The following is a rough transcript which has not been revised by High Signal or the guest. Please check with us before using any quotations from this transcript. Thank you.
===

cara: [00:00:00] We're not gonna have BI in the future. It's all gonna be ai. We're gonna be talking to our data. We're not gonna be looking at dashboards anymore. When I use chat GPTI don't ask it to serve me up a dashboard. I'm asking it questions through prompts. 

hugo: That was Cara Daily predicting a future where we talk to our data rather than staring at dashboards.

Cara is the VP and head of data strategy at Early Warning, the parent company of Zelle. Cara's career is a masterclass in high stakes data leadership from the early days of online advertising at DoubleClick to building the data strategy for Nike and holding chief data officer roles at Bank of the West, Silicon Valley Bank and t Rowe Price.

Now Cara's navigating what she calls her product era, shaping the data strategy for early warnings, decision intelligence business, where she leverages rich financial data and data science to drive fraud monitoring and modeling. In this episode, Cara shares her progress over perfection approach to governance, why she's abandoning [00:01:00] monolithic platforms in favor of incremental data products and HER 80 20 rule for balancing operational rigor with innovation.

We also discuss why loving data isn't enough, you actually have to take care of it, and why AI is finally shining a spotlight on the often neglected fundamentals of data stewardship. If you enjoy these conversations, please leave us a review. Give us five stars, subscribe to the newsletter and share it with your friends.

Links are in the show notes. I'm Hugo Bound Anderson, and welcome to High Signal. Let's now check in with Duncan Gilchrist from Delphia before we jump into the interview. Hey Hugo, how are you? Hey Duncan. So before we jump into the conversation with Cara, I'd love for you to tell us a bit about what you're up to at Delphia and why we make high signal 

duncan: at Delphia.

We're building AI agents for data science through the nature of our work. We speak with the very best in the field. With the podcast we're sharing that high signal, so we covered a lot of ground with Cara. I was wondering, Duncan, if you would let us know what resonated with you the most. You know, most of us have heard some version of the claim you heard from Cara in the [00:02:00] intro.

Dashboards are dead. The future is just talking to data and sure. I feel that. And here working on Delphia. I love it. But here's the uncomfortable part. In most companies, the point isn't just getting an answer. It's being able to defend it to your team, to your boss, to your customer. And that couldn't be more true in financial services where you have a whole new set of stakeholders, regulators, auditors, risk teams, and that's why today's episode is so interesting.

At Early Warning, the company behind Zelle, Kara sits right at the intersection of ambition and that constraint. This is such an insightful discussion on what good really means and how data leaders can build systems that move fast without making things up. Let's get into it. Hey there, Cara, and 

hugo: welcome to the show.

cara: Hi, Hugo. It's great to be here. Thank you for having me. 

hugo: Such a pleasure and I'm so excited to chat about everything you're up to now at Early Warning, all your work across major financial institutions and all the stuff you've been doing [00:03:00] over the years in in governance, which as we've discussed before, is.

So important and fundamental, yet rarely, clearly not talked about e. E enough, I think. So maybe to start off, you've held senior data leadership roles across major financial in institutions, and I'm wondering what drew you into data as a career and what's kept you in in this space? 

cara: Well, I think I really like solving problems and there's nothing like.

Large organizations, data problems, they are the most complex problems. And so the role of the data leader really started in the business intelligence reporting space, and that's where I was, I was in, but in like pre-financial crisis timeframe, oh seven and oh six, I was in that reporting space. And when the financial crisis hit.

And I was at GE and we really needed to understand our data at a deeper level. I started to understand that it [00:04:00] wasn't so much about the tables and the rows and the columns, but it was also how it was organized, how it was stored. And that became very interesting to me that there, there was, that there was real leadership to, to be had around that.

So I really enjoyed solving big problems. I was always a great communicator with business and technology teams. I always walked the line in the middle and that's what really drew me to this whole data space. And lucky for me, I, or lucky for all of us, honestly, in this space, it really started to boom. I don't think any of us BI reporting folks back in oh six and oh seven thought that our roles would become.

