Breaking Down Barriers to Deploy GenAI Solutions Faster

Jim Marous (00:12):

Welcome to Banking Transformed, the top podcast in retail banking. I'm your host, Jim Marous. While banks have successfully implemented gen AI for internal processes and back-office operations, they still struggle to achieve meaningful ROI on customer-facing solutions like Agent Assist and Conversational IVR due to regulatory compliance requirements, data privacy concerns and model needs, this results in significant hurdles when deploying customer-facing application.

Jim Marous (00:44):

In this episode of Banking Transformed, we're joined by Manish Gupta, CEO of Corridor Platforms, and Toby Brown, Managing Director of Global Retail Banking Solutions at Google Cloud. They're going to be discussing their groundbreaking partnership and how AI and cloud technologies can transform banking decisioning capabilities, enabling institutions to compete effectively while maintaining regulatory compliance.

Jim Marous (01:11):

Our guests will also share insights on overcoming implementation challenges, managing gen AI risks, and creating sustainable frameworks for long-term success in digital banking. Financial institutions have significantly invested in gen AI capabilities with major banks committing billions to AI initiatives, however, deployment being held back due to governance concerns.

Jim Marous (01:38):

A new solution from Corridor Platforms and Google can enable financial institutions to advance from gen AI better verification to production with assurance. So, Manish, before we begin, can you share a bit about your team's backgrounds and what key market shifts led Corridor to focus on decisioning platforms?

Manish Gupta (02:00):

Well, thank you so much Jim for hosting us, and I'm super excited to be here with you and Toby, I think it'll be a very interesting and topical discussion. Let me tell you a little bit about Corridor Platforms. Corridor Platforms is a responsible AI governance solution, which Gartner has coined a new term called Decision Intelligence Platform.

Manish Gupta (02:21):

But what that really means is just think about it simply as a centralized shared workspace in a company where different teams in a company can truly collaborate easily and harness the part of all the data that company has but do it in a controlled and governed manner.

Manish Gupta (02:40):

Basically, it also automates a lot of the things that were done manually making the decision workflow more efficient, more auditable, and hence also very governable. The goal basically is to enable highly regulated entities like banks to truly take in-house the latest innovations in technology and analytics to improve internal efficiencies, but also to create amazing customer journeys.

Manish Gupta (03:09):

So, where did we learn this? The founders of Corridor basically learned their art at American Express, all of the founders are ex American Express. And myself, including my co-founder, Ash Gupta, was the chief risk officer of American Express for close to 30 years and truly created a culture of responsible innovation.

Manish Gupta (03:28):

We were usually first movers and so when you think about some of the biggest transformation of the last few decades, I think the last one we'll talk to about gen AI later was when big data came about in the late 90s. I'm sure Toby also recognizes that as a huge step.

Manish Gupta (03:48):

It kind of gave birth to traditional AI and machine learning being used to harness the part of all the data that banks had but it was quite a tough journey. So, when you start thinking about using that at a bank, even with all our resources it took us a lot of time and years.

Manish Gupta (04:05):

We had to do talent uplift, we had to think about a lot of automation, we had to really think about all the new risks and governance and risk standards that we had to develop to kind of put things into production, similar to I think what will happen with gen AI. We had to spend a lot of time with the regulators, OCC and the Fed to kind of get them along and truly work with them to kind of harness the power.

Manish Gupta (04:29):

And eventually obviously the regulators came out with what was SR 11-7, but that was five, six years into the journey to give guidance. So, the idea was, it was so difficult for us to do it, we felt like there would be a fairly large gap in the industry and need for somebody to kind of help banks in this transformation.

Manish Gupta (04:50):

And that's when we came up with what seemed to be a very brilliant idea then, and probably responsible for all my gray hair now, where we said, "Why don't we take a group of real, senior, experienced executives who know AI, banking, governance and regulation,” and then we said, “Let's team them up with really some of the best young engineering talent in the world so that they can come up with innovative, creative next gen solutions." And that is why the Corridor Platforms was born. So, that's the history of Corridor.

Jim Marous (05:24):

So, Manish, the reason for our podcast today is you have an announcement with a new partnership with Google. So, can you provide some background into that partnership and how the solution addresses some of the major challenges financial institutions are facing today?

Manish Gupta (05:42):

Sure, sure. So, just to continue that thought, we launched our traditional AI, I'll just call it traditional, to make a difference between gen AI and AI about five years ago into the market and we have banks and credit unions who use it and we are lucky enough to have some top tier banks.

Manish Gupta (06:01):

So, when one of the top tier banks was started experimenting with gen AI, and we also got involved in thinking through how do you really govern gen AI? We felt that the user experience and the nuances of gen AI pipeline building were different enough that we had to develop a new solution still based on the same backbone of governance that was in the platform, but a completely new user experience.

