeah, it is. Right, right. If you literally had to sit there and discuss every permutation of why you would make a turn or hit a brake pedal, it would take forever. And yet autonomous cars are doing it all over the place now, instantly. But, you know, and as you say there. Right. So now you've seen a lot of different organizations as you go through, when you kind of do that evaluation, as you go into these organizations and you think about, all right, are we ready to embrace AI? Right, because there's more to it than the technologies available. Right. Is the data ready? Are the constructs ready? Are people ready to race? What do you look at when deciding is this company ready to take this step or not? And, you know, then communicating out what it takes to be ready when maybe they aren't? Yeah, I mean, I'm sure the answer differs slightly by the scale of your organization and how many different sort of pockets there are, if you will. But in general, I would say, look, data is the lifeblood of an organization. It's literally like the blood in your body, right? And different organs use the blood in different ways. And everybody wants access to the same data. You don't want different versions of your blood. Want dirty blood. You don't want any of that. Like, you want all of this flowing easily, seamlessly between the different parts of the organization that need to use it. But most corporations don't operate that way, right? They operate in their own silos and say, well, I'm only going to do what I need for my part. I don't care what somebody else needs. Right? And it's a normal human tendency. You care about what you care about and you don't think about necessarily what other people care about. But the truth is, if you broaden your lens a little bit, everybody benefits. Like, I'll give you an example back from, like, my banking industry days. I remember asking somebody, I was based in Singapore, so it was like I was in Malaysia. I'm like, okay, how many loans did we sell yesterday? So apparently the first challenging part of that sentence was the word yesterday because they only used to get a monthly report. I'm like, all right, well, at least go find out how many sold yesterday and then come back. We sold, whatever, 75 loans yesterday. I'm like, great, now give it to me by time of day. Like, we don't have it by time of day. We didn't capture time of day. I'm like, okay, you don't need it to calculate your salesperson's commission. I get it. But call center needs it, marketing needs it. Like all these people need it. Why didn't you capture it? Right? And so people don't think about broadening their lens of like, what data is needed, how could it be used? Right. They're only after the task at hand. And if people just took a step back and thought about that, that would benefit everybody. And also you really run into single source of truth problems. So the things you look at is, where's your data? Is it accessible? Do you have more than one version of the same thing? And all of these kinds of questions start to force yourself to examine, like, what are you really doing? Because with a tool like some of these gen AI tools, you'll just make bad decisions faster. Like this is not the goal. Right. And so yes, it improves a lot of data wrangling and whatever else, but if it's sort of dirty or you don't have a good semantic layer on what does the data mean? These things will love to interpret their own meaning of the data and give you answers. And if you're not careful, you don't know if the answer is right or wrong. Like you don't know what is going on. Right. And so many organizations, you'll go like, for example, like back to the same example. When I was in Singapore, I was doing data analytics across multiple countries. Every country had their own definition of what an active customer is. Okay, it sounds so simple. How many active customers do you have? Well, every country has their own definition. You're not going to get the right answer. Right. And so trying to clean that up, basically, data governance has become sexy again. All right. Is the way I'd phrase it before, it's like, oh, that's just whatever janitorial work. I'm like, it is definitely in the trenches, but it is what is the foundation of using this information to drive value. So where is it? How's it captured? Who owns the definition? Do you have single source of truth? Like all of these sort of quote, basic questions are an indication if you're ready to put powerful tools on top of that. Right? Yeah. No, I mean, yeah, I know we're talking to a lot of customers now that they're trying to get into what to think about with agentic AI. And it's one of the topics we're having there too is no, you want great data for it, otherwise it's just going to use the first data it finds. Right. Like an agent's job is to get the job done not to evaluate do I have everything I need to do it. It's just going to find what it needs and it's going to do it. Right, wrong or otherwise. Yeah. So you know, outside of just this scenario, I mean, because they are rushing to judgment, right? Like they're rushing to have to do this AI craze, right? Like boards are telling people, you know, we need AI, automation, AI built into things, but they're not ready. So I mean, I think you articulated a few. But what are some of the other just common mistakes that you see organizations making is they go down this push to be, you know, AI first AI ready. Well, another sort of common mistake is this belief and you can hear it in the