Why AI Leaders Must Get Their Hands Dirty ft. Allison Sagraves

Speaker A: Hello, we are Bringing Data and AI to Life, a, uh, podcast by Informatica from Salesforce. I'm Amy Horowitz. I lead solutions sales here for North America. If you've ever been 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 that have rocked these arenas. We're here to bring clarity to the chaos myth, busting the confusing parts and providing insights and guidance for complex data problems by delivering trusted data for analytics and AI. Hello, everybody, and welcome to today's episode of Bringing Data and AI to Life. I am your host, Amy Horowitz. Today we're joined by Allison Sagraves, one of the most respected voices in enterprise data, AI and analytics strategy. Allison, I am so glad you are here. Why don't you give our listeners some background and tell us how you ended up here.

Speaker B: Sure. Well, I was one of the first chief data officers in, uh, industry. I was a chief data officer for Top 25 US Bank. And I held that job for five years, which I'm really proud of because I think what's the average tenure is, what, two years, three years? It keeps shrinking. So I did that role for five years and then I became an advisor and I do consulting in the AI and data space. And also I just came back from Carnegie Mellon where I am, um, a faculty member and coach in the Chief Data and AI Officer executive program, which is really awesome. We just graduated with more than 90 students across the globe, really, who come to Carnegie Mellon to really learn everything about AI and data management and how to get to that next step in your career. It's a great program.

Speaker A: That's so amazing. I will definitely hit you up on the end of this podcast, but I have a question about where people start, so we'll hit on that. But let's talk about trusted data and the future of AI. With your background, you've been in this space for a long time. You see all of these executives through Carnegie Mellon. You mentor, you find, folks, organizations are moving quickly to adopt AI. We see that, we hear about it, we talk about it, uh, water coolers all the time. But many organizations are realizing their data foundations are not ready. What does AI ready? And, um, I'm using air quotes. What does AI ready data actually look like in practice?

Speaker B: So I had a really interesting conversation with, actually one of my former students at Carnegie Mellon who is chief, uh, of engineering for company. We all know that's in the pizza business. And we were talking a couple of weeks ago, and he was talking about they're putting in their semantic layer. And he said, we're meeting with the CFO because we have to agree on, uh, what the term sales means. And he said, there's the marketing definition of sales, which is demand, for example. There's an operational definition of sales, which could be menu value or whatever, and then the financial definition, which might be recognized revenue. And I thought this was so interesting, and I can't wait to talk to him more about this, because it really hit me at that moment that people keep talking about data quality and how we have poor data quality, and that's just a kind of a meaningless phrase, because I think the problem that we really have is I think we've cleaned up a lot of data. And cleaning up data from a quality standpoint is probably. We're pretty good at that. There are a lot of tools that help us with that.

Speaker A: We.

Speaker B: Where I think the real issue is at this point is the meaning of data and agreeing on the meaning of data. And so I think because in order for him to get to the next step of his project, he was meeting with the cfo. To me, that was just a light bulb, um, moment where, like, that's what it really takes is to engage with the C suite, with executive leadership, to really agree on some of the core terms and foundations of what sort of the taxonomy. I'll get a little bit into taxonomy in my life. What the taxonomy of your data in your context, in your company or industry really means. It's not about, are we missing a digit? Are Social Security numbers wrong? People are fixing all of that kind of stuff. It gets down to real conversations about the meaning of data.

Speaker A: I could not agree with you more. So my background, by the way, I grew up in it. I'm in Austin, Texas. You know, I grew up at a system builder in Round Rock, and it was one of the biggest concerns back in the day. This was the early 2000s. So imagine with all the data that we've had to your point, you can't move forward today, fixing your data foundation if you don't really have people that are committed to that. I'll tell you a funny story, antidote to what you were just talking about. You know, I was at a customer and they could not agree on the definition of VIN number. So imagine in Japan versus North America versus South America, VIN number, similar situation. They went to the CFO and went to the marketing department, went to everybody. And I think you hit on something that A lot of our listeners are questioning how do we even do that? So thank you for bringing that up. I think it's such an important point. Let's move on to trust and human centered AI. This is another big thing that I hear about all the time. Again, you read the articles in the news, you hear customers and prospects talking and advisors like yourself through your advisory and teaching work, which is really important. You've seen how trust impacts AI adoption. You just talked about this a little bit ago. What in your opinion, Allison, separates organizations that are successfully built on trust from AI and those who struggle to gain internal adoption.