These senior executive C-suite roles, head of data and AI that they are today. So I think the story arc for the role of the data leader has been really interesting and I was so thrilled to be a part of it. 

hugo: Totally. And you've [00:05:00] worked with and at so many interesting places since then. So maybe you could just give us a brief history of what you've been up to.

cara: Sure. So I started my career really in software and not even in the data space. I started it in online advertising. So I worked for Double Click and a couple of websites way back in the day. 

hugo: I do feel like that's like. OG analytics and data science in? In some ways, 

cara: yes, because we had very interesting ways in which we were targeting ads and it was all data related and it was good.

Right now you've got Instagram and algorithms and like we were way before our time, and so I've always been very drawn to that data-centric type of work. But then I pivoted away from the online advertising and I moved it to a role at Hyperion and I was. This is, I was much younger and I was in software and I really liked software development and I liked engineering and, and, but I was always on that frontline [00:06:00] customer role, so I was always customer facing.

I was technical support, I was quality engineering, but I was always engaging with the customer. So customer, the customer perspective was always top of mind for me in my earlier data roles. And then as I moved away from software into financial services, I started to take on roles that still had an internal customer facing element.

So I was in IT roles, but that's when I started to build more of an expertise around data and reporting and what have you. From there, I started to gain roles in the regulatory space. So data really grew up in the regulatory space and financial services with CAR as one of the major US based regulations that we had to comply with.

I've done CCAR, I've done Cecil. These are all acronyms related to capital planning and accounting and all of that. So I got really good on the regulatory side [00:07:00] of data. And then I pivoted into strategy, so that's when I started to. Hone my career around data strategy. I worked for Nike, I developed their data strategy.

I started to take on the role of the chief data officer, and that's when I was at Bank of the West in Silicon Valley Bank and LPL Financial, and most recently I was at t Rowe Price as their CDO. So I've, I've kind of had that like almost all versions of the chief data officer. I think that was like 1.0 to 4.0 and now the role of the data and AI officer is very interesting to.

To see how that's coming along. My role at EWS is slightly different. I'm not the chief data officer. I lead data strategy for the business unit that focuses on fraud monitoring and fraud data, and we serve that back to our financial institutions that we serve. It's a very interesting space to be in, and what I love about it is my role is [00:08:00] I'm developing a strategy for a data-based business.

So I feel like my era, my CDO era might be over and my new product era is coming up on board. So that's where I'm at. My trajectory, my career, the role of the chief data officer. And I certainly can share some of those stories as well as my new role, which is slightly different. 

hugo: I love it. And also, congratulations on the new position you've been at early warning for four months, three months, something like that.

cara: Yeah, a little bit. I think I'm just clearing my 90 days. I've developed, of course. They're like, I have a playbook for those first 90 days and develop their data strategy and working towards building a team and. It's very exciting. I think this is a great role. It's giving me a lot of opportunity to flex my product management muscles, which if anyone who knows me or hears this, they know exactly what I'm talking about.

So I'm very excited that I can be flexing those muscles. 

hugo: Awesome. And I [00:09:00] love that you mentioned your getting more and more interested in product because I, I think a lot of listeners will have heard of Zelle, which is something that Oh, yes. Comes out, everyone 

cara: knows Zelle. Yes, we are Zelle. So early Warning is the parent company of Zelle.

Which is the payments company pays, which is the Digital Wallet and Decisions Intelligence, which is the business unit that I'm a part of, and di, which we call the Decisions Intelligence is all about the rich data that we receive from our financial institutions, and it's the data science modeling and the fraud monitoring and all of that.

So it's a data-based business. 

hugo: Super cool. So I'm interested in how you go about. Defining the role of data in nearly 2026 in financial services, especially. 'cause it's an industry where we know, like accuracy, governance, trust, non-negotiable, not only, but also because it's federally regulated. 

cara: Yes. [00:10:00] Look, you have to define data as a strategic asset.

You can't think about it as an afterthought. It's not something that you just stuff in the closet and you forget about. You really need to focus on it as a strategic asset and you need to treat it like one. I love when people come up to me and they say, oh, you're the new data leader, or you're the new data person, whatever it might be.

I love data and I always respond to, I love that you love data. Do you take care of it? And they're like, ah, I'm a consumer. I, I, there, there are other people for that. And I don't think that's true. I think it's everyone's responsibility to be good stewards of the data that you manage, right? You get data from your email, from the websites that you're looking at, from the internal things that you might be working on.

