3564: Why Banking Is the Ultimate Test for Responsible AI

[00:00:04] Welcome back to the Tech Talks Daily podcast and today's episode is going to take us into the world of community banking, risk and the realities of deploying AI in one of the most tightly regulated environments imaginable. His name is Ravi. He's the Chief Product and Technology Officer at a company called Abrego. They are a fintech company supporting over two and a half thousand banks and credit unions across the United States.

[00:00:32] But today I want to explore how AI is moving beyond surface level automation and into the very core of lending, compliance and financial decision making. A world where trust, transparency and accountability carry very real world consequences. No room for mistakes.

[00:00:54] So whether it be invisible AI assistants working behind the scenes that we don't see or the importance of explainable models in high stakes decisions. Today's conversation will offer a grounded look at how technology can strengthen rather than destabilize the foundations of modern banking. Here at the Tech Talks Network, we now have nine podcasts and approaching 4000 interviews.

[00:01:20] And that is only possible with some of the great friendships that I've developed over 10 years of podcasting. And a company I'm proud to call friends of the show is Denodo. Because not only have they been on this podcast multiple times, they also help make sense of the AI data chaos that we're seeing now. Because the data world is louder than ever. AI hype, lake house complexity and pressure to deliver more with less. These are things that I talk about every day on this show. But Denodo is helping businesses make sense of it all.

[00:01:50] Because they provide a unified data foundation for trustworthy AI. So if you're ready to unlock real outcomes, simply visit Denodo.com today. But now it's time for today's interview. Let me introduce you to today's guest. So a massive warm welcome to the show. Can you tell everyone listening a little about who you are and what you do? Well, good morning, Neil. And thank you for having me on the podcast. My name is Ravi Namalikanti.

[00:02:20] I'm the Chief Product and Technology Officer at Abregal. We are a fintech company that serves over 2,500 banks and credit unions across the U.S. So that's roughly one in four financial institutions in the country. And that gives us a great mission. And it's actually quite simple to help community financial institutions manage risk and grow responsibly and stay ahead of innovation the best we can. And I'm sure we're going to talk about how we do that.

[00:02:49] My role is to make sure technology becomes an enabler of trust. Because, again, we are serving the community banks and the local communities primarily. That means using technology safely, data safely. And it doesn't matter which technology it is. Today we are talking about AI. You know, a few years ago we were talking about the public cloud, right?

[00:03:16] So it's about using technology safely, transparently, and effectively in a highly regulated environment. And our financial institutions, our customers, depend on solutions. Just to give you a flavor of what all we do in lending and credit risk, lending origination, financial crime prevention, and portfolio risk management. Those are the primary categories of products and solutions we serve up in these markets.

[00:03:46] And these are, like, abstract problems. They truly affect how money moves, whether businesses are getting funded, are we stopping fraud, how some of the credit decisions are being made, et cetera, et cetera. I mean, if you put all of that into a nutshell, it's really how local economies thrive. And just very briefly about myself, I've always been fascinated about technology and how that can elevate and amplify rather human judgment and not replace it.

[00:04:15] That's what excites me about AI specifically in banking. And it's truly not just about automation. It's augmentation. And the most successful systems that we've built over the last few years, they are all, their mission is to make people better and the workflows better, right? And the overall customer experience is top of the line. Well, it's a pleasure to have you join me on the podcast. So much we're going to be talking about today.

[00:04:42] And I think for the last three years, we've been continuously talking about AI, and we look destined to do the same in 2026. But now we're moving beyond the hype. We're talking about things like trust and transparency, as you mentioned a moment ago. And also, many people are starting to talk about responsible AI in broader terms. But despite that, banking still feels like a real test bed because mistakes have very real consequences.

[00:05:09] And it's a very risk-averse business, and quite rightly so. So why do you think financial services might be the strongest proving ground for AI that is both safe and useful? What are you seeing here? Yeah, well, Neil, that's a great question. Look, in banking, responsible AI isn't just a slogan that sounds good. It's actually in banking specifically, so it's a requirement.

[00:05:36] Banking is, again, one of those few industries where every decision, irrespective of the technology that we're using, has to hold up to whether it's regulatory, ethical, or any kind of operational scrutiny. So it's like a perfect proving ground because the stakes are so high. For example, if you take a false positive and fraud detection, that means we've added another friction point to the customer experience.