Speaker B: So I think that all data, data sets or main data terms truly have to have an owner who, who owns the definition and is accountable for that definition. In fact, yesterday we had our practicum where all of the students in the Carnegie Mellon program present a roadmap. And on one of the roadmaps they actually specified that all data domains were owned and defined and agreed upon. They said it in a really, I thought specific way that I haven't particularly seen on roadmaps and were agreed to by the C suite. So I'll tell a funny story about how this really hit home with me because I usually record things from my cabin in the woods where many, many people know of my antics as a beekeeper and a steward of a forest. So this year I got into foraging and this does all relate. So I get to that in a second. So I live in Buffalo, New York. So in April I go out into the woods. Nothing is green except there's this giant patch of wood, one green thing. And I got out my identification app and it said that it's ramps or wild leaks. And I'm like, wait, I think that's a food. And sure enough, it's like a food that like people have eaten for forever, but it's also a super like foodie food and trendy food. And so I thought I gotta make some stuff with ramps. And then my brother in law who has a restaurant in New Jersey, I'm like, would you like some ramps? So I shipped him ramps. I brought ramps to a restaurant in nearby. And I was like really into this whole idea of like ramps only grow. They're like whatever those mushrooms are in France. What are those called? The uh, truffles? Yes. So they always grow in very specific conditions. And I happen to have this huge sprawling area of ramps. So I got really interested in the idea of what foods grow naturally in a forest. And so I started and. Cause ramps are only around for a Couple of weeks. So I learned that there are things like nettles, wood nettles would grow naturally in my forest. So I started. I'm like, okay, that's the next thing that's coming up. But at this point, like, all kinds of plants are growing because now we're getting into May, and, like, the forest is turning green. And so I'm like, I gotta find nettles. So I'm wandering around the forest and honestly, everything kind of looks the same at a certain point. Like, is it a weed? Is it a nettle? Is it whatever? And so I was sure that I had come across nettles, which are very, very nutritious food, very high nutritional value, used in cooking, have been used for years and years. So I was super excited to get nettles. So I picked something that I had the app and it said, yep, that's nettles. I had the book and I'm like, yep, I think that's nettles. And so I started picking them and I'm like, you know, I'm gonna taste a nettle. I wonder what a nettle tastes like. And then I read there was like, one tiny attribute about the leaves of a nettle, that they are alternate leaves, not opposite leaves. I'm like, well, these leaves kind of look opposite. But I'd already taken a bite, and it turns out that I had picked white snakeroot, which is actually poisonous. And it's actually. Anyone could Google this. It's what killed Lincoln's mother. Anyway, so I took one bite. I looked up on Chatgpt, go to the hospital. I'm like, come on, it's one bite of a white snakeroot. So my point was that Linnaeus, back in the day, one of a few hundred years ago, started this whole taxonomy of plants so that we could identify the characteristics, the plant families and so forth across the world. Imagine doing this with, like, no real transportation or technology, but came up with an entire taxonomy, what is it? Binomial nomenclature. So that you could identify that a plant in one location was the same plant in another location. So I began to realize how important the specificity and taxonomy are to feeling trusted in classification, which is basically what data is. And I realized that if I want to be a forager and I want to provide food into the food supply in whatever way that is, maybe just local restaurants, I have to be able to certify and stand behind the fact that this is a nettle and it's good for you and it's not snakeroot. So it Sounds kind of ridiculous, but my first use case was super easy. It was ramps because that was the only thing that was growing in the forest. And, and I became a little overconfident. And then uh, when everything started growing, it's a lot harder to differentiate among plants. So I really feel that that is the same story with data. We can get a little overconfident. We can do that first pilot, it can be successful and then we start getting into the is this a nettle or is this a snake group? Pick your work example. That is a parallel and that's where having a taxonomy and doing that due diligence, really understanding the data as a practitioner and as an expert, subject matter expert is essential. And I think that we just overlook that a lot of things are more automated, but there still is a bit of a human touch that's required to ensure that you have a confidence behind your data. And anyway, that's my tale.

