The Role of AI in Risk Management and Compliance - with Miriam Fernandez and Sudeep Kesh at S&P Global Ratings AI in Financial Services ยท 20 Oct 2025 Source: https://aiandbanking.libsyn.com/the-role-of-ai-in-risk-management-and-compliance-with-miriam-fernandez-and-sudeep-kesh-at-sp-global-ratings Transcript: https://podcasts.opennash.com/shows/ai-in-financial-services/the-role-of-ai-in-risk-management-and-compliance-with-miriam-fernandez-and-sudee-647a0c/ (via OpenNash Podcasts, machine-generated) Welcome, everyone, to the AI and Financial Services podcast. I'm Matthew DeMello, Editorial Director here at Emerge AI Research. Across financial services, the rise of AI is redefining how organizations measure and manage risk, but it's also creating new vulnerabilities, from market instability to the growing complexity of regulation itself. As financial institutions adapt AI to detect fraud, evaluate credit, and model systemic shocks, the challenge is no longer just accuracy, it's accountability. Regulators are evolving in real time, and in many cases, the private sector is being asked to take a more active role in maintaining trust and transparency. Today, we're joined by two leaders helping to define how that future takes shape. Miriam Fernandez, director in the analytical innovation team specializing in AI Research, and Sudeep Kesh, Chief Innovation Officer, both at S&P Global Ratings. Together, the pair bring unique perspectives, both from the analytical and operational sides of the financial ecosystem. Together, we'll explore how generative and agentic AI systems are transforming risk management, what new forms of regulation are emerging across global markets, and why human judgment and trust remain the most important variables in the equation. But first, are you driving AI transformation at your organization? Or maybe you're guiding critical decisions on AI investments, strategy, or deployment? If so, the AI in Business podcast wants to hear from you. Each year, Emerge AI Research features hundreds of executive thought leaders, everyone from the CIO of Goldman Sachs to the head of AI at Raytheon and AI pioneers like Yoshua Bengio. With nearly a million annual listeners, AI in Business is the go-to destination for enterprise leaders navigating real-world AI adoption. But you don't need to be an engineer or a technical expert to be on the show. If you're involved in AI implementation, decision-making, or strategy within your company, this is your opportunity to share your insights with a global audience of your peers. If you believe you can help other leaders move the needle on AI ROI, visit emerj.com and fill out our thought leadership submission form. That's emerj.com and click on Be an Expert. You can also click the link in the description of today's show on your preferred podcast platform. That's emerj.com slash E1. Again, that's E-M -E-R -J dot com slash E1. Without further ado, here's our conversation with Miriam and Sudeep. Miriam and Sudeep, thank you so much for joining us today. It's a pleasure having you. Pleasure to be here. Pleasure, Matt. And it's a pleasure to be here with such fantastic guests. Just starting from the top, it's especially great to have the team from S &P Global joining us to share their wide-ranging perspectives on the global economy, especially at a moment when technology is intersecting with every industry in deeper and more transformative ways. Maybe it's just because I'm 38, but I'm part of that small generation that saw life both before and after the internet. But I can't help but compare this moment to the early days of online adoption. Back then, not every house had access to the internet, or at least the same kind. Some had dial-up. We knew Wi-Fi was coming on the way. There was a point where nobody had dial-up, everybody had dial-up. Some people had Wi-Fi, then eventually everybody had Wi-Fi. It feels like we're at the same kind of inflection point with AI adoption. It's widespread, but maturity varies. Some organizations are just beginning to experiment while others are embedding AI into their core operations. Just with that in mind, and especially from what we're seeing in financial services. Let's start there. Miriam, how do you see AI transforming financial risk management today and what key benefits are beginning to rise to the surface? Yeah, as you were mentioning, this is in fact a journey for the financial sector that did not start yesterday. It started decades ago. And some types of AI, like supervised and unsupervised machine learning, have been used in risk management functions by decades to either for example uncover hidden patterns in data like it's the case in fraud detection or for example to help with credit scoring for banks or risk profiling for insurance companies but what what is new with generative ai is is really the capability to extract insights from data and enhance risk management done by humans so such data can be structured or it can be unstructured And what this means is that financial services companies can therefore expand the risk analysis, but not only using the traditional data that you have from the financial statements and from your Excel spreadsheets on interest rates and so on, but also incorporating unstructured data in different formats to uncover new patterns and extract insights. And an example of that would be, for example, in the insurance sector, you can consider a company that not only incorporates the analysis from