Learn more

Advice Architects Ep. 11 with Cara Dailey

You'll Shoot Your Eye Out: Why AI and Data Products Are Damn Hard

Day Wachell
August 14, 2023

In this episode of the Advice Architects podcast, we are joined by Cara Dailey, formerly Chief Data Officer (CDO) at LPL Financial and Bank of the West. Cara has extensive experience leading data-driven initiatives in the financial services industry and shares her insights about balancing risk mitigation, compliance, and privacy while maintaining your firm’s objectives for business growth and operational efficiency.

Listen to the full conversation between Cara Dailey and Responsive AI’s Day Wachell.

For more information about Cara Dailey, you can find her on LinkedIn.

For more information on Responsive AI’s solutions for advice providers, contact us.




Day: Welcome.This week's Advice Architects episode is titled You'll Shoot Your Eye OutWhy AI and Data Products Are Damn Hard. This is Day from Responsive and today we are speaking with Cara Dailey, a transformation driven C-level executive with more than 20 years of experience directing teams, overseeing development and launch of challenging and innovating data projects and programs for industry leading organizations across the US, including as CDO at LPL andBank of the West. Cara and I have had a number of awesome conversations about how data products are hard to do, so I'm looking forward to today's show. Cara, welcome to the show.

Cara: Thank you.Today, It's very exciting to be here.

Day: Awesome. At the start of the show, we do a little questionnaire to get to know you and your relationship to fintech and healthtech. So, what's your job title and what do you actually do?

Cara: Well, I have been chief data officer for a number of financial firms and chief data officer, head of data, head of data and analytics. It's really all the same, which is you were leading the charge on helping an organization transform their fundamental data infrastructure, how they use data to make decisions and really help them be more efficient and drive revenue opportunities with this really interesting, but hard to understand, asset called data.

Day: And how long are you doing that for?

Cara: Forever. No, I've been I've been really doing my data roles since the early, or really, mid to late 2000s. I really earned my data chops during the financial crisis. I was working at GE Capital at the time. And, you know, data really wasn't a thing. And all of a sudden overnight with Lehman and everything that was happening in the financial services industry, data became extremely important.So, I started in software, moved into financial services, and then overnight started to work on this thing called data management. And actually, at that time it was called business intelligence and it was reporting. But really, what it was there was all this data work that led to the final reporting. And I was really interested in the end-to-end process.

Day: When were you happiest in fintech or data?

Cara: I am happiest when I really see people using the data and being able to go home early at night. I know that sounds very Pollyanna, but I really enjoy the fact that someone will say, Oh my God, this dashboard took me two weeks, night and day to pull together and it's just a press of the button and I can now go and do more interesting things. And that makes me feel like I really helped an organization or, you know, the other times have been when we've discovered a prospect because we ran a model that no one else would have known otherwise, and we brought new revenue into the business. Those are really exciting moments when there's efficiency, when there's revenue, when there's a real tangible outcome. That was driven by myself, my team, because it's always a team activity.

Day: Yeah, those magic, those magic data moments. Yes.

Cara: Yes.

Day: Why do you care about wealth management technology?

Cara: Why do I care? Well, I do believe that everybody being armed with the right tools to control their own wealth is really important. Whether you're a college graduate, your mid-level career, your retirement age, you’re a wife that just lost a spouse, you’re an ex-wife that's going through a divorce, having these tools and capabilities that are almost self-serve makes it so much you're armed with the knowledge and the tools. That's what I truly feel about wealth. You know, I lost my father last year and I watched my mother struggle through understanding what her entire financial profile was. Well, lucky for us, we were able to dig through the data and get the right tools in front of her and she could understand it. And we had a fantastic financial adviser helping us.So that's why I'm passionate about this this area. I'm passionate that it's not just for the uber-wealthy, that it should be for everybody.

Day: It's people's lives. At the end of the day, it is where they're headed. Yeah, so as you know and are probably tired of by now, there's a lot of hype around AI. I don't think we've seen such a blow out of the Gartner curve ever. But before we get into generative AI and all the magic, I want to talk about what the foundation of AI, which is data. And for those of us who know how the sausage is made around data products that have actual users and use actual data, there's so much complexity to get to day one, and some of it doesn't even have to do with the data. So, to start out, let's what is a data product? What does that mean?