hallways many places. I mean, maybe less so now, but for sure, when things are bursted unseen, it's like, we need to be doing some AI. Well, what does that sentence even mean? Like, right, we need to be doing AI. And there's this fear of falling behind or fear of whatever, right? And so sometimes from you'll hear it circulating among the top executives like we need to be doing some AI. What are my AI use cases? I'm like, look, the mistake is AI is a tool, but what business problem are you actually trying to solve now? Always starting with the problem. You can lose sight of that because you can get shiny object enamored with the technology, like wow, look at this. You can type in stuff and get answers like, okay, but what business problem are you trying to solve? And work backwards from that. Because yes, AI might be a tool to help you solve that problem, but it doesn't have to be like you can solve it lots of different ways, right? And so people lose sight of that and they get spun up around pilots and demos and oh, isn't that cool? And whatever else. And you lose track of what you're trying to solve. And you see it in various published sound bites like 95% of AI use cases fail or nobody's getting ROI or whatever is going on, right? You see a lot of that circulating and that some of it is the data problem we already spoke about, but some of it is what are you trying to do, right? And the power of AI, it certainly streamlines and automates a lot of different steps in what used to be manual processes. And you will get benefits from like hey, the agent used type of notes after the call. Now this thing summarizes the call automatically. Okay, great, right. And so you can take existing business processes and make them faster, better, cheaper for sure. But the real value Honestly, behind AI and agentic, AI is not taking existing processes and making them faster. There is benefit to that, but it's really reinventing processes. And so like this phone call. Yeah, I can make the phone call better by taking notes or whatever. But the real question you should ask yourself, why am I even taking a phone call? Like what is happening? Right. And so can I redesign processes in the company? And that requires much broader thinking, systems thinking. And I don't think at least the current state of technology is going to redesign processes for your company. Right. You can deploy agents to execute pieces of processes or even coordinate across a few pieces. But humans are still needed to think through the right way to solve the problem, not hell just do this. You know, instead of a handsaw, I'm using a power saw. I'm like, okay, I got it right, but how do I really think through that? And I think that's what we're all living through right now is really trying to harness this in different ways. I mean, the first phase is I'll take what I was doing and instead of somebody reading the page, the machine will read the page. Okay, fine, got it. Right. But now what? Right? Yeah, it is interesting and I, and I agree with you. I mean, I think we testified to this. You know, there's the efficiency projects, there's the high value, like transformative projects. And then when I say, you know, a lot of them are just the resume building projects, like how can I go out and do something cool to throw it on my resume that I did it and go. And again, that has value too, right? People are learning and pushing themselves and going with it. But maybe like you've actually worked in a bunch of industries that are kind of fascinating to me and I think are interesting how AI will adopt into it as well. Right. And like I'm always fascinated with healthcare one because I'm amazed at what our healthcare system can do. I mean doctors and the evolution of medicine is just crazy to me. And now with the notion that AI can help make this even better. Right? Read things, advise things, speed things up in terms. But I have to imagine, right, when you're in the crux of it, designing systems to make decisions that are life and death, like the pressure that goes onto it, the challenges that go with data that falls into that. How have you describe a little bit of that? Just with what are some of those challenges? And how'd you take one? Hey there listeners, it's Nick. I hope you're enjoying this episode. I wanted to let you know I about A special upcoming event Informatica is having. So the incredible Informatica World, our annual premier AI ready data management conference. So this year we're coming live from Mandalay Bay in Las Vegas on the 19th to the 21st of May. We'll have keynotes from industry leaders, we have a bunch of technical tracks, AI data initiatives, and of course, the Informatica Innovation Awards. It's not to be missed and we hope to see you there. Thanks for tuning in. Enjoy the rest of the episode. For sure. Healthcare complex beast. And there's many pieces of the healthcare system, whether that's providers, doctors that provide care, whether it's pharmaceutical side, whether it's the insurance side. There's many pieces to the puzzle. But there is a problem. One of the challenges in healthcare is disconnected data. And that's not just disconnect. And we talked about silos within a company. The problem with healthcare is not just silos within your company, it's silos across the ecosystem. Like, how many times have you gone to a doctor and they ask you to fill out a medical history? Have you ever asked yourself why you're filling out a medical history? Like 19 times in your life? How come they