[00:06:04] Or maybe an unexplainable or an opaque credit decision can affect a small business's longevity or a person's life or even a bank's reputation. So it's really where trust meets the true business model, right, that gets banking and banking customers up and gone.

[00:06:26] Beyond that, what makes financial services unique is the combination of a lot of unstructured data, a lot of legacy systems, especially in this nation, right? There's a lot of cores that are 40, 50 years old, the bank cores that are still in operation. And pair that up with quite a bit of governance and accountability. So every model has to be explainable.

[00:06:53] It has to be auditable and underlined with the compliance standards, right? The compliance standards are not changing just because we are deploying AI. They're the same standards. They're the same expectations, whether it's a regulatory expectation or a customer expectation. So that naturally pushes that innovation toward that responsibly AI. And that's why it's not a slogan that you have to have explainability, fairness and traceability just built into every workflow.

[00:07:21] It's not a feature that you add once you build the technology and the capability that sits on top. I often say that AI can earn trust in banking, it can truly earn the trust anywhere. So that's why I think banks are really at the intersection of trust in AI and the deployment of AI, if that makes sense.

[00:07:46] And before you join me on the podcast today, I was doing a little research on you and I was reading how you mentioned the rise of AI assistants inside banks are often running behind the scenes. So from your perspective and everything you're seeing here, how are these assistants really changing day-to-day financial work? And what early patterns are you seeing in efficiency or quality?

[00:08:07] And the reasons I say that is I think this year there's been an increased focus on, hey, let's not talk about AI so much and talk about the measurable impact, the difference we're making, the problems we're solving. So what are you seeing here? Yeah, another fantastic question. So what's truly happening inside banks or credit unions today is fascinating because most of the AI adoption is invisible,

[00:08:32] meaning behind and within the banks or credit unions back office. So it's not the chatbots that you see on a website. It's truly the assistants that are embedded into the workflows. It could be a lending origination workflow. It could be a anti-money laundering workflow. For example, take a lending assistant.

[00:08:53] It can summarize a 60-page financial statement in seconds or maybe flag anomalies in the way the statement is structured. Maybe the totals are not adding up. Maybe flagging potentially a fraudulent document. Or maybe even suggest which key ratios to review.

[00:09:14] So this is all grunt work that analysts have historically done, which frees up the analysts to focus on more judgment calls instead of data entry, data completeness, et cetera. So that's why in compliance or in lending, these assistants are truly taking those workflows to the next level. And these are all happening behind what a customer might be experiencing when they are interacting with the bank. And the same is true, again, in a compliance workflow.

[00:09:45] AI assistants, we have an AML assistant that can scam the transactions, that can write or understand what a normal looks like and write a summary of potential transaction that could be fraudulent or should be flagged as an alert that a BSA officer should be taking a deeper look at. These are things when you're looking at a volume of data that a BSA analyst has to look at. We are all human.

[00:10:12] We are not consistent in how and how much focus can we apply during an eight-hour day. Well, for AI, it doesn't matter. As long as you're deploying compute to the AI, the work is consistent and the output is consistent. So that's what we are truly seeing in the adoption of AI. And as we look at this, the last two things that I mentioned in terms of quality and consistency are rising to the top,

[00:10:38] which when we started deploying AI, we thought that, yeah, those are good additional features. But when we started talking to customers, they got very excited about the consistency of a narrative. It could be an alert narrative or a loan memo, right, and how structured it is. And our customers got excited about that, right? And that's what we're seeing, that sometimes you don't really understand where the usage will fit a customer's need and their workflows.

[00:11:07] And the less time to add to all of that, the humans spend less time on all of this routine work. It'll help them to focus on those anomalies and investigate those edge cases, Neil. And when it comes to technology and adoption, there are a few contradictions in the industry. Because although banks usually move very carefully with new technology, I think it's also important to highlight that they tend to adopt modular systems in ways other industries are much later to follow.

[00:11:34] So are there any other sectors or what can other sectors learn from how community institutions are approaching AI without creating unnecessary risk? Because it is an industry that seems to have mastered the art of this. Yeah, you're absolutely right. And when you refer to community banks, they don't have the budgets that a JPMorgan or Bank of America, they have. They're spending billions of dollars, right, to stay ahead of innovation.