Cara: I've described this in so many different ways with so many different analogies. So,I will pick one. You know, I envision data products as a bounded set of data, whether it's customer data or product data, but it's something that can deliver a tangible outcome. So, if you're looking at customer data within your firm and you're building a data product around the customer profile, you can do all these really cool 360 analytics and behavioral modeling and all of that because you have all your customer data in one place. And the most important thing I think, I don't know, people would disagree around building a data product is honestly the people. It's the person that becomes that product owner that knows how to shape that product, bring in the right data, work with the right business teams that understand and how to define that data. That's really important. When building data products, it's like building products. It's just data is the foundation of it.

Day: So, let's talk about why that's really frigging hard.

Cara: It is.

Day: And, what happens kind of on day zero. So, I'm a data product owner and there's this great idea and everybody's saying we should, we should head down this path and do this because it's going to create a lot of value. What am I going to find on my adventure that that makes this a long and winding road?

Cara: Well, I think it always starts with the people because you could build the best data product in the world, but if no one understands it and how it was assembled, no one's going to trust it. So, I think as a data product owner, part of the first steps is really understanding the business problem you're aiming to solve.What's going to mean something at the end of the day, is it going to be I'm going to be able to change a client address in one place and a hit, you know, push of a button, and it will syndicate across 100 systems that will automates o many people's lives. And you need to understand who those people are. You need to understand sometimes people don't want it like as much as the data product owner thinks that automation and this efficiency might be the best thing. There are people out there that that is their role. And so it's about talking to them and understanding their current pain points and saying this is how we could do this better so you can go and do something else. So that's one part of it is one is understanding the business problem and then defining those outcomes. The second thing and this is where I think data products and this evil word called data governance come together.

Cara: And this is where you can really drive good governance. We'll call it enablement because, you know, there's an aversion to the word governance. If you can get this data product built with the right owners, clearly defined data, good quality and controlled data, then your life is easier. I know the analytics teams are going to want to go to go grab the data and put it into something and deliver something. It's not going to mean much if the data is wrong, and that's where the whole connection to AI is going to become very important is if you're building enterprise data products, you want governance at the core or enablement at the core. You want people to eat their vegetables and build the state of product with good, clean ingredients so that at the end of the day you could hook up I to it or generative AI to it and you could ask it questions and it could like closed loop a decision process, right? So that's, that's where I think those are the things that are important to a data product owner a day in the life of a product owner. Those are the things that I think are the most important when building data products.

Day: Now why are data products political and how can they be political?

Cara: I hope they're not. I try not to think they are. But what it does come down to sometimes is ownership. And ownership is a real emotional thing at times with people, right? It's like, no, that's my business. I own that. I get to say what that number is and means, and you're taking that control away and saying, we’re going to automate this whole thing. And, you know, they'll they could feel left out. Right. And that's when things become, I guess, political or ownership can become emotional. And so, it's really about at the outset of creating those data products, really identifying who are those key business owners, who are those key players that have a say in what that data is, where it comes from, what quality metrics ensure that it's right, they're going to know the best. If you rely, and I love the engineers, I really do. But if you rely on engineering to make those decisions, it might not be the right, the right call at the end of the day. So that's where I can see where it can get a little bit political. Now, the other thing is, if we're thinking about AI, we are really getting even a further step removed from that business owner. And now AI is coming in and making these types of decisions. So again, I think it's really important to start with the people, start with the key players, understand who owns these business processes and who can say definitively where to get the data, how is it defined? That's really important to assembling those data products.

Day: So, data can be political because it can live in silos. And those silos have owners who have attachment to the to the outcomes and the data itself. But they're also afraid of things. There's risk. Can you talk about some of the things around privacy and security that keep these things in silos?

Cara: Well, I think with the increasing of privacy regulation, what's deemed personal identifiable, that definition is is on a sliding scale, right? Depending on how you combine the information together. And given that these privacy regulations are newish, companies haven't really thought through their privacy strategies at times. And, you know, quarantining off PII, personal identifiable data, that doesn't even make sense to some to a consumer goods company because customer information is everything. It needs to be readily available to be making sales and marketing decisions or to a wealth company. We need to be able to have PIN numbers and Social Security numbers. But the reality is that information can be tied together and cause a breach at a firm and that that information can get exposed. So, while you're building your data product strategy and thinking about all this great growth and opportunity and efficiency, you need in parallel to be thinking about risk mitigation, compliance and privacy, everybody asks me, Well, are you more of an offensive or defensive CDO? And honestly, on both, I think both are equally important and that you should build a strategy that weaves them together and it can be done. It just takes a little bit of time to put in more proper data quality controls around that PII data.