don't know this information? Right. And so that's because everybody only has a piece of the puzzle. And if the people had a more complete view of you, they could take better care of them. But they don't have a complete view. Right. And because the data isn't integrated and there's not standards and government is trying to push standards, but it's a longer road. So that is certainly part of the problem. And certainly we're not at the stage or I don't think we're at the stage where machines are making medical decisions for you. Right. I think humans are still in the loop. It's not completely automated, like, okay, this is the best medicine for you, period. Humans are still reviewing recommendations, which I think is actually still the way it should be for a while. Because like you said, these machines fill in the blanks. They'll just make guesses and you won't know what's going on right behind the scenes. And I think one of the things challenges as a society we are dealing with is we are losing or eroding the ability for us to think critically. Right. If we become so reliant on machines to do things for us, we stop thinking and we just overly trust output and soon people won't even know if it's right or wrong. Like, it'll just like, oh, okay, it said to do this or the answer is 22. I guess we'll go with 22. Like, what? How do you even know that sounds. That's right or not. So that's what I worry about. But healthcare, yeah, for sure is a lot of manual processes. I mean, anybody who's ever gone to the doctor and received some crazy explanation of benefits form, like six weeks after you go to the doctor, you're like, what is all this stuff? How does it mean? And so it's highly manual. People typing in information, people making mistakes, data missing, whatever. And so there's a lot of room for opportunity. I'm like, even at Bushill, California, we got a whole bunch of faxes every year. Like, what faxes? Like, yes, faxes are still around. And that's because there's no standards. So, Doc, doctors don't want to deal with 30 different insurance companies. They're like, just fax it. Like, fax it. Everyone's got a fax. Right. And so that's why it still lives on. And so there's a lot of friction in the system to be taken out for sure. Even before redesigning things completely. Yeah, I mean, who would have thought? You know, we're having this modern conversation about data in the AI era and you're bringing up how to accommodate fax machines. Yeah, I mean, it's nutty. And I think just to continue in healthcare, I think personalized medicine is not that far away, especially as we get an explosion in the number of sensors and inputs into the system. Like, just because your cholesterol is X, you should be on a statin and everybody who has cholesterol over this should be on this statin. Like, really, everybody's still different, you know, why should everyone take the same drug? Like, like there's so many different permutations of what's going on in your human body. And so I think AI is getting closer and closer to that concept of really understanding what's going on. Because as we talked about at the beginning of this call, more data, more inputs, more than, you know, humans are typically processing. Like even a human doctor is typically processing. Right. There's more data to be had if you can simply gather it and have access to it. Well, maybe, you know, you bring up personalization, we'll flip it a little bit maybe to your current opportunity with, you know, working for Taylor, a fashion startup. Right. Using AI styling customers better than they can themselves, which I could definitely use that assistance. So I'm going to have to check it out. But what excites you most about AI's potential in this consumer personalization side of it as well. Look, Taylor is a men's subscription rental company, right, where we try to understand the customers and use a combination of human stylus and algorithms to send you clothes that you are going to be excited by and help make you more confident so you can achieve your goals. And so it's a hard problem, for sure, from multiple perspectives. Just on the clothing itself. For example, anybody who's ever gone shopping knows a small is not a small is not a small. There's very few standards, right, in terms of what that is, and does it stretch or does it feel soft? And there's all kinds of metadata about every piece of clothing that you don't even realize you're processing as you're looking or wearing a piece of clothing or shopping for one. So that's a complex thing. Like, what is it about this shirt? I need to know, does it hang long, does it hang short, does it rod, broad shoulders, Is it this, is it that? And then there's a large volume of metadata about a human, okay, what's their size, their waist size, what are the chest size, height this, that, and long arms, short arms, whatever. And what is all of that data? And you're trying to match this data between a large volume of metadata about a person with a large volume of metadata about a piece of clothing and say, okay, does this one go for this person? And then you add onto that contextual information. Does the person need it for work? Do they need it for, like a date night? Do they need it for, like, what do they want this thing for? Right? Because things have different purposes. You have to take in, count the weather outside. Like, if it's hot, you're not going to want this. If it's cold, you know, whatever. And not only is all of this hard, then you got to do outfits. Does