[00:12:05] And when it comes to community banks, they are incredibly pragmatic innovators. They don't have, again, these massive R&D budgets. So they innovate. And as their partners, we innovate with that intent, right? These are small, modular steps that can be measured and untrusted. So what other industries can truly learn as you look at this is that modernization doesn't require a full rebuild or a full reimagination.

[00:12:34] So, again, going back to a community financial institution, they have these processes. They've had these processes and workflows for 30, 40, 50 years. They haven't changed much. And just because we have a new technology that can inject, let's say, 100% efficiencies, they cannot afford to just disrupt the customer experience, disrupt the operational workforce. So the answer might still be, you might still be 2x, 3x productive in a matter of time.

[00:13:04] But how do we go from that point A to point B and do that in a measured and trusted fashion is where I see an incredible amount of innovation happening, right? Not only in applying the technology, but the cautious way in which they deploy the technology. So they're still taking advantage of the innovation, but doing it in a very customer-centric fashion.

[00:13:30] So it's really low risk, high return use cases is where they typically start. And another lesson that I would maybe take away is just working with financial institutions day in, day out is that governance mindset. These institutions, they already operate under very strong risk management frameworks that you alluded to. And when they are evaluating a new model, it was machine learning models in the past.

[00:13:58] Now it's the AI models, right? They ask this exact same question. Can I explain this to a regulator? Can I trace its data lineage, right? Those are the questions that drive that safe adoption. So the beauty of the community banking model is that it's innovation in service of that trust and the stability. And as you said there, explainability is such a huge topic in regulated environments.

[00:14:26] So when you think about models that must show their reasoning, what does the explainability look like in real banking workflows rather than just theory? Anything you can share around that? Yeah. Again, this hasn't changed, right? Irrespective of the technology, we have to, our customers have to explain every single decision. It cannot be that opaque box, right? That you cannot go behind, right?

[00:14:54] You have to have very strong rationale and reasoning that's actionable for the person using the technology. Take credit risk as an example. If a model flags a borrower as high risk, the loan officer needs to see why, right? Is it a drop in cash flow? Is it an industry downturn? Or is it maybe increased credit utilization?

[00:15:22] So there's 30 different drivers that could flag a borrower as high risk. We have to know before we flag the decision, right? Primary decision to the loan officer. That level of transparency lets them validate or maybe even override based on the relationship that they may have with the customer or even contextualize the result, right? So that's where explainability comes into play. And at Abrego, we take that to heart.

[00:15:51] We design systems that surface these insights visually. If you're calculating a score, for example, showing those key drivers, the confidence levels, or maybe even peer benchmarks. So the humans that are, at the end of the day, responsible for the decision, they stay in control, right? And they feel that they have the information to be able to make, especially when it's an adverse decision.

[00:16:18] We also try to separate the technical and business layers of explainability. For example, if you're talking to a data scientist, right, they talk about future importance and validation. Well, you can't have that, right, with bankers. When you're talking to bankers, it has to be in a language that they understand, right?

[00:16:39] What's driving the scoring and going back to the different drivers that we talked about earlier, are there things that we can put in place as mitigating factors, right? So that's the philosophy that drives us, that in regulated environments, AI shouldn't just make these decisions, but rather show X work, right? That's how you earn the trust of both the regulators and our customers. This month, I'm partnering with Alcor.

[00:17:05] And if you've ever tried to hire engineers in another country, you probably know just how painful it can be. Different laws, patchy support, and partners who don't truly understand engineering roles. So Alcor approaches this from a different tech point of view. They specialize in Eastern Europe and Latin America, and they're able to combine EOR capabilities with recruiting.

[00:17:29] So you get one partner handling everything, and they help you choose the best location for your stack, find developers with the right depth of experience, and run proper assessments so they can onboard people quickly. And they also give you a model that respects both transparency and margin. Most of your spend goes directly to your engineers, and the fee will decrease as the team expands.

[00:17:53] And you can even transition everyone in-house at that time when you're ready without having to worry about a penalty. And that structure is why a mix of early stage and unicorn stage companies use them as they scale. So if you want to take a look, visit alcor.com slash podcast or tap on the link in the show notes. But now, on with today's show.

[00:18:16] And in your work, you do do a lot of help in helping community institutions operate through, I think, what you call a perfect storm of pressures. So how is AI helping some of these smaller organizations keep pace with regulatory change, shifting customer expectations, and the arrival of new kinds of competition? Because it is an incredibly delicate balance. But at the same time, it's not just the larger banks, the larger community institutions that you help.

