Today, on the Data Masters podcast, we delve into the complex world of artificial intelligence (AI) and its profound impact on the enterprise. As business leaders grapple with the implications of AI, they often find themselves facing a myriad of challenges. Executives, managers, and data experts come together, recognizing AI's importance, yet often feel perplexed about how to align it with their business strategy effectively.
In this thought-provoking episode, we speak with Elena Alikhachkina, a data and technology pioneer who has been nominated as a Global Data Power Woman for four consecutive years. She dispels the myths surrounding AI technology and explores a crucial question that should guide every organization's AI journey: "Why?" According to Elena, it's not enough to simply implement AI solutions. The real power lies in connecting those solutions with a strategic vision to drive tangible and meaningful outcomes.
If you've ever wondered why AI initiatives often fall short of expectations and want to unlock the secrets to effective AI integration, this episode is a must-listen. Join us as we explore the transformative approaches your business needs to leverage the true power of artificial intelligence.
Tune in now and gain insights that could reshape your AI strategy.
Intro/Outro - 00:00:02:
Data Masters is the go-to place for data enthusiasts. We speak with data leaders from around the world about data, analytics, and the emerging technologies and techniques data-savvy organizations are tapping into to gain a competitive advantage. Our experts also share their opinions and perspectives about the hyped and overhyped industry trends we may all be geeking out over. Join the Data Masters podcast with your host, Anthony Deighton, Data Products General Manager at Tamr.
Anthony Deighton - 00:00:39:
Welcome back to another exciting episode of Data Masters. Today, I'm thrilled to introduce Elena Alikhachkina, a visionary leader and trailblazer in the data analytics industry. Elena stands at the forefront of digital innovation with over 25 years of executive experience and data analytics at some of the most prestigious companies, including Roche, Danone, Johnson & Johnson, Nestle, and MABA. Her remarkable journey through the ranks of Fortune 500 companies and her substantial influence as a global thought leader have garnered her many accolades, including being named among the global data power women for four consecutive years. Let's extend a warm and enthusiastic welcome to Elena. We're excited. Thrilled to have you joining us today on Data Masters.
Elena Alikhachkina - 00:01:31:
Thank you so much, Anthony. I'm so glad to be here.
Anthony Deighton - 00:01:35:
I thought maybe we could start in a funny way with a question that I've often ended with because I think it can kick us off in a conversation about AI. I've often asked guests on Data Masters their view on Gen AI, specifically whether Gen AI is overhyped or underhyped. Sometimes we have this conversation and somebody says, Gen AI is as big and important as the internet. It's as big as the internet. And other people say it's just a stochastic parrot. Doesn't mean or do anything of value or interesting. Where do you stand on this debate?
Elena Alikhachkina - 00:02:19:
Holy... The biggest changes were made by the new input. This is when the internet came off really big. So that led to my office like 2006, 2007. Forms came in place. For warmth. So then we have a digital transformation, how it builds. So there are a lot of transformations. Excuse me. That's a long, long time ago. It's actually been going on for who came from. That's yours. You must say it. No. The iPhone is actually much bigger than the situation.
Anthony Deighton - 00:03:08:
Elena Alikhachkina - 00:03:09:
Out. Will last for a few days.
Anthony Deighton - 00:03:17:
Elena Alikhachkina - 00:03:18:
Changing the person's dog, right? But we all use social. We all have smartphones. We're all on the channel. We cannot believe what's out there. So I do see that the AI revolution is really some sort of singular, because all of us as consumers. Well, meeting. And we use a fabulous set of things. So they use Google Maps. Simple as this, right? So we use at home. Is it anymore? One of the most popular tools for salespeople Yeah. It's a long story, so I'm certain. Checking in with you. So when this is, I'm crossing this line. Same as we have been going on the ball. You know? You're right. Social events. I was looking for it most of the time. Hopefully John. And Yeah. Thank you, bye. Social media. That's so, so good. There's a big difference. Actually, me. To get it. Yeah. Okay. And the day it did not come to me, it's a big revolution. So it was really quite painful. And I think we are in the same wave. So don't think about it as hype, because it's kind of going to. So, consumers have a following, and we're interested in the following.