Speaker A: I love it. And actually, you know, it leads into the difficult role sometimes of a chief Data Officer. What you just described is what we see a lot of the time. You know, we have some of my friends that are CEOs are they have the budget or they don't have the budget, they don't have buy in from other C levels, they do have buy in. It's uh, a very difficult role. So let's talk about this. The evolving role of a CDO and chief data and AI officers. So you've helped so many people, you really helped the industry. You shape conversations around this, around modern data leadership through your work at Carnegie Mellon and across the industry. I, uh, want to hear your perspective on how this role of the data and AI leader changed over the last five years. You've seen it grow from starting to where we are today. What would you say has changed?

Speaker B: So certainly in financial services and in regulated industries, the initial mandate for chief data officers was uh, more on the defensive or regulatory side or kind of protection, compliance oriented work. And then after the data was established to be able to understand a customer, let's say for a 360 degree view for like let's say fraud, then people realized, well, we can use that for identifying their needs better. So we kind of went from defense to offense. And that was a shift in the chief data officer era. I think now we're kind of going back as we've introduced AI because it's a new set of opportunities on the offensive side. But I think that because the risk is so much greater, we're kind of back to that defensive piece again. So I feel like we're at ERA, let's say 2.0, 3.0, whichever ERA it would be. But I think there is a bit of a start again on the defensive side. So say for example, in financial services we used to say boring is the new sexy. So I think that's kind of true with AI right now where the use cases tend to be more fraud related, more stopping problems as opposed to on the offense. I think that's coming and maybe in certain organizations those use cases have been deployed. But I think once again we're approaching it from like uh, the defensive angle. And I think again we're still, I mean there's various levels of maturity, but most organizations are still kind of at that pilot phase. And I think part of it is because of what I talked about earlier is the meaning gap. There's not enough trust in meaning across major terms, across an entire company. And that hasn't been arbitrated to the point where people have confidence in their data, but I don't think that they know why. I think that it's characterized as a data quality problem when I think it's partly that, but it's also this data meaning problem. So I think that's kind of why we're a little bit stuck in pilot.

Speaker A: Pilot phase. Yeah. Well, let's talk about long term value strategy. How do you monetize what you're doing? This is the number one conversation I have on a daily basis. I have all these pilots going. I'm having a hard time, number one, justifying the use case. Number two, I can't really measure and if I can't measure, how can I justify long term value? So you talk about this a lot. You talk about tangible results from data. For leaders, this is a really important one. This is the money question for leaders trying to justify AI investments today. What metrics or business outcomes should they focus on beyond just efficiency? Do you have anything that you would recommend?

Speaker B: Well, I'd say I would look at the industry that you're in. And so in financial services, like I said, I think the initial use cases are really focused more on hey, how can we stop fraud? How can we stop money from going out the door? I think there are particular use cases that are contained in a business unit. I think ultimately, and I think it is going to be hard to justify AI for like a period of time. I really feel we're in kind of an organizing period which is probably going to be a couple of years. I mean, on the one hand you hear people Say like, well, if you don't have your act together now, you're out of luck. And uh, in three years the whole world will be different and I think it probably will be. But my experience is that to actually affect, to effect material change in a company, you really have to completely rethink your business model and workflows. And that is not something you can do in 6 months, 12 months, 18 months, even 24 months. So I think we're going to continue to see, this is maybe an unpopular opinion. I think we're going to see in mainstream companies, maybe not in the top top companies, we're still going to see less at the enterprise level and more solution based, kind of division based value. I mean ultimately where I think we're going to see value, I think uh, the value is going to be societally transformational. It will be, we discovered diseases earlier, we prevent diseases, we drive more safely, we are able to extend credit to more people. I mean the potential is transformational. I think trying to achieve these transformational results In a uh, 12 month roadmap is not realistic right now.

Speaker A: This is a great question. So if you had 30 seconds with the CEO that said, hey Alison, AI can do all this alone, it's going to solve all my problems, what would you tell them about the role of data, trust and context? What does that play in long term success?