Manish Gupta (06:27):

So, last year, the first six months we worked on creating a new product, which is Gen-GardX, which is the main product that we are talking about in partnering with Google, and we'll talk about that. So, six months we built that, and then because we were lucky enough to be live at a client, we said, "Why don't we launch it as a beta so that we can actually test it in a live use case with people's working and see if it's solving the issues."

Manish Gupta (06:52):

And Google also, luckily enough, was at the same client and helping them through that journey, so we really worked together. Us trying to really figure out what are the features, how do we fine tune them? We work together, for example, on things like, "Hey, how do you collaborate on prompts?"

Manish Gupta (07:10):

Because LLM pipelines have now new things like what's ground truth? What's the LLM you're using? What are the instructions which are prompts? What are rags and knowledge bases and guardrails? How do you stitch them together?

Manish Gupta (07:22):

So, we worked on adding features that we have not developed before, like how do you really do hill climbing with prompts? How do you do version management so that it can be really controlled because changing one line could change an effect so how do you control that?

Manish Gupta (07:36):

We worked on a completely new module, which is human in the loop monitoring. Once you have developed a use case, how do you let people test it in an analytical environment to break it, jail break it or make it do some things that it wasn't supposed to do. All very, very critical as we are thinking about how do you manage pipelines, develop them, manage their risk and take them into production.

Manish Gupta (08:00):

One of the other things that we spent a huge amount of time with on was governance, model risk management and fair lending, because there are no set standards for it. So, we collaborated with another partner of ours, an investor in Corridor Platforms, Oliver Wyman.

Manish Gupta (08:16):

And we got a bunch of gen AI governance experts together with us and thought through what are the initial tests, metrics, reports that would satisfy MRM and fair lending needs, both internally and with the regulators eventually and so, we included that as part of the foundational package so that now banks can have a head start.

Manish Gupta (08:37):

So, this is the exercise that we went through, and because of that, some of Toby's teams, which are really experts in LLMs and making them more accurate and efficient, worked with us on our platform, as I said, and it became clear that this was probably the same situation at many banks where if we could partner together, we can really help the adoption of gen AI.

Manish Gupta (09:02):

Maybe Toby can give us … therefore, when Toby kind of made the call and asked me whether we want to be a partner, obviously Google's calling a small startup, so we were thrilled, but I do believe that this is a fantastic solution that can help a lot of clients. Toby, maybe you wanted to add your thinking or your perspective.

Toby Brown (09:21):

Yeah, no, thanks Manish, and thank you Jim, and it's great to be here. So, I have the privilege of leading Global Banking Solutions at Google Cloud, and what we really do is think about how do you use all the incredible technology we have at Google Cloud and how do you apply that to banking specific challenges and use cases?

Toby Brown (09:40):

I came from banking myself, just like Manish, I worked for over 20 years at various tier one banks most recently at Wells Fargo and I lived and breathed a lot of these challenges as well. So, I love bringing that lens to work every day and pairing that with my deep understanding of the Google tech stack to talk about how we can help our customers, all these banks around the world, essentially use the technology to accelerate and amplify their own transformation journeys that they're going through.

Toby Brown (10:10):

Of course, increasingly this is an AI first conversation and not really generative AI first conversation and I think we've really seen the industry rapidly evolve in its thinking as it's gone from understanding the technology to figuring out use case ideation and getting a kind of quick list of targeted areas they want to pursue and look at to then going through this whole end-to-end process of actually operationalizing the technology.

Toby Brown (10:40):

And that's really what we started to see some real road bumps, that's where the rub is, particularly in a regulated industry like banking, because you have so much that you have to go through, all these traps you have to run in order to do that in a way that fits within your risk management governments, in order that complies with all the three lines of defense and everything that they do at a bank, and also that's going to pass muster with regulators and so there's so much to that. And we know that that work traditionally has been very manual and it’s not scaled very well, and it’s a very common frustration point internally at all banks.

Toby Brown (11:18):

And so, what's really amazing about this partnership with Corridor is that they are a massive accelerator for our customers in getting through all that and really helping them now create a flywheel, essentially.

Toby Brown (11:30):

This kind of virtual cycle where they can much more quickly go from the point of having a generative AI use case that they want to put in production working through everything that they need to align internally in order to deliver it and then using a workbench and a platform like Corridor and their Gen-Gard AI specifically to then accelerate the process and to really shorten the cycle time for being able to get that into production.

Toby Brown (11:57):

So, that to me is tremendously exciting, it's because everyone is of course, racing to figure out how you can turn, create this kind of factory model for generative AI so that you can get as much in production as quickly as possible and really maximize your business value. Of course, all the banks are looking around and they know that everyone else is doing this as well and they know that a lot of this is going to come down to execution.

Toby Brown (12:22):

Who can create the best flywheel around their data, who has the right highly performed infrastructure with all the security built in that you need and who will be able to line up all those internal enablers to be successful and to win in their business. So, it's a really exciting time to work with banks and it's a really exciting time for us to work with Corridor on partnering with those banks to help them transform.