All of this is information, and all of this is our responsibility to manage appropriately. So I'm a big believer in data governance and data [00:11:00] stewardship, but I also have a pragmatic slide. I do believe that too many rules is too many rules. We have to have a flexible way so that we can properly use the data at the same time, make sure we have good safety and security protocols.

hugo: Yeah, that makes a lot of sense. And there's a tension here, right? Because for it to be strategic and a first principle of any organization, it needs to be top down. In order for it to, for everyone to practice their own data hygiene and make sure they look after their data, it needs to be bottom up as well.

So I, I just wonder in terms of needing the top down and bottom up, how do you think about building an organization that is like a pragmatic, results driven approach to data and and governance? 

cara: That's a great question. I really think that it starts with the leader first. As that data leader, as that CDO, whatever your title might be.

If you are accountable and [00:12:00] responsible for the vision and mission of data at that firm, you need to be able to communicate to all levels of the organization, and you need to be willing to roll up your sleeves because data can be a dirty business. It's not easy, it's not straightforward. There's going to be complications.

And you need to be able to communicate the complex to your executive C-Suite, but also what does the executive C-Suite communicate that to your analysts? So I do believe in this, having a really strong communication around what are the overall goals of a data organization, what are the overall. OKRs and what are the things that mean most to the customer if you're a data-based product or service, right?

Those are things that are really important. I think also the thing that is really important is hiring good player [00:13:00] coaches, because oftentimes when you come into these organizations, you might be employee number one of the data company, right? And you've gotta build something from scratch. Not everyone is built for that.

So you really have to find out who are those key people to surround yourself with that are player coaches that they can play the role, and then eventually they can coach those players. So I do believe in having good communication skill, having a strong strategy, and then hiring really good player coaches.

Because in the beginning you're all players, even the leader. Everybody's on the field and everyone's playing the game. 

hugo: It makes a lot of sense. And last time we spoke you, we, you mentioned something that I find really fascinating and it's that there are challenge challenges that arise when people aim for perfect governance framework.

So I'm wondering if you could just tell me a bit about what an ideal [00:14:00] gov governance framework looks like and the pitfalls that happen when people aim for Perfect. 

cara: Yeah. I'm a big believer in progress over perfection and that it's a journey, right? So you can't just come out on day one and say, we're going to have an 18 point framework on data governance and data stewards will be actively remediating data quality and serving that back in to the controls framework and blah, blah, blah, blah, blah.

You have to show the road to get there. And so what I've done in my prior roles. Is I've established a very simple framework upfront, ownership, quality, and knowledge. I've said these three things and said, look, these are themes, right? Ownership is about what does it mean to be a data owner, a data steward, a data custodian, and here are the basic level of responsibilities.

Likely in your first year, you're not hiring a data steward or a data custodian. You're [00:15:00] assigning that role to someone who exists in the organization. When it comes to quality, start to identify the key data domains that are most important to you, and just say, okay, these are the five data elements that are the most important in the payment space or the identity space, or whatever your domains might be.

And then knowledge is really build a simple to follow fluency program, teach people about data, and don't be afraid to break it down into the simplest analogies. I'm a big believer in cleaning out your closets. 'cause data sometimes gets stuffed in the closet and forgotten about. And the role of the data officer, the data leader.

Is to come in and organize those closets into beautifully cataloged sections, and you can just go in and pick out the outfit or donate the clothes you don't need anymore. And you can really tie that [00:16:00] back to some data management practices and principles. To answer your question, I like to break it down into a simple three-prong framework with ownership, quality, and knowledge.

And then overall, it's really about. Being in this, like basic management, mature. 

hugo: So I'm also, I love the term data stewardship and I'm interested in how you can get everyone in involved out everywhere. I've, I've worked, most places I've worked, people have been very excited about being data stewards, but everyone's working at 120% capacity without even thinking about being a data steward.

So I'm wondering how you think about driving cultural and behavioral change so that people actually can just be more mindful with respect to their data. 

cara: So I, there's nothing I love more than ownership and stewardship because like my favorite element of a data governance framework, 'cause it's the people aspect, right?

This is all about inspiring people to take care of their data that [00:17:00] they own. Stewardship is all about data quality and remediation of errors. And so really finding the right people is important. So I think number one is, one, have a very clear definition of what ownership and stewardship means, and it's simple and easy to follow.