the policyholder, but also could incorporate data from Internet of Things, sensors and devices, and like climate data and climate patterns, and therefore be able to manage in advance and prevent in advance catastrophes, climate catastrophes, like, I don't know, like river overflow. and things like that. So just to summarize key benefits, it's basically about the potential to incorporate that multimodal data from different formats with generative AI, and also make sense of it thanks to closer to real time data insights. And all of that, when you bring all of that together and you put that at the service of humans that are experts in risk management, And what this does is that it amplifies their ability to make more informed decisions and more informed risk management. Yeah, just in terms of on, and I can wear my innovation hat as well. Like, so I think like what Miriam articulated was just a lot of sort of this combination of opportunities and the manifestation of these opportunities in the financial industry's capacity. And it's very wide, as she articulated. And just in terms of process, I think one of the important things that we're seeing, this kind of goes to your analogy with the internet, right, is that this becomes kind of a force multiplier on how we do business and so on. So it's not just taking the data analytics and being able to sort of apply it to something, but it's almost, it's really gardening, right? So it's just kind of like when you're gardening, if there's mature trees and you want to take the fruit from the trees, it's all well and good. But you also have to focus on what is the nutrients in the soil and the water content and all that other stuff that will grow trees. And it's a cycle. So I think by using generative AI, it's actually producing new data, which is the utilization of this analysis and of these data, machine learning and algorithms and so on. And now we're basically having some cognizance of what it's being used for. And then I think once you basically kind of zoom out quite a bit, a lot of industries, financial services or otherwise, are able to learn from each other's learned experiences using these kinds of things because we're using it. So I think it's a combination of sort of that linear progression in terms of using it for risk management and then augmenting with capability generation. But it's also kind of, you know, recursing back to be able to do a lot of things in new ways. Absolutely. I think what stood out to many listeners, and sometimes I get a sense of what they're thinking as we're going through the course of the show, is that theme of interconnectedness and how quickly information moves today. We've had about 15 years of flash crashes now driven by technology accelerating the flow of data faster than ever. AI is only going to amplify that. It's going to be part of the problem, but confusingly for many, part of the solution as well. That's where regulators and institutions tend to step in, trying to bring some stability to parts of the market that need it most. What steps are we seeing regulators and institutions take to mitigate these risks? So I think in terms of systemic risks, I see two vectors that can bring systemic risks from AI implementation. And the first one is around concentration risk. And that stems from basically the market tech oligopoly that we see around chips, around cloud services, around AI models and AI software. This is not something that is completely new. we saw it a few years ago already with cloud, but it's now with Generative AI, it's certainly posing some questions for financial institutions to diversify their risks in their relationship with third party providers. Any link to that concentration risk, I think it's also worth mentioning the interdependency that financial players have in their digital interactions. So banks interact with clearinghouses, with asset managers, with fintechs. And if one of them suffers from a cyber incident, and we know that now with generative AI, it's easier to perform those malicious attacks, then that contagion risk can come along and it needs to be managed by the best. So this is the first vector. And then the second vector is not any risk per se, but it's more the capacity of AI to be an amplifier of financial instability. And how can this happen? So basically, if we think about autonomous AI systems that are performing processes from end to end and with human supervision, but not direct human supervision. and this could be the case of, for example, algorithmic trading, we can see there that if the models are not well managed, we could see an increase in complexity and an increase in opacity of the workflow. And then this could accelerate the spread of contagion risk because in the end, these are highly fast transactions that happen in milliseconds. And then, therefore, contagion risk can be accelerated. So, therefore, leading to financial instability. Yeah, and Miriam hit the nail on the head in terms of really the interconnectedness. With the vector of digital, it basically breeds even more interconnectedness. Now, I think what's a relevant question is also is what is there to do about, you know, some of these different issues and so on. because in some cases it's basically it's a structural change in the interconnectedness of the system itself because of this you know this this digital vector which is somewhat you know permanent right it's just kind of it's a it's a structural this is