Day: Okay. So, we've gotten our data, we've built our data product and now our product has users and they got to use it right. What can go wrong at this stage and how can we ensure successful usage of the data product?

Cara: This is probably for me the most fun part. So, I know you asked me that question early on. I really like helping people understand there's a better way, right? And when a data product comes on the scene, no one really changes their business process. They're like, well, I'm just going to do everything the same way, butI'll just double check it against the data product that you built and see if the numbers tie. I mean, how many finance teams have we worked with that go parallel run after parallel, run after parallel run. So, it's a lot about building trust. That's first and foremost. And the way you build trust is you let those key players, those key stakeholders, those champions, be part of the process. If you go and build this in a box away from them and then come back and say, ta dah, I did it. We have this great product. It automates everything.You're going to get a set of dashboards at the end of the month that's going to tell you, you know, how much incentive you're going to be getting as a salesperson.They are not going to trust you. They're going to say, well, this is you know,I, I don't know how you did this. I don't know where this data came from. How can I trust it? So, one is build trust, make them a part of the process. Second is really good change management processes around things like training and adoption. You know, I should you should build your data products to a point where it's easy to understand, easy to access. 

Cara: You're going to have varying levels of expertise using that data product. Maybe you have data science teams that just say, give me access to the SQL, I'm fine after that, or you're going to have business analysts that say, you know what,I want to put this in some sort of query analyzer and I want to be able to run my own ad hoc reporting. And then maybe you've got some higher level executives that says, I want this data on my PowerPoint and that's it. So, you have to you have to flex for each type of user groups. So, while you build trust, you build training and adoption capabilities that that help them better understand. The other thing that I would say to add to this is you enable tools. You create tools that help them understand the data from a data literacy standpoint. So a metadata catalog as an example, you know, there's a number of vendors out there that can offer you this catalog where you open the front door and it's almost like a grocery store, a well-organized grocery store you see on each of the aisles what's down that aisle so that as a data consumer, you can go in and pull all your ingredients together for whatever you want to make. That's how I use analogies a lot because it helps connect the audience to something they already get. And I think those are the types of tricks of the trade that that have helped me in my in my own career journey.

Day: So, we built a data product now and it took some doing. We had to have the right people. We had to deal with ownership and risk and users string all these disparate things together. Now we have generative AI, right? And life's going to be so easy. And just kidding. What from your perspective as a data executive, what is the sincere promise or at least the hope for a promise of these large language models and generative?

Cara: Ai Well,I'm sure I'm not alone. And you've played with ChatGPT, which is awesome and terrifying at the same time. The thing about ChatGPT is it gives you an answer.That's the thing. It gives you an answer right away. And it's eloquent. It can write me a story, it can write me a poem, it can write me a PowerPoint deck that I can put in front of my boss. It can write my son's term papers. So, it's going to be amazing for so many things. And I can see from a wealth management standpoint, you know, this is going to be a game changer in terms of financial advice, in terms of financial planning. So, you know, there's so many wonderful advantages to using this technology. I think where people are going to have to be super careful is around the foundational data that you hook up to it. That data's got to be clean because if you start to take out the human element of checking and you just go with generative AI and it's giving answers on a screen directly to your consumers, your customers, and you're not in the middle making sure that it's not saying something wild. That's the thing that makes me nervous. That's the whole you shoot your eye out because you know it's only going to be as good as the data you feed it. I can tell ChatGPT or whatever, some generative AI to write me a story about, you know, myself. And I don't know, you can ask it to write it a story that information you feed it is now init. It's now in the algorithm, it's in the database. And now it can recall that whether it's right or wrong, right. I think those are the things I really wantt o see, you know, what others think about in terms of, you know, the the regulation around AI. I think it will be very interesting to see what will happen there. Yeah, because I think there's risks to it. But I also think there's so much opportunity that you can't turn away from the innovation.