Jim Marous (12:49):

Toby, you've been in banking for a while as you mentioned, both at Google and Wells and other organizations. And when you look at this from a gen AI perspective, there seems to be a lot more talk than actual walk in many cases, but what patterns are you seeing in how banks are approaching gen AI adoption today, and what are some of the common use cases you're seeing actually being deployed?

Toby Brown (13:19):

That's a good question. I think we've seen that evolve really quickly. So, I think that there was, I don't know what you could call it, skepticism, I guess initially on kind of the product market fit side of it. How much business value we're actually going to be able to generate using this technology within a regulated industry like banking, but I think we've seen that move really quickly.

Toby Brown (13:40):

And we actually just did a survey towards the end of 2024, where we surveyed a couple thousand different financial services institutions about the ROI they were seeing on gen AI and the results that came back actually surprised me.

Toby Brown (13:55):

They really positively surprised me on the upside because what we saw was that for those who have moved gen AI use cases into production that 90% of them were experiencing revenue growth, that they 80 to 90% were seeing huge efficiency gains, but long story short, they were seeing real business value. And I think if you had done that survey, another year prior to that, you absolutely would not have seen that.

Toby Brown (14:23):

So, we're seeing just the pace of change and the pace of the industry's ability to adopt and operationalize it really improve which is obviously phenomenal for the entire industry. I think when we look at the use case landscape and we look at how banks are actually using the technology, we really see three kind of big pillars.

Toby Brown (14:44):

So, the first is using it of course to — well, let me back up one second there. I'll say that the other kind of frame we see on top of that is that most banks are using a two-pronged approach for going through and implementing generative AI use cases.

Toby Brown (14:58):

The first stuff is really the low hanging fruit, it's all the optimization work that you're doing to improve productivity of all these different types of employees and it's the things that often always have a human in the loop.

Toby Brown (15:12):

It's where you can get the most business value with the least amount of effort and allows you to kind of prove out and iterate with the technology. And then I think in parallel to that, we're seeing the more transformative use cases, and we're seeing those increasingly start to come online.

Toby Brown (15:26):

So, where you're creating new experiences and you're also just changing the way that you go to market as a business or as a bank rather and I think that's obviously really exciting to see too. But back to the point around the use cases specifically we see three big areas.

Toby Brown (15:46):

So, one is using it, generative AI to help grow, to create new organic revenue streams, which is of course, a huge challenge in banking, particularly post 2008, most tier one banks, most banks of any size have struggled with creating sustainable organic revenue growth. There's some macroeconomic tailwinds and things that that have helped in recent past but overall, to generalize, that's been a big challenge.

Toby Brown (16:10):

A lot of that comes down to data. So, looking at how you can really harness the strategic value of your data to both acquire new customers and to deepen your relationship with the existing ones. We see banks now using gendered AI to do a way better job of targeting potential customers.

Toby Brown (16:32):

So, using generative AI to do a better job of selecting segments in a novel way and in a faster way so that you can get signal quickly and go out to your customers and have that data-driven personalized interaction that everyone's trying to achieve. So, there's obviously a marketing use case there that's kind of top of mind in default in terms of being able to generate new campaigns quickly, new offers quickly that are truly targeted.

Toby Brown (17:00):

And then there's also like a next best action, next best conversation angle to that where we see a lot of customers who are completely replatforming their contact center tech stacks as an example and they're creating these, what I'll call customer service hubs instead of the classic old school kind of phone bank where was built back in the day for taking calls from landlines.

Toby Brown (17:26):

And now you've built a new customer service hub that's taking in all these different interactions across all these different channels. We see them using generative AI to maintain context as interactions switch between channels and then also to improve the quality of those conversations. So, usually the only stuff that's making it to a human now in a contact center is more complex interaction types.

Toby Brown (17:50):

Most of the simple interactions have been migrated over to an app or a digital experience, and so that requires a different skillset and different training, and it's harder and we're seeing banks really successfully use generative AI to improve their performance and to improve the quality of that customer experience.

Toby Brown (18:09):

So, when a customer calls in and says some obscure question, “I'm traveling to this airport lounge and I have your credit card,” or sorry, “I'm traveling to this airport and I have your credit card, which lounge do I have access to?” It's not something an agent knows off the top of their head.

Toby Brown (18:25):

So, instead of having to put the customer on hold or hold tab between different applications and find that they're now getting prompted with that correct information in near real time, and then they're using that to then broaden the conversation and focus on what else they might be able to do to help the customer at that specific point in time.

Toby Brown (18:41):

So, maybe it's, I see you're traveling, so I like to talk to you about travel insurance, so I see we have this credit card, which is a better credit card, so it turns from a servicing conversation into a sales conversation, which offers huge opportunity both for the customer and of course for the bank as well from a business value perspective.

Toby Brown (18:57):

So, the last pillar is really around risk. So, we know that fundamentally of course, banking's a risk management business, we know that all the risk organizations have grown massively since 2008, and we know that a lot of those business processes are extremely manual, and a lot of the control environments are extremely manual.

