AI Adoption and Governance in Banking – Techstrong AI Podcast EP60

Hello, and welcome to the Techstrong AI Podcast. I'm Amanda Ani, and with me today I have Samir Gupta. He is the Financial Services AI leader at ey.

How are you doing today? I'm great, Amanda. Thanks for having me.

Happy to have you on the show. Can you share a little bit about ey, what services do you provide? Yes, yes.

Happy to. So EY is a large global fi, uh, services firm. Uh, we cover all sectors in over 180 markets.

Uh, specifically I focus on our financial services sector. Uh, we provide, uh, professional services among our many service lines. EY also does work in the audit and tax and transaction transaction space.

Uh, the part that I belong to is our consulting services business, where we work with FY and skin, everything from strategy to adoption of tech, of new technology, uh, refinement of their business processes and the like. Wonderful. Thank you for sharing.

So, our topic for today is AI governance models and AI adoption in banking. So, as banks adopt ai, they must ensure their production environments are equipped to handle sensitive data and foundational assets like large language models, and they need to have a good cohesive strategy. So from your experience, uh, what mistakes do business leaders or companies make, and what advice do you have?

Yeah, yeah. So just, just taking a step back, uh, AI is one of the most, uh, uh, or the fastest, uh, moving technology right in, in recent times and financial services institutions as well as the broader corporate environment is just trying to keep, keep, keep pace with that. So in that context, uh, some of the, uh, trends that we have seen over the last few years have been quite consistent across the financial services business.

One of the trends has been to, uh, treat this like a shiny object, right? And to jump, jump onto the AI bandwagon, uh, start a whole bunch of proofs of concept without necessarily thinking about the business value. Um, so one of the mistakes I would say that, that we've seen financial institutions do is to start, uh, and try too many things at the same time without really, uh, thinking through the business connectivity.

The second part is, uh, in terms of thinking about this narrowly as a technology challenge, uh, and thinking of it as something where if they could just find the out, uh, the, the right mousetrap and everything will, will, will, you know, just solve for itself. Whereas, uh, what they're starting to realize now is that AI is just one piece of the puzzle, and they have to solve for many other things in terms of process adoption, in terms of governance, in terms of, uh, changes to the organization structure, in terms of, uh, people and training. Um, and, and I think that, uh, just, uh, coming to terms with that is, uh, has been a journey.

Uh, lastly, I would say there is, uh, the, their, the, the existing processes in terms of how organizations manage their, their ai, uh, and related risk has not kept up, uh, not kept pace with how quickly the technology has changed. So one of the big, big mistakes has been to rush into the tech development without thinking about some of the governance related challenges, And especially in banking. That's definitely, um, a, a big concern.

So what advice do you have when it comes to integrating this AI technology and making sure, um, that there's no vulnerabilities or, um, security issues? Yeah, yeah, that's a great question. So, so one of the things, uh, related to the, the governance and, and security, right, is, is the policy in itself.

A lot of banks have gone through the journey of, uh, realizing that even their existing policies related to ai, uh, and analytics, uh, has, has holds some gaps in it, right? So the starting point is to say, what is the policy that's needed, not just to keep up with the regulations, uh, which have also been changing and evolving, but also to meet up with the internal standards in terms of risk and controls and security, uh, in order to not show up on the front page of Wall Street Journal, right? As, as a place where, uh, AI got misused or put put customers at harm.

Uh, moving on from technology, it is about building the right set of processes and controls, right? Which includes, uh, incorporating those controls and how the technology itself is managed and governed. Uh, this can be things like controls on data, how the traditional data related controls, which are often role-based and access-based controls, how do they translate into a world where large amounts of data is being, uh, ingested into large language models and running a whole bunch of applications, right?

So, so, so putting the right set of controls on it, both from a process standpoint as well as adoption of the right technology, uh, it, which is helping administer those controls to the data and to the AI assets that are getting built. So there's a ton of work going on in this space. Banks are doing their own work.

A lot of, uh, large tech providers, a lot of large platform providers are, are, are doing work in this space. Uh, so I think those are, those are the two, two main things, the policies as well as the technology and the adoption of controls. Can you provide any examples or use case scenarios of problems that maybe banks face or processes, um, that could be, uh, improved with AI tools?

Yeah, for sure. There are, there are so many examples. Uh, one of the, one of the things that we've seen, uh, with ai, uh, adoption, especially with generative AI and our AI, is that it is still a relatively new technology, right?

So there are concerns about, uh, how, how much of a risk it creates because of this. The use cases have been, uh, more what, what one would consider on the safer side of the spectrum, right? Internal, internal applications and use cases, things that are for the most part, uh, facing employees of a bank, of a financial institution as opposed to direct, direct to customer, even though the latter has started to come into play.