this is what evolution looks like but then as part of that evolution you have to take it into to to in stride that that interconnectedness breeds more systemic more and different systemic risk as miriam articulated Right, right. Another way I've heard these dynamics described, and I'll give direct credit to a good friend of the show, Nick Lewis, who leads one of the financial crimes units at Standard Chartered Bank, is through the lens of shifting responsibility. With so much political pressure to cut bureaucracies across agencies, especially in the West, the private sector is using AI to streamline its own operations. As a result, more of the enforcement burden is being handed to private institutions. In Europe especially, we're seeing banks take on a larger role, tracking know-your-customer data, KYC data, staying ahead of emerging risks and keeping regulators informed. And regulators, meanwhile, are acting more as overseers than operators. Their position is keep us updated, share your data, and we'll determine whether something crosses a threshold of criminality. We'll grade your homework, basically. But it's the financial institutions that are expected to spot those red flags first to do the homework. Sudeep, what other innovations in regulatory strategies are you seeing take shape in the future of AI and finance? Sure. Yeah, yeah. And so I think that there's been quite a number of regulations and you basically see different regulations in various parts of the world with respect to, and this kind of, this is predicated on the fact that all of this is complex. So you have financial market regulation on top of communications regulation, on top of AI decision-making regulation. And that manifests itself in various different ways, depending on the jurisdiction. You're operating in the nature of the economy, the interconnectedness of the system, and so on. Some of the common things, I think, are connecting on intent. So in Europe, for instance, you have this principle of do no significant harm, right? So that's kind of embedded in the EU AI Act, for instance, and some other parts of regulatory proposals and otherwise, where basically the intent matters. It's not just what you're doing, but why you're doing it. And if you're going to do this with some element of malicious intent, that's bad. And that's going to be regulated against. What I think you're going to start to see is a bit of an evolution, though. And this goes across a lot of geographies. A lot of regulation right now is punitive in nature. So what I mean by that is that you do something bad and then you have to basically pay a financial penalty to restore the injured party, for instance. In the case of AI, some of these things are actually very, very catastrophic. and they're irreparable. So I think what we're going to start to see is this regulation, which is aimed at basically how can you facilitate interconnectedness among parties, so between governments and private institutions and nonprofits and so on, to be able to prevent harm as opposed to rectify harm. So that's a bit of sort of a future thing, but I think it's coming to a head with a lot of the issues around cybersecurity, the interconnectedness of these different systems, disclosure standards requiring a lot of companies or organizations, I shouldn't just say companies, requiring organizations of significant size to really help with sort of AI adoption and AI learning programs and things like that. So we're all better educated on how to sort of keep ourselves safe, how to keep our organization safe, and then therefore keep society safe. That's becoming more and more of a priority. Now, I think that there's going to be a lot of changes in terms of, you know, already happening in cybersecurity land with some of the requirements of the SEC, for example. So I think it's reasonable for organizations to really think about what systems they need in place to be able to kind of document their intent, their processes, how these things are used to provide some transparency where it's required, or at least sort of what I call internal transparency, where you can at least connect that back to some of the learning programs and so on. So that way, everybody's kind of operating from the same playbook or the same songbook, so to speak. And I think this is a little bit of an offshoot of your question, but I think it's pertinent is, you know, around cybersecurity, then you start to, you have to think about this quotient of trust. So basically, what's that quotient for your stakeholders, whether they're customers with regulatory, you know, authorities and so on? How do you operate in a way that basically is conducive to trust? And if you do that, even if you were to have a cybersecurity incident or something like that, or something where artificial intelligence has ended in a less than ideal outcome or something, if you maintain trust, it's still manageable. If you're actually negligent, then it tends to erode trust. And then that can actually be existential for a company or for an organization, because a lot of that value is derived in that trust. So I think that's increasingly going to have, whether it's specific regulation in terms of law, or if it becomes on a first principles basis, this is the right thing to do. I think you're going to start to see a change in that conversation. Just a couple of points, particularly from the European lenses, from the European regulatory lenses. I think it's