Speaker B: I would just really encourage them to get their hands dirty and have a real working session with their engineers and subject matter experts in a given domain and to find out what is their equivalent of what are my pizza sales? I would ask a question that fundamental. So whatever industry you're in, whatever business context you're in, ask a question that you think is so basic and then sit with your engineers, sit with your business people, sit with people from different departments to say, well, you know what sales can mean these three things or whatever can mean these three things and appreciate for yourself why there is more complexity to this and learn how to build a basic agent, learn what can go wrong. I think you really have to be hands on in order to be confident enough to be a leader, to be able to ask the right questions of your team. I think that if you don't kind of practice with AI yourself and incorporate it into your daily life, I think there's a confidence gap that can cause people to not feel. You have to be able to apply your own intelligence and ask questions with confidence in your own business, with confidence and humility, because nobody knows everything. And I think that this is an industry where people are shamed where if they don't know something and they feel stupid for asking a question. So I think you have to create an environment where you can be stupid and learn from the ground up to be able to start to be confident, to ask real serious questions. So I think we're kind of in this spiral where people are not free enough to admit what they don't know, which we all don't know, all kinds of things because the environment changes every day. And to have that kind of permission to learn and be hands on and then you start to get a more realistic view of what is possible in your organization with your technology and with your partners.

Speaker A: So to summarize, the mind shift change that you feel executives today need to make is it's okay to ask questions, it's okay get your hands dirty, let's get into it, let's be vulnerable, let's ask questions, let's learn. Would you say that's the biggest mind shift you think needs to happen?

Speaker B: Yeah, I think you have to do the equivalent of put your boots on, walk out into the muddy forest, pick a, uh, green, taste it, risk a, uh, little poison so that you can be confident when you've actually picked a nettle instead of, uh, snakeroot. It's that tangible. It's that hands on. And that doesn't mean that you will be the one that will always be doing that. But you have to appreciate what it takes in order to be realistic about what is required to execute something well and responsibly.

Speaker A: I have one question that's not on our sheet. I feel that I hear on the news daily that companies are laying a bunch of people off.

Speaker B: Off.

Speaker A: We had one in Austin that laid off 60,000 people or something crazy like that. Do you feel that's appropriate? I think you hit on this. I just want to hit on it again. You said not jumping straight in with both feet. You need to explore and figure out how do you tell a CEO that AI is not taking everyone's job?

Speaker B: I honestly think that I, uh, have a contrarian take on this, so I'm almost afraid to say it. I think that most companies have a lot of people that they never really knew what a lot of these people were doing. Most companies are run by 20 to 30% of the people. That's just the reality. But also nobody can figure out who those 20 to 30% are. So I think that that's not new. That's been around for my entire life. So I think that they can say they're laying people off because of AI. But honestly I think there's like an accumulated efficiency gain that has lagged the technological and data revolution over the past 20 years that for my entire professional career we would automate something and then people would say, well it's fingers and toes. It's fingers and toes. And I'm not like advocating, well fingers and toes should add up to people. I'm not like saying that this is a good thing, but it never did. And so much automation has been introduced into the world in the past 20 to 30 years and yet we did have productivity gains like as a country. So I think that honestly some of this is just catching up to previous automation that wasn't really accounted for. So I see it as a lagging indicator instead of a leading indicator. I really do worry about what is going to happen to jobs. I don't think we are all there yet. I think that like I said, what you're seeing now is just kind of a retroactive catch up. But I really do worry about the impact on jobs. I think nobody really knows. I mean but it you one sees oneself when you go to do something like suddenly I'm able to answer medical questions that I might have either not ever gotten to ask my doctor for or if I did, I now I can answer it myself. In many cases I'm able to do basic legal work that I would have had a lawyer do. So you kind of see it every day and it's naive to think that it's not going to have an effect on jobs. I think these are massive societal questions we're going to have to grapple with.

Speaker A: Thank you for that. Okay, now here's the fun part. I'm going to do five fire questions that are going to be quick, just right off the top of your head. No prep necessary. Here we go. Are you ready?

Speaker B: Yes.

Speaker A: Okay. What's one AI buzzword you wish would just disappear?

Speaker B: Human in the loop.

Speaker A: What is the most underrated skill in modern data leadership?

Speaker B: Arbitration.

Speaker A: Which industry, according to you, is handling AI adoption the best right now?

Speaker B: Pharma.

Speaker A: Yes. What's one data mistake organizations continue to repeat Oversimplifying. In one word, what should AI feel like?

Speaker B: An adventure.

Speaker A: I love it. Alison, thank you so much. This was amazing. I really appreciate you being here. Thank you again.

Speaker B: It's been great. Thank you so much for the opportunity.

Speaker A: Of course. So thank you to all of our listeners out there. Please remember to subscribe like find us wherever you get all of your podcasts. We are on every platform and we look forward to our next adventure with you. Thank you everyone. Have a great day. 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 here at Informatica, thank you for listening.