Toby Brown (19:20):

And we now see banks using the technology, generative AI to improve the way that they predict risk and to help automate a lot of those processes.

Toby Brown (19:28):

So, everything from environment scanning, so looking at how the regulatory environment is changing and shifting to then understanding the implications of that on your policies and procedures and how you implement and stay compliant with those as well as sometimes even going all the way down into your code base to make modifications to stay compliant.

Toby Brown (19:48):

So, a new risk-weighed asset calculation changes, you then bump it up against your internal policy bank, sometimes that goes all the way down to your code base. You can see banks starting to pull the thread there across all of that to figure out how they can just really accelerate and improve their cycle times for, and in turn improve the way they manage risk as a regulated bank. So, those are some big ones that we see a lot of traction on and a lot of excitement around from the industry.

Jim Marous (20:15):

So, it's interesting, Manish, we had a pre-call and we were talking about the fact that there's so much opportunity in the generative AI area and the AI area in general, and you look at it, but at the same time, we see financial institutions having barriers, having challenges, having hurdles as you were, and what does that mean? What are some of the major challenges that financial institutions of all sizes are facing from the standpoint of deployment of these solutions?

Manish Gupta (20:49):

I think I'll just build on what Toby said. One is gen AI is truly transformation. I'm convinced the more and more I get into it that it will change the landscape of banking in the near future, and we can talk about that a little later. One beautiful thing it does, it kind of democratizes decision making and modeling in some sense.

Manish Gupta (21:12):

Because we speak about larger banks, when you think about the next tier of banks who have struggled to keep up with the same talent or same size of group, when you come to gen AI, the actual model in Toby's team or Google and other people are creating phenomenal progress in really providing models that can provide answers and obviously you have to now learn a new skillset, but everybody's at the same starting point.

Manish Gupta (21:43):

So, therefore it is an interesting time that hey, now you have to figure out how to instruct the model, which is an art by the way, I'm learning. I thought it was going to be simple, but it is an art form, which is a complete learning skill. That how do you provide the right information and the guardrails?

Manish Gupta (21:58):

So, I think the main challenge is in learning something completely new for both big and small banks and it comes with the same issues of how do you think of the talent, of shift in talent to do this now? How do you then think of, like Toby said, with this kind of newness and new ways of even building and testing where you're giving inputs and outputs to an answer, to a model, I think of it as it's easier for me to think of it as like a teenage mind, which is an LLM model.

Manish Gupta (22:33):

And then you’re providing instructions to that mind and giving it information about your local scenario, local bank problem or use case and you are getting very, very smart answers. It solves a critical need.

Manish Gupta (22:45):

When I was in banking, I used to go to the telephone centers, and it's surprising the amount of churn that you have at agents in a bank because it's not a high paid salary. So, kind of to manage interactions with high quality, with high consistency is very, very difficult.

Manish Gupta (23:04):

If you could train this very, very smart teenage mind to criterialy give proper answers all the time, it would be a phenomenal improvement in customer service, revenue, efficiency, and obviously in some sense cost efficiency too.

Manish Gupta (23:19):

But now you have to think through how do you manage those risks? Because you can train the teenage mind but it is truly a teenager that does not always do what you tell it to do and that's the interesting generative part.

Manish Gupta (23:34):

And that's something that banks, big and small, will have to think about. How do you take your risk tolerance? How do you manage guardrails and come up with interesting ways of applying and harnessing the power but within the risk tolerance, which is banking.

Manish Gupta (23:51):

So, to me, there are this kind of just going through the journey, governance is going to be a huge issue because traditionally, banking has been a very low tolerance kind of industry, regulators will have to pay their part, the hype and the fear of gen AI both are a boon and a curse because regulators are being very conservative right now because they want to make sure that something doesn't blow up.

Manish Gupta (24:19):

So, I think everybody will just have to come along and then obviously there'll be a completely new era in my mind. And it'll be interesting to see because I do think the players will change depending on who adapts small or big.

Manish Gupta (24:32):

One other very important point that comes to my mind, Jim, which is as you think about traditional banking where credit unions and regional banks or local banks all had a captive member base and you were manually going there applying.

Manish Gupta (24:51):

There, there was this concept of, "Hey, let the large guys who have more resources be first movers, and we will be fast followers," though nothing is that fast in banking.

Jim Marous (25:02):

That's right. Fast followers is a-

Manish Gupta (25:03):

Yeah, it's a trade.

Jim Marous (25:05):

It's dichotomy, yes.

Manish Gupta (25:06):

But it was okay because you didn't really lose your customers, you just figured out how to give them better treatment a little later. Now with the digital age, that's all gone because you are competing for the same customer against the larger banks, the digital banks who are now harnessing the power of all of the data coming up with fantastic experiences and clients, and your customers can move in a jiffy, we saw that with Silicon Valley Bank and First Republic.

