How Financial Services Leaders Operationalize Safe AI - with Dr. Oscar A. Rodriguez of Citi

**SPEAKER_1** (0:12)
Welcome everyone to the Emerge AI in Business Podcast. Today's guest is Dr. Oscar A. Rodriguez, Vice President Data Analytics at Citi. Dr. Oscar joins Daniel Faggella, Emerge CEO and Head of Research on today's episode, to explain why AI breaks down inside financial institutions when teams compete to be first, filled in isolation and push models forward without shared definitions or ownership. He argues that AI only survives past the pilot stage when governance and accountability are treated as built-in operating habits, but not after the fact controls. So organizations know exactly who owns decisions when models fail or regulations shift. Today's episode is sponsored by Securiti AI. In this episode, we cover how AI breaks when governance isn't built in early. To go deeper on this topic and learn how financial institutions are digitizing paper-based records to unlock usable data for AI and using alternative data like public web and social signals to enhance risk assessment, download our free PDF report. AI in Financial Services, Executive Cici, at emerge.com/fcs1.
That's emerj.com/fcs1 to download your copy. Now, the conversation with Dr. Oscar.

**Daniel Faggella** (1:44)
So, Oscar, thanks for being here.

**Dr. Oscar A. Rodriguez** (1:46)
Oh, thank you for having me. My pleasure to be here, thank you.

**Daniel Faggella** (1:49)
Absolutely, well, your background is in a space that is only going to be getting more attention every five minutes, which is sort of the security and AI intersection within financial services. FinServe is way ahead of the ballpark in terms of seeing the fundamental changes ahead. I want to start with just sort of what holds back AI projects sometimes, I'll ask it for you like this. When you look at the patterns of what stops AI initiatives from moving forward within the regulated financial institution world, and I guess particularly in the security intersection space where you hang out, what do you see as those big hurdles today?

**Dr. Oscar A. Rodriguez** (2:20)
You know, this is a great question, and I've seen it since day one, when I started out with AI and implementing AI, is there is a lot of enthusiasm and I can't blame them for that. But the pattern I'm seeing is that enthusiasm repeatedly outpaces the readiness. I remember situations where there were multiple business areas and they wanted advanced analytics or AI capabilities at the same time. But when you look underneath, each group was working from different definitions, right? So they had different data sources, they had different priorities, and the technology itself wasn't the bottleneck. It really was, as simple as it sounds, alignment.

**Daniel Faggella** (3:05)
I guess let's kind of drill down into what alignment means conceptually, then maybe we'll tie it specifically to, again, some of the issues that might affect folks in kind of the security ecosystem. Alignment, when I hear it, I sometimes think about, okay, does executive leadership, who has some kind of ROI goal and maybe a bigger strategy, does that line up with the folks within whatever the department is, who know the outcome that they're going to try to do?
And then the boots on the ground folks that are going to use this new tool, or maybe their existing tools are going to be augmented, do they have buy-in? Is this going to help their workflows? Do they feel a sense of ownership? You know, when you say alignment, those things come to my mind for you. So, what comes to your mind here?

**Dr. Oscar A. Rodriguez** (3:43)
Yeah, for me is, you know, teams experimenting independently. And what I mean by that is, you know, there's sometimes, and I've seen it firsthand, different teams within one organization wanting to get there first, wanting to be the ones to tout and say, hey, we solved this problem. While leadership is probably focusing on different things, different ideas, probably a little more concerned with what the future is bringing, right? In terms of risk, compliance, teams are creating just duplicate effort with inconsistent standards and competing priorities.
And I can't stress this enough, AI is only going to be as good as the data behind it. I'm sure you know this and listening knows this.
Financial institutions, for example, they often have customer risk compliance and operational data. And it's right across disconnected systems. And that will limit the model effectiveness. So when you think about it a little deeper than that, right? I also can go a layer beyond that and talk about governance. I would say that that frequently is introduced after the model has already been developed. So kind of what I meant earlier on, right? About teams are so enthusiastic about throwing it out there. They'll build it out. OK, well, where is the governance behind it? They'll develop it first, and then they'll try to adjust and append all these governance in order for there to be risk compliance.