Everybody should get it right. Data steward is responsible for data quality, and then look for the right individuals in the organization that. Have been in the operations space, or they've been very hands-on? Actually, the people that I really like to bring in as stewards is customer facing individuals.

People that have been in client services, the frontline, they understand the customers pain points. That makes data so much more valuable when you tie it to customer experience and [00:18:00] customer impact. So that's the thing that I like about stewardship is I think it's, I think it's really important. I think it's a role that can be really hands-on and execution oriented and you can feel a lot of value at the end of the day.

It's hard when it's someone's side desk job. I really do believe in hiring separate data stewards that are implanted into the organization and that's it. I just, I love stewardship. I think it's a great role. 

hugo: And another great thing you're doing there is by tying it to customer experience and customer impact, you're tying it to revenue as well.

cara: Sincerely, I really think that you can take data and treat it like a product, and that's where the whole customer impact comes in, right? It's really important that you can tie what you're doing to the customer and how they feel at the end of the day. You gotta be able to do that. That's what's gonna make [00:19:00] any data strategy highly valuable all the way up to your executive C-suite.

hugo: Awesome. As we've mentioned, you've been working in high stakes regulatory environments for a long time where data quality impacts outcomes seriously, and I'm wondering if you can tell us a bit about what those experiences have taught you about what good enough actually means, and perhaps even talk us through some relevant examples.

cara: Oh, I've seen good enough, I think. There's been instances when I've wanted to build the most perfect data platform, the one data platform that's gonna solve it all right? I think I'm not alone in this, is that many data officers go on this journey to build the one data platform. I think what I've learned from being a CDO and doing this strategy over and over again is sometimes putting all the data in one spot.

Is, it's too long of a journey. So what I really like to do [00:20:00] is I've changed my strategy and have started to build things called data products, and I see that as almost the good enough solution. When you are on your data journey, you have to deliver value. At every year in your three to five year journey, right?

You can't. You can't just say, okay, I'm gonna go put data in a box and it'll be ready in two years time. That's not good enough. What you need to do is you need to give some quick wins, some early deliverables, build some smaller data product on-prem in a database. But that's good. That's still very good.

That helps the data scientists, that helps the analytics teams. Those things are good enough and to have a pragmatic data strategy that gives capabilities and products to your users, whether they're internal users or they're external users, [00:21:00] giving that early really gives value to, to, to your customer. So I do believe in picking apart the big platform idea and bringing it into products and say, okay, this product's gonna be delivered.

By this date, it's gonna have these data sets, it's gonna have these capabilities. Maybe it has a marketplace on top of it, but it's not the whole thing. I know. Do you understand what I'm saying? Yeah. So that, that's definitely, and I think with ai, that is something to really consider is the journey to put all your data in one place so that AI can run on top of that.

That's not always a strategy for success. I really do believe in certain data sets need to live together in a data product. Put AI on top of that. If you start to put all your data in one place, it's a long journey and you may never get there. 

hugo: For sure. And we may get to this, but it, what's happening now [00:22:00] isn't only putting AI on top of things.

It's embedding AI throughout, and even in building ETL pipelines or whatever it is, like AI permeating all layers of data systems, I am ready for the robots 

cara: systems. Yes, I think. AI is not new. LLM and Gen AI is new, but AI's been around for a long time. We've been doing predictive analytics for a long time.

There was RPA many years ago, which is robotics process automation. So AI has been around, I think with Gen AI and now agentic ai, there's a lot more capabilities. We never even dreamed before, and I think about how agentic can start to talk to each other and really create a closed loop type of decisioning.

There's a lot of opportunity. I think it's still early days. I think when people ask me, Hey, where do I start with ai? I say, start small, start safe. Use data sets that [00:23:00] are fairly clean. And that you feel you have a lot of high trust in and run chat GPT on top of that, see what happens. That's what I had done, certainly at some of my other firms, and those are really high value use cases to any organization.

Chat, GPT, you got your Microsoft Co-pilot, you got a lot of LLMs coming out now. Those are all really good things to just start to play with and see how it works. 

hugo: I totally agree. AI's been around since the term was first coined at the Dartmouth workshop in the fifties or whatever. I, and I've always thought ai, uh, I've always thought of AI as more of a marketing term than anything else, and I actually think it still is that, but it is more useful these days.