Manish Gupta (25:35):

So, now you can easily, if you are not quick to kind of adopt and adapt to the same level of sophistication, giving the right price, right offer, your customers are gone and you are basically left with the people who nobody wants, which is a huge existential issue. I mean, it's going to be a huge issue in this.

Manish Gupta (25:53):

So, therefore, how to figure out how a usually slow-moving industry now adapts to a speed of change where you can't really be late or to the game. That's where I think fintechs like ours are going to play a critical role.

Manish Gupta (26:09):

The whole idea is how do you kind of package everything, automate everything, make it efficient, but also give a knowledge transfer so that banks can really be fast in adapting transformation. Anyway, so those are some of the things that I see are really needed to really get the full ROI from gen AI.

Jim Marous (26:28):

Well, it's interesting, Manish because on top of all this, you have government and regulation and playing a game of catch up. So, what happens is, no matter how fast you want to run, the reality is regulation in some ways is slower to adapt because they're also trying to set the guardrails for all this.

Jim Marous (26:48):

And from your perspective Manish and I'll ask Toby the exact same question, how are you seeing government regulations and the concepts of keeping this all manageable and risk averse? How do you see in government actually responding to the potential of gen AI and I guess on the same sense, are they playing catch up and how do the small banks not use that as an excuse and the big banks not use it as a hindrance?

Manish Gupta (27:17):

That's an interesting question. To tell you the truth, I don't think anything will happen differently here than what we saw and went through with the AI or the big data revolution when AI and machine learning came along.

Manish Gupta (27:32):

Regulators, at least in the U.S. and probably globally in the Fed and the OCC, have some of the sharpest minds. But basically, the rule book that they follow, which a playbook that they follow, which I have gone through many years as leading risk in organizations is probably the right one.

Manish Gupta (27:52):

If you want to adopt a new technology, they asked the big banks and the more sophisticated banks who are first to really think through the risks themselves because it's you who want to do it and think through what is it that you need to do to manage the risk.

Manish Gupta (28:07):

And then when you are discussing things with them and they have a few proof points and their role is to enable that a little bit so that they don't really create a stance where everybody's scared to do this and I think there's a role they play there.

Manish Gupta (28:22):

And then once they have a few proof points, they do a very good job of taking everything that's been done and summarizing it into a white paper and then creating hopefully a guidance, which a lot of banks can follow because they don't have the internal power and clear coming up with the regulatory mandate or MRM requirements like they did. And I think this is going to go through that same process and therefore it's just a question of working with that dynamic.

Manish Gupta (28:48):

I think when it comes to big and small banks, I think it's going to be okay because smaller banks in general are not going to do the kinds of things like building your own LLM or really fine tuning an LLM or really doing something super fancy in developing pipelines with rags, et cetera though there's a requirement for it.

Manish Gupta (29:12):

And I think Toby can add to this, there's a whole field of agent tech AI thought processes where solutions already being billed by the sales forces of the world, et cetera to really give a plug and play for smaller banks to do it.

Manish Gupta (29:27):

But the problem we see with the smaller banks is the same. They're starting to use some of these plug and play facilities but they still have the same risks of a teenage mind and a gen AI. So, they still have to get comfortable that before I use this agent that, "Hey, can it be jailbroken? Can it say something because I'm responsible for that agent?"

Manish Gupta (29:47):

So, they'll have to also think through governance and testing, and there's no excuse for that. Anyway, I mean I think I may have lost the plot somewhere there, but I think those are the kinds of things that truly make this a very interesting journey.

Jim Marous (30:01):

Toby, what do you see from your perspective as far as, because it goes beyond financial institutions in your organization with regard to governance and regulation, but the reality is in healthcare and in financial services, it's amped up because there's so much data and there's so much privacy and individualization there that the government, no matter how much it changes politically, continues to be on top of this. Because it is front and center and it only takes a single bad character to throw this whole dynamic into a tailspin. So, what are you seeing out there?

Toby Brown (30:43):

I think that we're seeing regulators approach it very cautiously and very thoughtfully. I think it's tough because the pace of the technology is also changing so quickly. So, how prescriptive and specific can the regulation really be when the technology is evolving at the pace that it's evolving at? So, where to me that likely lands is something more principles based.

Toby Brown (31:09):

You also have the whole political piece of it and the change that's happening right now and the impact that's having, like on the U.S.' approach where we had the beginnings of kind of a AI governance framework but that's now being completely rethought.

Toby Brown (31:26):

So, my point is that you have change on a lot of fronts. Both the technologies evolving, the political environments evolving and so I think being able to actually land on some really fine grain, specificity around how AI will be governed in a regulated industry like healthcare or banking is going to be extremely difficult if not impossible to do.

Toby Brown (31:46):

And so, I think what we see today, where banks are successfully getting through all the lines of defense and the regulators to actually get the technology to production is being very use case specific.