important to highlight that there is a risk-based approach, right? So the higher the risk, the higher the requirements. We have already seen that with DORA, the Digital Operational Resilience Act, and how, for example, for critical third-party providers of tech for financial institutions, there are additional requirements and assessments to be conducted. as part of the law. And we are saying that with the EU AI Act and the foundation models, for example, they require special systemic risk assessments and adversarial testing that on top of what we're seeing for other models that are not foundation models and not that large preset. Yeah, I think that really highlights one of the advantages of these systems moving through the private sector first and reaching regulations later. We want to be sure the technology is fully vetted before it ends up in the hands of people in greater positions of power. It's also important that the tools are tested, proven, and understood well before they're used at that level. There's a lot of perspective to unpack there. Go ahead, Sudeep. Well, I mean, and what you were describing, as well as what we've kind of seen as an evolution over the past, say, 25 years or so, it's really been the financial services being an ecosystem. So it's basically, there's different participants in that ecosystem, ranging from providers of capital to users of capital, and the regulators are also part of that ecosystem. So it's not layers, I would say that there's basically, there's interconnections, and there's a lot of different responsibilities, which are shared, because it's one of these things that, like, I think it's implicit in the gentleman's point from standard charter is financial institutions have to protect themselves in terms of trust and goodwill with their customers and so on. So by actually sort of taking proactive steps to be able to help protect information and prevent harm is actually good for their stakeholders and then subsequently good for their business as well. And I think it's this notion of compliance not just being a regulatory box that you have to check off, but actually being something that's good for you. It's almost like the exercise that you have to do to have better governance processes and so on. I think that's just coming to a head in terms of just advancement of the ecosystem itself and recognizing that these systems, maybe historically, were always interconnected. But now digital technologies are giving us the opportunity to see those connections and actually sort of help develop and maintain those relationships in kind. Yes, I mean, obviously, financial institutions are also, as Suleep was mentioning, part of that ecosystem and embracing AI adoption as part of the innovation. I think one of the topics that we're starting to hear more and more is agentic AI and these autonomous systems that could be transformative in the sense of filling the gap between the generation of insights versus the execution. And that brings that brings that close to the gap in terms of coordination, execution and also learning from the environment. The environment can be digital, can be physical, can be ingesting regulatory insights and from the supervision perspective, incorporating that as well into the analysis. And that goes both ways, right? So it also goes to the regulator. And regulators right now are also starting to use generative AI as part of their processes to classify documents and interpret data. But something that I wonder, and I think we're going to go there, it's around how the regulator is going to also conduct the analysis of the AI risk models from institutions before you were mentioning that self supervision. but it's also true that it looks like some new metrics are going to have to be incorporated also from a supervisory perspective to prevent biases and to prevent certain risks. And I like to call one in particular that I read recently from an economist. His name is John Danielson. He's a professor from the London School of Economics and specialist in systemic risk. And he calls for some homogeneity scores, which basically analyze the percentage of AI systems that are using the same or similar training data sets across different counterparties. And this could be financial sector institutions, for example, because in the end, if you're using the same training data and the training data is biased or has some sort of discrimination factors, this could lead to hurting behaviors in the market and therefore systemic risk. So this is just an example of things that need to be considered for the future, because as the technology evolves, so we'll need to evolve the supervisory metrics to be able to analyze those models. Yeah, but I think that it's not necessarily the case where you have sort of this first mover advantage in the traditional sense, because now with a lot of these technologies, you're seeing this kind of renaissance of open source technologies, first of all, but then also a lot of these things that are basically, they're bespoke developed solutions to problems that others may have. But then they're basically the usage of them is put in the public domain. And then you have sort of these these economies of scale that are not just deep, but they're also wide. And I think so. So I don't necessarily think that it's necessary. It's there's no guarantee that a lot of these technologies will be developed privately by a financial institution. First, you may you have a