I, I think to use, particularly with the advent of generative ai, and of course agents have been around in a variety of incarnations for a long time. As well. I, I, I love that you mentioned starting clean, starting safe, and getting a small data set, and then building something on top of that and seeing [00:24:00] what your capabilities are, because that dovetails really nicely into something I wanted to talk about, which is financial data is extremely sensitive data.

So even in these examples, how do you think about the intersection of data security, encryption, tokenization, even when someone at a financial institution. Taking something like chat, GPT, of course, a secure version of it, quote unquote, whatever that may mean on top your internal 

cara: secure version. Of course.

hugo: Yeah. 

cara: Yeah. I think there's a bigger intersection more than ever with the chief data officer and the ciso. I think now's the time to really partner between the two teams. Data security, data privacy, data governance, the chief data office. Those teams have to be highly linked together. And when you think about data tokenization, it really runs across, right.

Your platforms [00:25:00] need to be safe and secure. We need encryption, we need tokenization. We need DLP. We need all of these different security and control capabilities. And also, when you're talking about financial data, you're also thinking about second line risk, right? Risks and controls. I was very lucky that when I started my first chief data officer job, I started that at a bank and it was Bank of the West, and I learned data through the lens of finance, like really becoming a good data governance officer, what's good enough for regulators, what best in class looks like, and how to build the roadmap to get there.

And so. I saw from an early like years in a, as in a CDO role, is that it's really important to be a leader that can tap into these cross-functional areas, and that data [00:26:00] governance is not just owned by one person. It's the responsibility of everybody and really making sure that all of these different capabilities can connect.

It's really important. So data tokenization, data encryption, those are all really hot topics these days, especially with cloud. And there's a lot of vendors out there that can give you those capabilities as part of their platform. I think it's all about making sure you're partnering with the right people internally in your organization, getting the right outside in perspective.

So talk to the vendors. Talk to the consulting firms. They've been there. They've seen that at other firms. I think that's really important, and I definitely think that the roles in data are now coming even closer and closer together between security, privacy, risk, governance, the data officer and the data steward.

hugo: Awesome. So we mentioned [00:27:00] AI briefly. Briefly, and so I think AI is just reshaping. Expectations A A across the board. And I'm just wondering to your mind, what foundational data capabilities need to be in place before a company can actually responsibly a adopt serious AI capabilities? 

cara: That's a good question.

I think your pay to play is you have to have a good data foundation. You have to have a good set of. Data products that are well organized, cataloged, defined, and owned. You've got AI is going to shine a light on data governance. I think that was one of the, as a, in the last couple years as AI was to really taking off and it was becoming part of every CEO's narrative.

I was thinking to myself. I feel for those data governance managers because unfortunately, [00:28:00] data governance isn't always a fully funded organization remit role. It's not always supported. And now with ai, there's just going to be a bigger emphasis on having data governance in place. So I think what you could do is you could pivot your strategy to be data enablement for ai, but.

However you wanna market it. I think your basic fundamentals of data governance need to be in place in order for you to really take advantage of AI on in a scalable way. 

hugo: And what are the base set of fundamentals there for data governance, do you think? 

cara: Oh, the base basics are having a good data foundation.

Data infrastructure, right? So having a good, solid data platform. Having a good data catalog that organizes the data. You know, AI needs metadata. That's how it like, understands what it is, right? [00:29:00] So metadata will be very important. Lineage is gonna be really important. Quality, right? Sometimes you don't have to have the best quality data, just like good enough 80, 80, 20 rule.

But AI is going to. You know, garbage in is garbage out. So it's gonna look at that data, it's gonna read that metadata, it's gonna report back what it thinks, and you don't want hallucinations. So that's important, that metadata and lineage and just having good data catalog and good data practices on how you define data.

Those are the basics, I think. 

hugo: Yeah. Great. So we mentioned briefly AI systems and. They've said, we've said this would be the year of agents. And as it turns out, it was in, in many ways a lot of exciting developments happened. But as Kafi said on the eSSH podcast recently, it looks like this is actually gonna be the decade of agents.