Day: Yeah, it's been incredible. You know, people have always had a healthy, in my mind, even though I'm an AI vendor, they've had a healthy skepticism towards AI and it seems like this thing is so compelling and hypnotic, people are willing to bring it into the castle walls very quickly. And it is interesting to see. I think it makes people feel very smart and very special, which is, in my mind, very dangerous phenomena for human beings. Um, so let's go ahead. Go ahead.

Cara: One thing I would love to add is when I first tried it out, it made me feel like my job's going to get easier now because I spend a lot of time working on PowerPoint decks and writing just the right message that's going to resonate around driving value with data and literally, like you just pull it up and you can input it all that information without any controls. So, I think that's the that's absolutely the shiny object syndrome that people are having is my. It's a very personal thing. My life is going to be easier.

Day: So putting on your CDO black hat someone is brought this into your business.Let's just say you have a wealth, a wealth business. You work for a wealth business. What questions are you going to ask your team about this project and what risks are you afraid of if you if you do decide to use this?

Cara: Question I would ask is what data are we going to feed it? Meaning like what data foundation is this going to sit on top of? We need to carefully catalog the information in that data set. We don't want to let this loose on your enterprise data warehouse because it's going to expose all your warts very, very quickly. Personally, from a design standpoint, and I don't know if this is right or wrong, some AI specialists would say I'm wrong, I'm sure is that you put it on a very well contained data set so that you can start to train that algorithm and look at the output and get comfortable with the answers it's delivering. Because again, you just unloaded onto the data lake, data swamp, that you've been working on to migrate to something else over the last two years. And you're going to have a problem. You're going to you're going to see all the data quality issues front and center. So, I think that's one of the things I would ask my team is, okay, what data set, let's contain it. Let's make sure that there's no PII, no privacy data that can get exposed inadvertently. And then let's also understand what kind of questions would we be asking this capability and really start to think about what other data sets do we need to augment that is safe to use and then try it? I mean, you're everyone would be crazy not to try this and really pilot it, build an MVP around it, demo it, get people excited about it. You don't you're not going to win at a company if you say, we're not using AI, we're just not ready. You have to be ready. It's more about, okay, let's just control the use cases and like build upon them. That's my perspective at least.

Day: Okay. So, we're getting we're getting near the end of the show, which is which is time for the deep question. Clearly, this is a transformative technology and it's going to change the nature of everyone's work, even data executives. So, when you're thinking about this new technology, how do you think it changes your role? And how do you think data executives should think about their role in the future?

Cara: This is a great question.

Day: You can take some time if you want to. If you want to kind of think about it, don't feel like you have to answer right away and riff. So feel free to take a moment to kind of compile. It's a deep question.

Cara: I think every CDO should look at AI as a strong business partner and that you want to leverage these capabilities to do more to be a force multiplier. AI is never going away. The Pandora's box is open. Whether it's for good or for evil, I think that it's for good. And make sure you understand what you're walking into, that you have some key principles around the use of AI and the types of things you want to feed it, the types of data you want to feed your AI capability. That data officer could be the champion for AI at any firm. And I think this is a real opportunity for those CDOs to get out of the plumbing and into the seat of business value and outcomes, because that's really what's going to make you stand apart from the other CDOs that may be struggling in the data management minefields that you're leaning into. It's a strategic partner for you and it's driving those value and those outcomes. I think that's how data officers need to look at it. And the last thing I'll say about that is on the CDO role is, you know, and it's a little controversial, but I always think that I'm just working myself out of a job. Eventually, one day, the data should be clean. It should be well organized. It needs to be accessed by everybody.You've got your data assets working. You now have AI that can do a lot of the human elements of it. So you can look at AI as that partner and that that helper to let you do things that are different to work on different things.

Day: So, it seems like you're saying that the role is changing from one focused on data governance and data management to insight and business value. So, lots to think about for us data heads. Thank you so much for coming on the show. It was totally awesome conversation. Really glad to have you.

Cara: Thank you.Day It's always a pleasure speaking with you.


Day Wachell

Day studied AI in the SymSys program at Stanford and Film at Columbia. Their innovative thinking drives the company's success by bridging technology and the arts, leading to a culture of creativity and out-of-the-box thinking.