the shirt even go with the pants? Like, I don't know, you're not going to wear the Hawaiian shirt with the dress slacks necessarily. Right. And so how does this happen? So long way of saying it's a complicated problem, one that it's a little bit subjective. Okay, I agree with that. But at some point, it's not subjective anymore. Like, that looks bad and that looks reasonable. And some human stylists are good at it. They can style you, right? They'll look across the inventory of everything you could be wearing, understand you as a customer, understand that different clothes match you up. And what we're trying to do is almost capture the essence of that styling Capability and scale it. So this is for men who don't have the time or the interest or the skill necessarily to be able to style themselves. But can we get something that can create that for you and augment that for you? I mean, of course we could just hire a whole bunch of stylists as the customer base grows, but that's not the plan. The plan is to somehow scale the skill set of the good stylists to be able to apply it to a wider audience. Right. And so it's a hard problem and one that people can relate to and experience every day. And can we apply some intelligence to help assist in this problem area? So that's what we're working on. Well, I could tell you as a 6, 8 man that's frequently is looking at big and tall, even just the notion of, hey, can you help me compare all the different random sizes? Does it run long? This and that. Even that simple problem is super hard and I would love help with. So we're heading up on time here. I always do like to end these with a little bit of advice. So if you have some aspiring people that want to get into this space, right, the data space, the AI space, what's your advice to them? What's the best way to break in? What's the best prep they can do in order to be successful? Technology keeps evolving quickly as we're all experiencing, right? So it's not about, oh, go learn this one tool and you'll be good. I think my advice in general is to really ground yourself in side of it in terms of like, what problem are you trying to solve? Is this going to matter? In what way will it matter? The technology will follow to help you do that. And like I said, it evolves over time. But approaching things with a technology first lens is where I've seen most things fail. Like, oh, we'll just build this great database and everything will be solved. I'm like, no, it won't. Right? Like, what are you trying to solve for? And that means thinking about, this is the problem, this is the way I can solve it. To solve that, I need this data, right? And so working backwards, not working data forward, but working business backwards is what I found to be successful. So that's a vague answer to your question. I understand, but it's not like, oh, go take these three courses on Coursera and you'll be good. I'm like, yeah, maybe for the next week, but then it'll change again, right? And so I think really having that critical thinking skills around Thinking outcome oriented, that's where I think most people are falling down is they're not thinking outcome oriented or scale oriented. Like in this world especially, you can do most things once. You can do this analysis once, you can piece together an agent once to do this thing once. But will it actually scale? Have you thought about operating a scale? Have you thought about the change management required inside of an organization to adopt whatever it is that you've created? People also ignore the change management side of the world and say, wow, this is a better way to do it. I don't understand what's wrong with you. Like, what's wrong is people don't like change, all right? So you've got to embrace that, which is almost the softer side, I'll call it, of technology. People are just too enamored with only the hard side of technology. So that's my advice, actually. And for those, you know, subscribers that are listening, they'll catch on. Like, it's a trending answer in terms of a lot of successful experts here of, you know, I bubble it down to curiosity, persistence, right? Like, understand what you're actually trying to do and be persistent enough to push to make sure you're actually doing it. And yeah, embrace change because the one thing that will happen is inevitable is things are changing. Things are changing faster than ever, right? Like as fast as things are changing today, today is the slowest you will ever experience. It's only experiencing and accelerating. Awesome. Well, Zohair, thank you so much for taking the time. I think a lot of these insights, one I think are trending. So hopefully our listeners are starting to get this grasp of lots of cool things can happen, but there's some tactical advice on how to make sure you ensure some success as we go. So thank you for sharing and being on the show and appreciate it, listeners. Hopefully you enjoy it. Thank you for having me, Nick. Bringing data and AI to life is brought to you by Informatica. To find out more about Informatica and how our intelligent data management cloud can help you achieve better business outcomes, head to informatica.com stay tuned for more illuminating discussions until we meet next time. Keep harnessing the power of data and AI to bring transformative outcomes to your life and business. Make sure to click subscribe so you don't miss any future episodes and tell your friends about us too. On behalf of the team at Informatica, thank you so much for listening.