'cause there's so much room to to grow. And it really does seem like early days. [00:30:00] So I'm just wondering, what opportunities and risks do you see for agentic AI in enterprises and regulated en environments over the next 1, 2, 5, 10 years? 

cara: I think it's gonna amplify our productivity, not just from a individual user perspective, but from a business unit perspective.

I think about my day-to-day role, right? As a strategist. I have a digital intern, a strategy intern chat, GPT. I'm able to use our internal chat, GPT to help me start to define the data strategy. That's a huge help. Like it's a huge help. I think, and I always use it in a very safe way of course, because we have an internal capability around it.

But also in my personal life, I see, I use my external chat GPT in so many different ways. I ask it to develop a workout plan and a [00:31:00] help me plan my Christmas Eve menu, and I use it in so many different ways. It has. It has amplified my productivity. Now, I haven't measured that productivity. That's the thing that I think we're gonna, we're gonna, the bubble might burst, is we've deployed all the AI things and we've done ATech ai and we haven't truly measured the value.

I think that's the thing to watch out for and to always keep top of mind that if you're developing an AI strategy for your firm and you've got AI use cases. Capture the baseline capture. What is the process before ai, and then what does it look after? Because at the end of the day, like this is just technology and innovation, right?

We still have to do these processes, make sure you capture the value. That's really, I think, the bigger thing in all of it. 

hugo: Juan, what's the lowest hanging fruit, do you think? Particularly for places [00:32:00] like EAR early warning and even products such as Zelle. 

cara: I think that there's a lot of opportunity in the fraud monitoring space around ai, but we're very thoughtful with how we would go about that.

Right now, we are not deploying any of those use cases Right now we're using internal chat GPT for just employee productivity gains. But if I level up to maybe what I've seen in the past, you know, that other peers. AI to help investment research. That's always been, you know, any type of research or summarization or synthesizing that can get you information quicker in a more concise way.

I think those are things that AI and definitely LLM has certainly helped with. I do think that as we get more comfortable with Agen ai, it's going to help us with. Basic business processes that are just have taken us days cut down to hours. Like it's [00:33:00] just gonna be really interesting to see how I've taken a lot of manual processes into automation.

I also think, now that I'm thinking about this, and you got my wheels turning in my head, is that. Maybe we start to to, or we stop doing non-value add things, right? So if we feel like in order for us to continue these manual processes, we have to automate it, maybe these manual processes are not as important and we focus on the things that are important.

So hopefully maybe they, that will AI will help us recognize what's the most valuable. 

hugo: Yeah, I think it will help un uncover a lot of those different things. I also, this is something we've discussed with several people on. The podcast before, but it seems like just with the amount of stuff you can do with AI and the amount of experiments you can make, that an emerging skills important skill is not necessarily.

The ability to do an experiment, but the ability to, after doing a hundred, to spin down 98 or 99 of them. [00:34:00] And I'm wondering if that's in order to really focus on what can you value add, and I wonder if that's applicable to the financial space. 

cara: I think it, I think so because there's just so many business processes out there and having the ability to inventory those business processes.

Capture what are the time weights and the, you think about the six Sig. Six Sigma folks like Agen AI is like a six Sigma on steroids, right? And so I think there's a lot of opportunity for process improvement, efficiency, but also what I'm my hope is as humans is that we start to recognize that. Maybe not all of that is necessary and that we've, that we've focused on the high impact and the high value things.

hugo: Yeah, and I do think when starting out, I always encourage teams and organizations to build a matrix of risk versus value, and at least initially, don't go for moonshots and go for high value, low risk. Uh, [00:35:00] 

cara: go for things that are, start small, start safe. That was always my mantra. Start small, start safe. 

hugo: I also wonder, and I'm not asking you to talk about anything that's happening at early, early warning, uh, of course, but what space there is for agents to help with data governance?

And let me give some form of analogy or an example. I've actually been quite amazed. I wasn't short term bullish on agents and I've been proved wrong this year seeing. In the past couple of years, it was a couple of years ago, I was copy and pasting code from VS code into chat GPT and error messages back and forth.

Right then we had auto complete tab completion, that code completion stuff. Then suddenly we had products like Cursor and Claude code where I haven't, I can chat with an agent and it writes the code for me. Now we're seeing, um, agents embedded in. Slack where I can ping or discord, I can ping an agent and say, can you fix this documentation or open this pool request and it will do that.