Anthony Deighton - 00:05:07:
Right. And so I think we understand the AI phenomenon has a personal touch. And I think your point is well taken, because if I think about even just members of my family, extended family, how many of them have ChatGPT accounts or ChatGPT-pro accounts? And also, if you simply look at the number of users of OpenAI's technology, I think it may be the fastest growing new rollout of technology ever. Yet, the implication for this in the enterprise, I think, is causing people lots of challenges. You've talked a little bit about this idea that executives, managers, data people will get together inside a company and brainstorm, okay, AI is happening. You know it's important. What do we do for our Business Strategy for AI? And they have many ideas, and yet very little of them actually turn into production deployments. You have a point of view about why that is. Why is it that we're not seeing companies, organizations able to harness the power of AI effectively? So maybe that'd be helpful to people if they understood that.
Elena Alikhachkina - 00:06:20:
Yes, you actually said this. It's exactly the question why. Because when I see this happening, Discussions, relations. People by nature, solution. So people come to the table to talk about solutions. And the solution discussion takes most of the solution discussion is focusing on certain applications. Discussion: how bad is our data? Discussing, for example, the need to hire a volunteer, you're jealous of my age. So, for me it's a solution. So how would you recommend 20% Basically, a little bit more imagination for your sessions. And stayed in time. And discussing. Now, what they would say, and it's surprising, Thank you. It's kind of funny. Thank you. So I'll just make it like that. Thank God. Development When you're in a public organization, let's say a sales organization. They are much smarter. So when you take them at the end of the session. Okay. So I think this session was somewhat successful. Mm. And the question at the end Shade on. So we are ready to go. So as you know, Right. But by nature, we all jump into solutions. Let's fix the data. Let's do something because everybody do it. So start from what?
Anthony Deighton - 00:08:13:
Start from wax, which is great advice. I would add something to that that I think is AI specific. So starting with why, connecting the work you do around AI to business strategy makes a ton of sense. Other piece of it that I think people forget is scale. And I think the disruptive piece of AI is it allows you to break a constraint on a business, which is a scale constraint. You mentioned this just as an example, you might have a sales organization and the challenge is getting more salespeople and more effective salespeople. So why might we need more sales capacity and we need more effective salespeople. And then the disruptive thought is, well, what if you could add an infinite number of salespeople? Certainly a thousand times more salespeople because you can use AI to break this scale constraint in this example, your sales organization. So I think one piece to add to your idea is starting with why, but then also removing the traditional constraints. Oh, we can't hire salespeople. We can't do that. There are only so many copywriters in the world. There's only so many whatever. I'm like. Break some of these logical scale constraints.
Elena Alikhachkina - 00:09:31:
So I'm guessing you really care about the balance of the scalability. So there are two almost aspects of scalability. The first aspect is the balance of scalability, and the second one is the sharing of the balance of scalability. All the things that say Oh, my God. May he, yes. It's gonna be in the video. So the pen is not going to be created if not when you build another one. So it comes into, have imagination, with this salesperson, home marketing person, how huge is the number? So what can an individual in the future check? And we really try to And then you can see what's going to what adoption means, right? Because as an example, scalability could mean not just a technical side, could be the skill So it can be a scale. On the scale of the data scale, cross-organization. Or for example, adoption also could mean a lot of stuff, including personal being honest. Many people don't have to.
Anthony Deighton - 00:11:10:
And another thing you've written about and talked about as a constraint on the system is the quality of data. And we know when it comes to AI and even Gen AI, finding quality training data, quality input data is a really important part of making these models. The risk associated with any AI project is around questions around bias or questions around hallucination or overfit models. And these are all really symptoms of poor quality training data, poor quality input data. And the traditional wisdom, the conventional wisdom around data quality is that you have to clean up your sources, go through and fix everything. But you have a contrarian view on this. And given the importance of the quality of data in the context of these AI strategies, what do you suggest to organizations on quality?