Toby Brown (31:58):

Leveraging the known worn path around things like model risk management and talking in a very use case specific way about what they're going to do and why this is their view of the risk of the kind of foundational components they're using, these are the controls they're applying to that and this is why the residual risk on the backend, it fits within their risk management framework and their risk appetite and all those things.

Toby Brown (32:24):

And I think that there's such a robust approach already in place there for the industry like the model risk management is something where every bank has a 2 to 300-page pack for every model that goes into production. It's something that's been vetted seven ways from Sunday with everyone, including the regulators.

Toby Brown (32:43):

And so, I think the rails there aren't broken and don't need to be transformed. I think there are tweaks on the margin and things that need to be done a little differently in light of some of the uniqueness of generative AI. So, I think that it's going to be interesting to see how granular versus kind of principles based the regulatory guidance is going to be. And then it’s going to be left to banks to interpret that and to essentially evidence that what they’re doing adheres to those principles. It’s kind of like a BCBS 239 kind of thing to me.

Toby Brown (33:17):

So, we know that those principles are going to be there at some point, and we know that each bank is going to have to demonstrate that they can live up to those principles. And of course, all that robust model risk management work is going to have to be in place before any of these use cases go into production.

Toby Brown (33:33):

And that's somewhere really, back to our earlier conversation, gets back to your speed and your cycle times for doing this kind of work and it's going be a huge differentiator for those who are able to really cut down their cycle time.

Toby Brown (33:48):

Manish made a point earlier too, around the model versioning and model penny, and that's a huge point for staying on top of this technology as it continues to evolve. Because the typical process internally for a bank is at least 9 to 12 months, something like that, typically for a new version of a model, a model changes-

Jim Marous (34:09):

Not going to work.

Toby Brown (34:11):

All the way back through the whole process to then have it in production. But now you're having new models come out, at Google, it's at least it's monthly, quarterly, multiple times a year. And so, you've got to evolve your processes to keep pace with how that technology is changing.

Toby Brown (34:29):

And the ability for banks to do that is going to be a huge driver of competitive advantage, I think, in our industry. So, it just gets back to the power of using platforms like Google Cloud and tooling like Corridor AI to help you accelerate and create that flywheel as best you can.

Manish Gupta (34:55):

Jim, let me just add and maybe just nod the eye there to Toby's point because when we're working with clients' life, this is a real-life problem. I want to make sure as, let's say a model version improves, the benefits of it are immense, both in terms of the solution and the cost.

Manish Gupta (35:17):

Because you can now do the same thing in a much smaller model, which is much better and faster and the cost can be twofold, threefold lesser, so you can't really wait for that change. So, automating governance, which is the whole theme around how do you take it, standardize it and automate it so that you can do those reviews that used to take 6, 7, 8 months into a weekly basis requires a very manual process to be fully put on into a systematic platform.

Manish Gupta (35:51):

There is no two way, that was the basic need that we saw in the market when we launched and with AI, but with gen AI, it's now almost impossible to do anything without it being automated. And therefore, I feel that is something, whether they use Corridor or develop it themselves, et cetera, it'll have to be a problem that they solve to be competitive.

Jim Marous (36:12):

Well, it's interesting, we take talent and the personnel that it takes to run all these things for granted, but the reality is we see the highest levels of innovation no matter what we're talking about in today's marketplace happening, where partnerships take place, where the people that you're partnering with have a vested interest in the final result therefore, a financial institution may only have to, I'll say tweak, it's much more than that, but tweak 20% of what's being developed as opposed to the entire 100%, that doesn’t matter what size organization you have.

Jim Marous (36:46):

I mean, yes, there's hundreds if not thousands of people at Chase working on these things daily but it's easier to work on them when there's a partnership that takes an even bigger organization such as Google with a very specialized organization such as Corridor brings these two together and say, "We'll still make it ours."

Jim Marous (37:05):

But it's these organizations partnership that makes it so it's not only looking over their shoulder at what's going on with governance and things that are changing every day on top of everything else but are also deploying it at speed and scale.

Jim Marous (37:19):

On the same hand though, on the alternative hand, you have the financial institution's legacy mindset that is risk adverse, that is slow to react, that it was used to still at many organizations, annual planning processes that only evolve on an annualized basis as opposed to weekly, monthly, in some cases daily.

Jim Marous (37:40):

How much of a challenge is it or number one, how important is it to have these kind of partnerships where you have people that are focused entirely on this solution set? And then at the same time, how hard is it for your organizations to work with financial institutions that just in a legacy mindset way say, "Yeah, we understand that you can come to us faster with all this, but it really works against our overall culture." How do you do that Manish? And I'll ask Toby the same thing, because it's something we're up against all the time, especially in financial services.

Manish Gupta (38:17):

I think you are hitting the nail on the head and I think Toby also referred to this, the speed of change is at a totally different level right now. I think in the next year we'll probably get more change than has happened in the next last 10 years and that's real.

Manish Gupta (38:37):

I think some people would say it'll be more than that. The whole concept of Moore's law is completely different in this concept. Every two, three months we see huge improvements in both the solution and tool kits that companies like Google are providing.