We're [00:36:00] seeing them in CICD, right? Yeah. We're seeing more ambient agents and background agents. So I'm with things and I think this is a harbinger of what's coming for knowledge work more generally. And I'm just wondering your thoughts on the future of ambient agents or agents that are there to make sure data governance policies are applied well and this type of stuff.

cara: I think that's an excellent use of agen ai. If you go, I agree. I think there's a lot of opportunity for agents to be that digital data steward, that data governance bot that can go in and cleanse data and tell you, you can converse with it. I think, you know, it was just talking to our CDO not too long ago, well, today in fact, and she's talking about how.

We're not gonna have BI in the future. It's all gonna be ai. We're gonna be talking to our data. We're not gonna be looking at dashboards anymore. And I agree with her. I [00:37:00] think that when I use chat GPTI don't ask it to serve me up a dashboard. I'm asking it questions through prompt. So I think it's gonna be a different way to engage with data and the more of the agentic is, gets embedded in the data and the tools.

I think, like I said, it's gonna amplify productivity. You're not copying, pasting in all of that. And, uh, I think what we need to be careful of is those hallucinations and making sure that we're really looking at our results before we take a human action. 

hugo: It's so interesting that you say that, and it does show how it's such early days 'cause I'm so sick of conversations with ai.

I'm so sick of looking at text or hearing text. Because I do a lot of voice chat as, as well, and I love asking AI to show me dashboards currently, particularly how do you do Yeah, particularly [00:38:00] in, in code scenarios. And I do think, ah, part of the future of like software development may not be in cursor or vs code or this type of stuff.

It may be a brief chat with an AI system and it shows me visually what's up. I 

cara: think it's so interesting. So I have a 15-year-old son and I think it's so interesting. The future for him and vibe coding and building apps. And he's not gonna have to learn to code. He's gonna just talk to an AI and create.

What he thinks in his head, and that's amazing. I mean, we had to do it the hard way. I don't know if you remember the eye editor, but that's ridiculous now when you think about it. So I think there's a lot of really good opportunity ahead. I think we have to be eyes wide open on the wrist. Right, because we're feeding it all this information and it's mining it in the background and that's for [00:39:00] good.

But also there's always a flip side. 

hugo: I, I also, I agree that I think the service area of software is gonna, is expanding so much with vibe coding, with ephemeral software. The ability for me when building AI powered products to spin up a basic trace data viewer, see what's happening. A lot of frameworks, vendors doing incredibly good work, but.

Like they can't satisfy all the needs of this incredibly, no, of the application layer built over this incredibly horizontal technology. Having said that, I do think people who know at least one programming language very well will be SS Superpowered here. Yes. And I do, I half joke that maybe I'll start like a bootcamp, a coding bootcamp sometime like in a hermetically sealed cave on a Greek island or something like, totally like no internet.

No AI and teach people to code so they can come out and then be superpowered by agentic AI as well. 

cara: I like that idea. I think you should start that. I think that would take off 

hugo: awesome. Actually. Why am I going to, I think I just wanna go to a Greek island. [00:40:00] I could do it in New Zealand or the desert here as well.

You could do it 

duncan: forever. 

hugo: Yeah, exactly. I think I just wanna a holiday as well, late December, and it's been a long, long year, hasn't it? I'm interested. How do you. Think about empowering teams in your organization to use AI responsibly while maintaining quality and oversight? 

cara: I think it's, it, it starts with the, the customer outcome, right?

So if I have product teams, I'm talking about, you know, what is the outcome we're driving and does it need an AI innovation? And sometimes it doesn't. But I think it's empowering that product owner with the right tools. And the right capabilities to deliver on that outcome. So I really think it's more about being really knowledgeable of what's available and also what's, what is condoned by the organization, right?

You can't go rogue on this stuff. You can't just go pick up [00:41:00] something off the internet and say, oh, I'm gonna implement this internally. You can't do that. Maybe if you're building something in your side hustle time. Well certainly be aware. Take the time. I was, I was a big fan of Google Time. Did you ever hear the Google time?

You put some time on your calendar that's just focused on like external thinking and dreaming was the art of the possible and really get to know what tools are out there. I think product owners, think tech leads, they all should be doing that and learning about what should be possible. And then ask your organization maybe, maybe you have it already, maybe you don't.