Elena Alikhachkina - 00:12:08:
So I've been getting suggestions so many times, let's just spend another two million to create the source. And you can do it every single year, spend two million to create the source. So, It means so much. Yes. This will mean everything. All right. You can imagine the tennis courses, we're going to sell some of channels So if you're gonna start the source,
Anthony Deighton - 00:12:44:
A road to nowhere.
Elena Alikhachkina - 00:12:45:
Exactly. And as soon as you clean, you're going to need it, I did. You definitely need to kind of, you know, rethink about What does it mean? And now, and it's something like, who has the theories, who has the ability to connect the data with the specific data. So definitely the beta quality here. Could be done. We do it. By using AI. So I think this is where AI can No. Yeah. Also, I think the data community side is no chicken side. So also more like a self-learned type of system, right? So it's not necessarily that you are now creating the rules, which I think 95% of them I think there is only like 5% of companies who are thinking about that. They are being biased. No. And of course, we're going to make them religious. So if you are sources by sources, and get advice. The question is, can we use more synthetic So again, the eye can come. And it could be done much, much easier. So as a data professional, I'm really excited how much day to day management. So. Because then they complete. My goodness.
Anthony Deighton - 00:14:40:
Yeah. And in particular, I hope that the data management is much easier. It's more inclusive, more people are participating in it, and we're not relying so heavily on high-skilled data engineers to do the work, but we can rely on the system, the AI, to do much of the work. For us, there's another subtle point you make I want to draw out. The idea that you could spend $2 million a year cleaning up sources, yes, it's the cost $2 million, but it's also the fact that you're essentially never done. It's $2 million a year forever. And this brings up this idea of a more agile approach to the data quality problem. You've been very much on the cutting edge of this concept of data products, an agile approach to managing data, thinking about treating them like a product and using traditional product management strategy. So maybe you talked about AI, it's very important. You talked about the importance of the quality of the data. Maybe we can talk a little bit about how data products fit into that. Maybe we could start with something really simple, like what's your definition? And are the data products, does everyone seem to have their own definition?
Elena Alikhachkina - 00:16:04:
Information. Like five years back sort of that kind of for scientists in this way. You want to add something? See? Yeah. So, when we implement a finish, Awesome. So for me, the main component is the question of Now this is something he showed me on the hall. Forms. So this could be a writing to the lab, even Syrian-backed communications. Me, the hour holder. And we've Sorry. Yes. So it's leaning consumers to the capital position. I also started kind of thinking because right now We have no movements. You in a Yeah. So we have technology transformation and I'm going to take that. So for me, it's more of a message. So we take this cell exposure. Bye. Vacation. So this application is now developed in a data product And the data layer. Is now developed into application. So then the concept of IT-related transformation And this is how I see it. And I get it from me, the kids are gonna change, not the kids. And visits. Is gonna make additional sales. He's a lot of kids. Are different levels of consumers who are right. You send the data.
Anthony Deighton - 00:18:25:
I want to highlight something there because I think that's really important. This idea that the data product strategy starts from these consumption endpoints and these business problem solutions as opposed to starting with sources and the conceit of trying to manage every source. Old enough to remember in the late 80s, but certainly in the 90s, this idea that the solution to all of our problems is that we're going to get all of the data to one data warehouse and that that was going to solve the problem. Or even the conceit that operationally that we were going to solve the problem by getting all of the data into SAP or to some Oracle or whatever the operational application of your choice was. And your point, which I think is really important, is we're never going to solve the problem that way.
Elena Alikhachkina - 00:19:17:
Anthony Deighton - 00:19:18:
And we therefore must start the conversation through the lens of how the data is being consumed and used. Is that a fair way of saying it?
Elena Alikhachkina - 00:19:26:
Exactly, actually, I can give you a point that some cases which Let me see your glasses. Before this was done by multiple salespeople. Right. So then, This is the problem. I now need to come to the customer and show the full portfolio. So then we go from this incident, That's why by the end. Yes, delivery of portfolio of pets. And Getting sales, getting customers, getting customers. And some other cases we want to be more profitable, right? So this is what we're going through. So I'm the person who is actually maybe in finance team who is really working on a, So how I can work across all products and all directions. In the holistic way. Smiling and talking. To really Exactly, going from the decision I want to make, or action I want to make, going back to the solution and going back.