Manish Gupta (38:55):

And so, I think it'll be very, very important for banks to really think about how do they change their nimbleness. This is when I was saying, I think the landscape is going to change because whether you are big or small, I think banks move at a pace which is not at the same of this change and I don't think there is a way to get around it.

Manish Gupta (39:24):

So, if you are slow in automating or adapting to the newness or adapting to the constant improvement and somebody else can do it faster than you, big or small, they're going to create better customer journeys and better efficiencies and better P&L metrics that they will soon start taking over.

Manish Gupta (39:43):

If you remember, Jim, I know you will remember this, if you look at, I'm just giving you a tangential example, the number one to five banks who used to issue mortgages 10 years ago and if you look at that list now, there's not even a single name, which is the same, rocket mortgage is number one and everything is different. So, I do believe this is going to be one of those transitional points where five years from now, I don't even say 10 years, because so much is going to happen in five years.

Jim Marous (40:13):

I even think five years may be long, too long too, I agree. Well, it's interesting because Manish, I’m used to saying change has never happened this fast, but it'll never happen this solely again. So, you had mentioned the ability to be a fast follower, you can't because you got to still run faster than the trains moving to be a follower which makes it very dynamic.

Jim Marous (40:35):

And as I mentioned, it also makes it so that partnerships of this nature and all what we're seeing in the marketplace today, you find the specialists in certain areas and say, "Well, Manish, your partnership with Google is saying we can't do it all ourselves. We would really like to have a partnership with Google."

Jim Marous (40:51):

At the same time that Google is saying in the financial service division, "It's good to find that specialist who's dealing with this specific problem and then the financial industry doing the same thing."

Jim Marous (41:01):

This is what excites me more than anything in the industry right now because it really takes scale of the financial institution a little bit out of the equation. Because any organization of any size can deploy it at the speed and the perfection that's the biggest ones can do and, in some cases, faster, because they don't have legacy leadership that can hold it back.

Jim Marous (41:24):

I mean, we see it all the time in the discussion we have on this podcast that some of the most astounding growth and innovations are coming out of some of the smallest organizations because they don't have all the baggage. And if they have the baggage, it only takes the leader to say, "I'm going to get rid of this baggage. We're going to do something different. I'm not going to rest on my laurels."

Jim Marous (41:46):

So, Toby, from your perspective, how does this make it exciting from a scalability standpoint, from a deployment standpoint to have these types of partnerships that really move something that is holding the industry back on a deployment basis to the top of the ability to change?

Toby Brown (42:06):

I think it's fascinating because in banking, there's always this tension internally between prudently managing your risk and growing revenue and acquiring, expanding as much as you can, so there's tension between that appropriately, so. And it's not like it's a move fast and break things culture, that doesn't go well, if you do that in banking, it's a move as fast as you can without breaking anything.

Jim Marous (42:37):

You can't go SpaceX on this.

Toby Brown (42:40):

No, no, you can't and that's a good thing for the whole financial ecosystem that it's like that. And there is, we do see new ways of thinking even at large-tier one incumbent banks, we see CIOs now being named heads of retail banking, you see people upending the way that they are structuring their operating models relating to deploying technology like AI.

Toby Brown (43:08):

And so, I do think that there is real willingness and engagement from leadership at banks all over the world on genuinely wanting to rethink the model and how they can evolve it in light of this new technology and the power of it and just how the world is moving.

Toby Brown (43:26):

But there is that third rail that's there and will continue to be there that you can't touch. That you have to still ensure that you are doing it safely and responsibly. From our perspective at Google, that’s priority number one.

Toby Brown (43:42):

We're all about responsible AI, we're all about an enterprise ready platform that allows you to deploy this technology in a way that is adequately controlled and that is safe for deployment and that is secure.

Toby Brown (43:56):

And then I think as you move increasingly up the stack, these partnerships become critical to, like I said earlier, to accelerate your ability to operationalize the technology within the confines of your institution.

Toby Brown (44:12):

So, within the way that you manage, model risk management within all the different processes and things you have to do to stay compliant using a partnership like this, like using the power of something like Corridor to help you automate and accelerate that work is like I said, just a huge driver of competitive advantage now in the industry. So, I think, to simply answer your question, to me it's going to be utterly critical.

Jim Marous (44:38):

It's interesting too because we look at these kind of partnerships and I'm loving it because I'm a legacy banker from 45 years ago and the reality is to find partnerships like this that are so specialized but so dynamic in that your existence depends on these type of partnerships hitting the ground running very quickly and evolving quickly faster than your partners can do. Not your partners within the two of you but your partners from the financial institution standpoint.

Jim Marous (45:12):

And then at the end of the day, it depends on the leadership of the financial institution to actually deploy it the way it can be deployed within the guardrails that are set by the government. So, Manish, to close out this podcast and this introduction of the exciting partnership that you have here, what excites you on the horizon?