Maybe. You know, you can get it through, through the AI intake process. There's a lot of opportunity now with AI tools and I think those product owners should really start to look at what is available to them and see how they can inject innovation into their product to deliver on that outcome. 

hugo: I couldn't agree more, and I do, I'll, I'll link to it in, in the show notes.

We [00:42:00] had a recently had an episode with Liz Koa, who's chief of innovation at the Behavioral Insights team. So like the Nudge group among other things. Right. And they're doing a lot of work on how to adopt ai, how to augment our capabilities with it. And they have an incredible four put paper called Augment, adopt, adapt, and Align on all of these things.

And one, one aspect is really the need to carve out time for experimentation. 

cara: Yes, exactly. And 

hugo: for dreaming. And to allow, yeah. Have systems which allow people to do that together as well as teams. 

cara: And there's tools to brainstorm, have AI help you brainstorm, like I think that's, those are all interesting things.

And to flex those muscles all the time 

hugo: without a doubt. Although I will say I don't think, I have not yet seen a good AI tool that is multiplayer where Hugo and Cara and someone else can chat with it and 15 turns later it remembers who is what. 

cara: That would be [00:43:00] great. This is. We shouldn't talk about. 

hugo: So maybe, maybe I won't even put it the, and I actually think the theme 

cara: podcast.

hugo: Yeah. I think maybe is our, the issue is in the instruction tuning, 'cause they do, they instruction tune on question, answer pairs among other things, which are just user ai, user AI system response, L LM response. So I wonder, maybe we'd need to do some fine tuning on it 

cara: would be so cool to have some kind of brainstorming AI.

Like, it's almost like a project team member that's with you that's like helping you whiteboard that has architecture background, that has an engineering background. Like you can load it up with all different personas and then it just becomes a part of your team. 

hugo: Yeah, 

cara: and it logs into teams meeting. I mean, think about that.

hugo: That's what I want and I've had the need, need for it. I haven't, yeah, so I'm also wondering. What advice would you give to data and AI leaders trying to balance innovation [00:44:00] with operational rigor in highly regulated settings? 

cara: I think right now it's a 80% operational rigor and it's 20% AI right now, or 20% in innovation.

I think that we need to start putting more innovation time on our calendar and starting to think about how can we shift that. There's a lot of opportunity. I know sometimes. As leaders, we feel almost bogged down by the need to be perfect and that the only way to get there is to be through that operational lens, through the tools that you have, through the processes that are laid out before you.

But there's opportunity and innovation and there's breakthrough. That glass, take a small use case and see if you can go fast on it. Those are the things that will start to make you believe that you can start to change more things in that direction. So I do think that you have to [00:45:00] make more time for innovation, especially if you're a leader that has to balance operational and innovation.

I always like having a small innovation team that's working on that kind of stuff. And they're thinking about the bigger strategy, they're thinking about the future. I think that's always good to have as part of your organization as well and push the boundaries in a safe way. 'cause again, we're regulated entity.

hugo: Absolutely. So to wrap up, I'm just wondering, looking ahead, what excites you most about the future of data governance and AI in in financial services? 

cara: I'm always excited about the automation. That's always been my thing is I love when. Processes become really well organized and parts of them become automated and you get to go home early, right?

Like the, I always thought about that even when I was just doing bi dashboards years and years ago. I always thought about, you know what? I'm making life simpler [00:46:00] for this one person that's taken all this time to build this report and report on these numbers. I'm really looking forward to the days where.

You know what? We're just conversing with ai. We're getting the information we need. It's immediate, it's accurate, it's well controlled, and sky's the limit on what we can do with that information. New products, new services, new markets, new channels. A lot of commercial opportunity ahead. 

hugo: Fantastic. Thank you for such a thoughtful conversation, Cara, and your time, and sharing all of your wisdom and experience with us.

cara: Thank you, Hugo. I really enjoyed this. 

Thanks so much for listening to High Signal, brought to you by Delphina. If you enjoyed this episode, don't forget to sign up for our newsletter, follow us on YouTube and share the podcast with your friends and colleagues like and subscribe on YouTube and give us five stars and a review on iTunes and Spotify.

This will help us bring you more of the conversations you love. All the links are in the show notes. We'll [00:47:00] catch you next time.