Anthony Deighton - 00:20:40:
Yeah, and the common theme you hear there is when you start with that business problem, The constraints to solving the problem is the data is stuck inside of it. If I could, in your example, as a sales rep, understand my product portfolio across the different product lines, Now I can deliver the cross-sell and up-sell. Or if I can deliver profitability because I can review costs across different business units, then so good. Now shift a little bit from a user perspective because the other half of the data product strategy really centers around, I sure really like the way you framed it. You have these analytical use cases and these operational use cases. Typically, these organizations are at odds or at minimum, denied and neglect. They don't know each other. But you view data products as a mechanism of cutting our cost flows. Or is that a fair way to go?
Elena Alikhachkina - 00:21:38:
Absolutely. I actually got criticize. And I can't criticize, but some people just say, no. And how many publications, right? But I see actually that there's because the eye is application and is available So which means that we're gonna see the convergence technical obligation and that they thought God bless you.
Anthony Deighton - 00:22:42:
Yeah, so it's almost like this common language between these analytical use cases or analytical processes and these operational processes. And at some level, maybe that's the definition of a data product, which is this idea of cutting across. I think maybe this distinction between analytical and operational is something the industry has created because it serves its purpose as opposed to something that users care about if they go to solve problems.
Elena Alikhachkina - 00:23:11:
Yes, absolutely, definitely. And I always just say, People don't step into the data product management rules. I think this will be a great opportunity for business. There's a technique called skill. They are in the roots. I actually wrote a lot of quotes myself recently. So, right. So you can do a lot of stuff, right? But what cannot be replaced extremely replaces this. Right? And for example, if you uh... Market Salesforce. You're manufactured in Boston, right? Considering thinking about your career in data management could be a good opportunity. Because this is where your skill of knowing the business with some at the stage of the data management applications which you personally want to use, Good theory, okay?
Anthony Deighton - 00:24:14:
So let's pull on this for a second. So a data product manager, maybe step back. The people who are listening share how you define a data product manager. And then maybe also a little bit about, because I know you've stood up teams that have data product managers and are thinking about a data product strategy. Maybe share for those people who are listening, who are interested in building a data product strategy inside their organization, maybe some tips and tricks or ideas for how to enable, empower, train these data product managers. So first, what is a data product? So someone's got to write a job description. What do you write in that thing? And number two is like, okay, now you've hired one. How do you make them successful?
Elena Alikhachkina - 00:24:58:
Because it was the first time I thanked you. Make the crowd-sounding session which has been a wonderful technology team. We started thinking about that Product Mindset And Robert Meyer said for applications, let's say I'm a senior in application, Thank you very much. So we spent quite a lot of time with my team, that's why. So if I want to take a project manager, I'm not watching this. I have a product which I develop, I grow, I measure results, I know the cost of this product, I know who is buying it, and I'm delaying my customer. So these are really generic type of descriptions, right? But basically, I need to have data product manager Product Mindset, which again, product and let's take any product in a store. The same mindset. I don't have a product, I'm sitting, right? So the second is to understand the life cycle of this product. And data products have really define the cycles, or you need to define them. And the sort of measure is, value it creates. So I need to And I think the cost is actually, Quite a follow-up tricky question, because the majority of current data management organizations, they are treated as a cost center. But not the video. I hate this. This is where I was the product manager, and told myself like, what is my ROI? And you prioritize the teacher unit, the development of the student. Which is gonna get more money. Or we're going to get more engagement with our consumers, right? So it's a really different mentality. I'm not just creating the dashboard because I love doing it, but is the salesperson going to pay me for the dashboard? I think this is a great question. And there are no We defined, I would say, roles to take product management. I did a lot of search and assessment recently. You're saying that, right? People are, okay, for example, master management sometimes, or there's product management, right? So sometimes architecture, for example, either all of us.
Anthony Deighton - 00:27:52:
Which is a very different mindset because I think most data people approach the task as a technical task. It's about writing Python code, or it's about scaling a data lake, or it's about adding governance rules or something, and not really thinking about, to use your words, the consumer. And I love this idea of thinking about the users of your data product as paying for it. Now, at a practical level, in many organizations, they don't pay for it. Like they just work together. But is there a way that you've seen organizations make that more front and center, make it feel like you're paying for it, or how do you get people to think about it when they're not actually paying?