Jim Marous (45:30):

I can't go five years anymore, so in the next year to two years, what excites you right now? Because it seems like every moment, I open up my feed and I see something else new that comes out, you go, “Boy, I didn't quite see words converted to video in a really, really new dynamic way.”

Jim Marous (45:49):

What excites you about what's going on in the marketplace right now and what your organization and Google and as well as other partnerships you see can bring to the marketplace for the financial services industry?

Manish Gupta (46:02):

I mean I think we've kind of gone on the same vein already. What really excites me is the speed of innovation, which is creating the possibility of practical and implementable solutions and I'll give you an example of that so that it's not kind of vague.

Manish Gupta (46:24):

When LLMs came, and this is not long ago, about a year ago, I was conceptually after all my experience was thinking, "Oh my God, in the end, banks will have to think about how to really build their own LLM because they want full control of it, or how do they really fine tune it because it needs to be about their particular industry and use case and personalize it."

Manish Gupta (46:49):

But the work that the industry has done in making these LLMs smarter and smarter so that they can incorporate a lot of the knowledge already in them and only a small amount has to be taught now has already changed my view to, I don't think most banks will ever build an LLM or need to.

Manish Gupta (47:10):

But they then need to really understand how to guide this LLM and therefore the speed is phenomenal. Now, I'm sure they'll build specific LLMs for specific industries, and so the heavy lifting is evolving fast, which would've been a challenge.

Manish Gupta (47:28):

Now, then it comes the most exciting moment that if it's becoming so much easier to do these use cases, at least the raw material is there, then I think it's exciting for a company like me, because like you said, in our focused area, which we want to remain focused, we have to be really running fast.

Manish Gupta (47:49):

But because we are just doing one thing, how do you take the governance of all of these developments and work on them, use all your experience to figure that out and how do we automate that? We will definitely try to keep fully up to speed on it.

Manish Gupta (48:03):

And therefore, like you said, if you want a financial institution to keep up to speed, you need to partner with the Googles of the world who will give you better and faster and cheaper models plus tools, which hopefully we'll also integrate but we'll really focus on governance and then you focus as a financial institution on truly what's important.

Manish Gupta (48:24):

How do you make it and choose the improvements that you want to do for your internal efficiencies and I think the most exciting part is going to be customer journeys, which are obviously higher risk. That is where the money is. Money as in a broader sense and that's super exciting to me.

Jim Marous (48:44):

There's no doubt about that at all. And I think the thing that you're trying to address, which is one of the challenges and barriers that financial institutions have, which is the comfortability with governance, with regulation, things like this, you're trying to take that off the table so that they can deploy faster, which is the end of the game, is something that I get frustrated with daily. What’s the opportunity out in the marketplace versus what actually gets deployed?

Jim Marous (49:09):

I've said it for I think the five years of our podcast. I get frustrated that my current two financial institutions know everything about me but they never show me they know what's going on about me.

Jim Marous (49:20):

I think that's the same thing with Google, with your Corridor Platforms, what's possible, we're not getting near what's possible, all we're trying to do is get rid of all the hurdles that people put away saying, "Well, I'd like that but," to get rid of that but and it keeps us all busy.

Jim Marous (49:36):

I mean, there's not a day, I'm sure, Toby, that you and Manish don't go to work going, "I have no problem with what I have to do for the day." It's a matter of, is anybody going to listen and actually then deploy. The talk at every organization's trade show, whatever it is, it's all about generative AI.

Jim Marous (49:51):

I'd like to see more from the standpoint of what I can actually feel beyond the efficiency side of it, the back-office sides, which banks can embrace that forever because any way I can save money is good, it's those non-defined goals from a consumer basis that you only find out later how well they've done.

Jim Marous (50:12):

So, gentlemen, we can take another podcast on this but I'm excited for your partnership, I'm excited to see what comes out of this and what happens in the marketplace. I think financial institutions benefit greatly from these type of arrangements that take less money than ever before to deploy the best technology you've ever before with partnerships that are pretty exciting. So, thank you both for being on the show today.

Manish Gupta (50:37):

Thank you so much. Hey, thank you.

Toby Brown (50:38):

Thank you, Jim, great to be here.

Manish Gupta (50:39):

Same, man.

Toby Brown (50:40):

Great to see you, Manish.

Manish Gupta (50:41):

Thanks.

[Music Playing]

Jim Marous (50:43):

Thanks for listening to Banking Transformed, the winner of three international awards for podcast excellence. If you enjoy our work, please give us a positive review. In addition, check out our recent articles on The Financial Brand and the fantastic research we're doing for the Digital Banking Report.

Jim Marous (50:58):

This has been a production of Evergreen Podcasts. A special thank you to our senior producer, Leah Haslage, audio engineer, Chris Fafalios and video producer, Will Pritts. If you haven't done so already, please subscribe to Banking Transformed on your favorite podcast app and on YouTube for more thought-provoking discussions on the intersection of finance, technology and leadership.