[00:18:45] It's all about those smaller and medium-sized organizations too, isn't it? Yeah, well, perfect storm is very delicately put now. It's not just AI, right? If you think about just banking more broadly, we are talking about deploying AI and maybe that changing the customer experience. And on a completely different tangent, we are seeing the rapid adoption of stable coins, tokenized deposits, right? That might change how money moves, right?

[00:19:15] Which is bread and butter for community banks. So this is a lot going on in the space. And yeah, you have to stay ahead of this, which means you have to find means in which you can become more efficient and then use the time to get ahead in these other tangential but very important forces that are at play in the market, right?

[00:19:39] So needless to say, AI has to become a force multiplier for these financial institutions. So on the regulatory side, as an example, AI can now digest hundreds of pages of new guidance, summarize what changed, and flag policies that may need updating, right? Take Genius Act as an example, right? Going through all of that and understanding what is it that I have to do as a community bank or at least need to know.

[00:20:09] Maybe in the past, it was maybe a consulting engagement. Maybe a team, right, going through this at quite a bit of detail, taking weeks on, right, to get to some of these actionable conclusions. Now it takes hours, right, if not minutes, right? So that's where AI comes into play. On the customer side, AI does help these smaller financial institutions deliver that personalization at scale, right?

[00:20:39] From faster loan decisions, right? We alluded to the types of drivers that are in the process of making a loan decision. How do you get to those decisions faster? How do you float those drivers up to the loan officers in a faster way? Or maybe even smarter fraud alerts so they can compete with digital first players.

[00:21:02] And on the internal operational efficiency side, again, AI helps quite a bit on the optimization front. Automating tasks from, you know, data ingestion, reconciliation, report generation, or maybe even document classification, right? That frees up a lot of time for the staff to focus on that relationship-driven work. Which, by the way, especially in the community banking segment, relationships are what drives the business.

[00:21:32] So if you can take a minute away from this grunt work and deploy that minute to continue to build a relationship with a small business that operates in your community to know them better, that's the marijuana state for these community institutions. So every minute that we can free up is gold.

[00:21:53] So in short, you know, AI helps these institutions, right, both on multiple fronts, regulatory front, customer front, and internally on the operational efficiencies, right? That increases their efficiency, decreases a lot of complexity that they have to take on in serving the customers. And we are recording this at a time where we're preparing to usher in 2026, starting to think about the future and what lies ahead.

[00:22:20] So on that side of things, when you think about the future of lending and credit risk, where do you see AI further supporting better outcomes for both institutions and indeed customers? And do you think human judgment will always remain essential because there is that fear in certain circles that we end up in that world where the computer says no and a human doesn't step in at all? So how do you see all this evolving and keeping that human in the loop?

[00:22:47] Yeah, I don't see, and we as a company don't see any time in the near future where humans go away from this process because this is so integral to how our communities thrive, right? So AI's biggest value in my view is that it accelerates the decisions, but you still need humans to prove those decisions.

[00:23:17] AI will help recognize different patterns, you know, maybe spot correlations across, various, you know, non-traditional data sets, cash flow trends, maybe even industry risk and how that's shifting, maybe early delinquency signals. So there's a lot of opportunity here to become more proactive and continuous.

[00:23:47] I mean, take portfolio monitoring as an example. Today, for the most part, it's a passive once in a, you know, maybe at some frequency, right? Somebody's going in and reviewing the entire portfolio of assets and loans, industries, et cetera. That could be a very active, ongoing thing once you have AI deployed in portfolio monitoring.

[00:24:11] So those are the types of use cases that we see where we will see a lot of adoption within lending and credit risk. And it's not just about math, right? End of the day, it's judgment. And we as humans, right, we understand that nuance where maybe a borrower who missed payments, right, is that because maybe there was a natural disaster but has a long record of good performance? How do you code that?

[00:24:41] How do you help AI see that, right? That's where I think there's still a long way to go to maybe even code human relationships. And that's why we think it's maybe a distant future. But there's certainly a lot of potential between now and then where we can help AI grasp the broader context and intent and start to help see the nuances but have humans in the play.

[00:25:04] So this is more of a partnership where AI provides the why now, maybe humans provides the why not yet, right? In other words. So it's really about driving that consistency and speed is where AI will help. And humans will provide the final arbitration, right, of bringing in that maybe empathy, maybe some exceptions, and bring to bear the trust, right, from the relationship that maybe is 15 years old.