lot of sort of this crosstaff. But similarly, I think it's one of these bits where autonomous systems and intelligent autonomous systems are very much being actively developed right now. And because in the concept of, you know, AI agents, they're basically generating their own trigger for, you know, different decision systems and so on. the notion of accountability becomes more, you know, ever more important where, yes, a system is making this decision and operating as a, you know, in sort of agentic or custodianship of our own sort of interest. But ultimately, we as humans need to be accountable for these things. And I wouldn't accept any substitute for less than that. But I think that that's one of these things that we're going to have to have an increased dialogue about what that specifically means, because I think that's the bit that's unprecedented is, and to your point, you know, we've had algorithmic trading for a long time and so on. That's more autonomous, not intelligent automation. But there's this concept and notion of intelligent automation actually goes at the decision science. And that accountability, I think, needs to be pretty clearly defined. And I don't think that's just the responsibility of a regulator or just the responsibility of a bank. It's basically all of our responsibilities to get some commonality of definitions. Right. And I really appreciate the nuance you're bringing there, especially since in a 25 or 30 minute podcast, we inevitably have to cut ourselves off before giving every topic the depth it deserves. Of course, there are plenty of stories in the news about regulators using AIs in ways that may not be fully tested. And that's probably a whole separate episode, right? We might need to bring you both on and have you both back for that one, though I imagine the comms teams might have more to say about it. Still, it all reinforces an old saying, right? Nobody has a moat right now to sort of quote that Google memo towards the beginning of the generative AI hype cycle that we saw a few years ago. That applies not only to major financial institutions, but to regulators themselves. It's a brave new world, and we're going to have to watch it evolve in real time. There's so much more to explore, and I'm sure we'll have plenty of material for future episodes. Miriam and Sudeep, thank you both for joining us. It's been a real pleasure. Thank you for telling us. Thank you. Before we wrap up, let's recap a few critical takeaways from our discussion today with Miriam Fernandez and Sudeep Kesh of S&P Global Ratings. First, AI is expanding the boundaries of risk management. Financial institutions are no longer limited to traditional data sources. With generative and multimodal AI, they can analyze structured and unstructured data from policy documents to the Internet of Things and climate sensors to surface insights faster and anticipate risk before it compounds. Second, systemic risk is evolving right alongside innovation. As financial systems become more interconnected, dependence on the same AI models, chip providers, and cloud infrastructures increases exposure to contagion and concentration risk. Regulators are adapting, but institutions themselves are taking deeper responsibility for managing stability across digital networks. And finally, trust is emerging as the new foundation for financial governance. Whether through self-supervised AI models or autonomous agents, accountability remains human. The institutions that embed transparency, documentation, and responsible AI practices into their workflows will be best positioned to lead in the next phase of global finance. Are you driving AI transformation at your organization? Or maybe you're guiding critical decisions on AI investments, strategy, or deployment? If so, the AI in Business podcast wants to hear from you. Each year, Emerge AI Intelligence features hundreds of executive thought leaders, everyone from the CIO of Goldman Sachs to the head of AI at Raytheon and AI pioneers like Yoshua Bengio. With nearly a million annual listeners, AI in Business is the go-to destination for enterprise leaders navigating real-world AI adoption. You don't need to be an engineer or a technical expert to be on the show. If you're involved in AI implementation, decision-making, or strategy within your company, this is your opportunity to share your insights with a global audience of your peers. If you believe you can help other leaders move the needle on AI ROI, visit emerj.com and fill out our Thought Leader submission form. That's emerj.com and click on Be an Expert. You can also click the link in the description of today's show on your preferred podcast platform. That's emerj.com slash E1. E as in Eric. Again, that's E-M -E-R -J dot com slash E1. We look forward to featuring your story. If you enjoyed or benefited from the insights of today's episode, consider leaving us a review on Apple Podcasts and let us know what you learned, found helpful, or just liked most about the show. Also, don't forget to follow us on X, formerly known as Twitter, at Emerge, and that's spelled E-M -E-R -J, as well as our LinkedIn page. I'm your host, at least for today, Matthew DeMello, Editorial Director here at Emerge AI Research. On behalf of Daniel Fagella, our CEO and Head of Research, as well as the rest of the team here at Emerge, thanks so much for joining us today, and we'll catch you next time on the AI and Financial Services Podcast. Thank you.