So, we see two or three main areas where, uh, there is a general trend around adoption. The first one is, uh, knowledge, uh, knowledge tools. Uh, so basically AI driven interfaces, which, uh, are between, uh, uh, a bank employee and a corpus of information that they need, whether it is policies and procedures.

Uh, so for example, uh, let's say there's a collections agent, uh, agent, right? How they're supposed to use the bank's policy around collections and use it for, uh, when they're, when they're handling, uh, a, a delinquent account or, or in, or in fraud and claims or disputes. So in general, the single most, most, uh, prominent use cases about these AI based interfaces, which move from keyword based approaches to, uh, to tools that are using generative AI to, uh, be able to better handle, uh, uh, natural language search and present, uh, the bank's employee with a more, uh, precise and contextually relevant information such that they can, uh, do their jobs better.

There's so many others, I could keep talking about those, but, uh, we'll, we'll turn back to you. So, as fast as AI is advancing, when you look to the future, say, you know, five years from now, uh, what does the banking industry look like as far as AI use? So it's probably gonna look a lot different than it does today.

Uh, one big trend that we are seeing is the, uh, application of agent ai and that technology is maturing, which is really building upon what, what has happened with generative ai, right? So ability to be able to, uh, interpret and, uh, un un unstructured content, right? Such as, uh, a query or, or a human, uh, you know, response, uh, a conversation, uh, a question that is posed, uh, you know, in, in a natural language, if you see the, the trend as to where that's going to go, a few things are likely to happen, right?

One is, uh, just a large scale usage of agent based systems for handling majority of a, of, of a bank's processes and operations. And this can be customer aligned onboarding, it can be risk assessment and risk management, it can be service operations. It can be, uh, many other things that are today handled by, either by humans or combination of humans and some rules-based systems, right?

So there'll be some large, uh, large scale automation of that. Second thing, uh, would be, uh, sort of a democratization of how these, uh, tools, uh, get built, right? So today, uh, it, there are often, there's a demarcation between a role between a business stakeholder and somebody who works in technology or somebody who works in, uh, on the data side, right?

So let's say there's a, there's a finance function or a CFO or a, or a controller might be working with a data analyst to, to run queries for them. Those types of roles will transform as, uh, the, the, the, the, the business stakeholders will be able to much more easily interact with the data, right? And, and, and have, almost have copilot or personas that are helping them with, with their business decisioning and day-to-day functions along the way, right?

Uh, large scale, uh, processes that are mostly employee driven and supported by decisions, uh, uh, based systems will likely get inverted where they're mostly, uh, fully autonomous or semi-autonomous agent based systems where humans involvement is really for the purpose of, uh, exception handling and for the purpose of, of having some sort of oversight, right? So today it is humans doing work with some tools based, uh, oversight, uh, or, or tool supported oversight. It may get totally flipped into, uh, you know, armies of agents doing work, uh, with, with some human oversight on top of that, right?

And lastly, I would say there's, uh, the expectations of customers and clients, uh, would, would change. And this is not a trend that's starting today and will change in five years. This is a trend which has been there for a long time in terms of expectations of a superior customer experience, right?

The days of like waiting on a call to, you know, to, to, to, to speak with an agent, those are going to significantly change. Most people don't want to speak with an agent or have a human interaction in the servicing context. And as these sorts of AI-based systems get better in two ways, one, in being able to better understand the nuance and context right, that causes each of us to, you know, navigate past the IVR and speak with a human, those systems will get more sophisticated understanding what we truly need, and they get more powerful in terms of fulfilling that need, right?

So the need for getting to a human is going to reduce, which is gonna fundamentally transform how, uh, how customers, uh, interact with financial institutions just in the way as, uh, the usage of internet and usage of, you know, banking acts has changed how we interact with institutions today in pretty much, uh, everyone or at least 90% of the way, uh, of, of our day-to-day needs, uh, needs and actions. Alright. Well, if there was one key takeaway you could leave our audience today with, what would that be?

Yeah. So hard to say one, but let me, let me, let me try to, uh, cover it in three, right? So one is expect a ton of change in terms of, uh, customer experience as well as employee experience.

Uh, expect a fundamental shift in terms of the transformation of underlying business processes and the application of technology and ai. Uh, everything in the past is probably 10% of what's happening, uh, what's about to happen in the, in, in, in, in the next few years. And lastly, I would say that the effort required to do this goes a lot beyond choosing the right technology and working on upgrading and the, the technology and having it, you know, future ready and futureproof.

It goes into just as much effort, if not more, and nuanced and complex effort in building the controls and, and establishing the control and putting the right governance on it, and in thinking about the processes and the, and the people aspects of that adoption. Okay. Well, thank you so much for sharing your insights with us today.

Thank you. Thank you again for having me. And thank you to our audience.

Stay tuned. There's more.