Elena Alikhachkina - 00:28:37:
So, I think, yes. The response. So we, And this is also the arms that shift. And there's a small female here. So. You know, choices. Way to go, Nellie. So here I definitely think my idea from before is, so when we define a product, we guess the value, we guess the benefits, my technical decisions are going to be making, making points about it. And then you can count the virus to create those cells. So then I need to account for the cost of this product, which is giving me the opportunity to create . Isn't it cost savings? So I saw some cases I found as people in India And get close to this because of the data but do not account for the cost of the data product, right? And people being free. And do pay for the target, which is interest in it. When you check on that young person. How you can put investments into technology and data. I see that there are less conflicts And I want to say that watching this. Yeah, it's the least... Solutions are changing souls. They can hold their sight. I mean, of course it was. Yeah. And that they both look like a little... So building this business for the solutions was not an easy challenge. Because we can see the organization's business. All the stars.
Anthony Deighton - 00:30:56:
Yeah, so it's much more Operation EX.
Elena Alikhachkina - 00:30:59:
It's more Operation EX.
Anthony Deighton - 00:31:00:
So let's cast our eye to the future, but not too far to the future, for next year. So we're on the cusp of the end of the year. We're heading into 2024. Given your outlook and what you see, I'm curious if you could share some trends that you see impacting us in 2024. What changes next year as it relates to data and analytics and anything that you care to comment on?
Elena Alikhachkina - 00:31:30:
Which will continue. One is that companies are changed. We have to make sure. So if you go to the data and So this I think people have to resolve. And how you can engage your customer and consumer in a different way, which is honestly. So I didn't know what to do. So you really need to. Actually, quickly can show results. And secure the next column. So consumer play, engagement, or customer play is probably much longer. But definitely the nice thing Amen. You will end up And this is where they will make it. He was dead. Yes. And Hmm. But at the same time, Because I do believe when such big challenges happen, Hm? Yes, you can see the link. And some are trained in the office. Form. YouTube and a native local solution.
Anthony Deighton - 00:33:50:
So just to summarize, Wardmew is the big trend for 2024 is that the data projects specifically, but probably more generally, need to be focused on delivering bottom line profitable results over simply thinking about how to create big infrastructure investments with uncertain payouts going into the future.
Elena Alikhachkina - 00:34:13:
Yes. So, the formalization items look like formalization and non-formalization, because it sounds like patient. You need to know. Because you basically tend to do the cross applications and focus on the
Anthony Deighton - 00:34:39:
So to say it a different way, if you can look across data silos, you're more likely to find solutions which deliver business value that's profitable because these are untapped.
Elena Alikhachkina - 00:34:50:
This is unpacking and surprisingly, people think it's such a difficult exercise, it's so expensive, and that we need to spend so much time on infrastructure. I had to get involved in construction. Then Brian really defined me as a donor. So let's say you're a financial officer at Bell New York. I want to improve to operate by ground points and then pull back and really see how this could be done. There are so many solutions which are post-adoption, which post-adoption work out substance abuse people.
Anthony Deighton - 00:35:23:
Good. Well, Elena, thank you so much for joining us on Data Masters. A fascinating conversation, starting with AI and Gen AI. So I think you have anchored yourself squarely in the camp of it's a big deal to how we think about the value of high quality data, and specifically data that cuts across silos as a way to really take advantage of that, but to using data products as a strategy for achieving that outcome. And the big trend for 2024 is making data projects profit, which is, I think, almost certainly right. So that's great. Thanks for joining us.
Elena Alikhachkina - 00:36:02:
Thank you so much, I'm super excited. So let's roll our sleeves, we're going to do it more.
Data Masters is brought to you by Tamr, the leader in data products. Visit tamr.com to learn how Tamr helps data teams quickly improve the quality and accuracy of their customer and company data. Be sure to click subscribe so you don't miss any future episodes. On behalf of the team here at Tamr, thanks for listening.