[00:25:32] And if we continue on this future mindset and as we continue to look ahead into 2026 and beyond, are there any other developments in compliance automation or credit decisioning that feel genuinely transformative to you? And if there are, are there any particular ones that are still at the stage of hype that need to mature before banks can trust them? Again, what are you seeing here? I suspect there's a bit of a mixed bag. It certainly is a mixed bag, Neil.

[00:26:01] The truly transformative, rather, innovations are those that automate, in my mind, the interpretation and not just calculation. For example, take compliance automation, right? And it's about moving beyond the checklist.

[00:26:21] We now have AI that can read a regulatory update, compare it to the institution's existing policy, and highlight conflicts or maybe gaps, right? And that's truly transformative because it turns a highly manual, high-cost process into an intelligent, smart, early warning system, right? And the same thing's true in credit, right?

[00:26:49] AI-driven risk scoring, for example, that continuously learns from performance data is another breakthrough. So these are the sorts of things that help banks fine-tune underwriting without introducing new biases, right? Or maybe fast-tracking existing bias. So how do you make sure that you're balancing those two?

[00:27:11] So to the last part of the question that you asked, I think there's still a lot of hype around fully autonomous decisioning systems, especially in a highly regulated environment like ours, where large language models making those interpretations without any human review, right?

[00:27:35] These are promising, but they're not yet ready for high-stakes financial use cases because there's still a lot of hallucinations, right? If you are actively using AI in your day-to-day, right, you must have seen these hallucinations. You get very, very confident answers, right? That if you don't know the context and you don't know the industry, right, that's a disaster in the making. It really is. And we've covered so much in a short amount of time today.

[00:28:03] And anyone that wants to maybe dig a little bit deeper, find out more about you, the work that you're doing, how you're helping these community institutions, banks, et cetera, where would you like to point everyone listening to them? Yeah, you can absolutely find us at our website, abrigo.com, where we put out a lot of content, sharing insights and solutions around AI risk management and just broader innovation for community financial institutions.

[00:28:33] We also put out a lot of webinars around these topics, right, where we see a good number of financial institutions show up and hear our perspective. In fact, today we have a webinar on change management in deploying AI within financial institutions. That's happening this afternoon, right?

[00:28:57] And personally, I share my perspectives and practical lessons on LinkedIn, especially around how AI can drive trust and transformation in banking. And of course, we are always happy to engage directly with listeners who are curious about building AI the right way and just reach out to us. Well, I'll have links to everything, make it nice and easy for people to find you.

[00:29:22] And I, for one, I've just loved going beyond the hype with you today and taking a sneak peek behind the curtain and why banking may be the ultimate proving ground for responsible AI, how AI assistants are quietly reshaping financial workflows. And of course, the role of explainable AI in highly regulated environments. So much food for thought. I would invite everyone listening to get in touch with myself or indeed you and let us know your side of the story and what you're seeing.

[00:29:51] But more than anything, thank you for starting this conversation today, Ravi. Absolutely. Thank you for having me on Tech Talks Daily, Neil. Really enjoyed our conversation and appreciate the thoughtful questions. Wow. That was a fascinating deep dive into just how AI is already reshaping financial services, but in practical and responsible ways.

[00:30:16] And yet we discussed why banking has become one of the strongest proving grounds for trustworthy AI, but also how community institutions are adopting cautious yet effective approaches to innovation. And what explainability really looks like when every decision can impact our lives and even livelihoods. So if you'd like to learn more about Ravi's work and how Abrego supports community banks through this evolving landscape,

[00:30:46] you'll find the links in the show notes. And as always, I'd love to hear what you're seeing in your own organisation, in your industry. So feel free to reach out, share your perspective, head over to techtalksnetwork.com. You can leave an audio message there or send me a good old-fashioned written DM over on LinkedIn X Instagram, just at Neil C. Hughes. But that is it for today. So thank you to Ravi for joining me on the podcast all the way in Dallas, Texas.

[00:31:14] And an even bigger thank you to each and every one of you located in 165 different countries. A big hello to you all. I'd love to say it all in your native language, but that's a podcast episode on its own right there. So thank you for listening as always, and I'll speak with you all again tomorrow. Bye for now.