Data trust in the AI era: Building customer confidence through responsible banking

Speaker A: Hello and welcome back to fstech's podcast series in partnership with HCL Tech. I'm Jonathan Easton, editor at uh, fstech, and if you caught episode one, you know we're diving deep into the transformative potential of generative AI in financial services. In our last episode, we covered what it takes to get started with Genai in banking. And today in episode two, we're shifting gears to explore something even more foundational. Data trust, transparency and the responsible use of AI.

Speaker B: Data is now one of the most

Speaker A: powerful tools at a bank's disposal, but it's also one of its biggest risks. Customers are more informed and more cautious than ever before. They want personalization, yes, but not at the expense of privacy. So how can financial institutions turn data trust into a strategic advantage in the age of AI? And what does responsible AI implementation actually look like in the real world? To help us break it down, I'm joined again by Sudip Lahiri, Executive Vice President and head of Financial Services for

Speaker B: Europe and the UK and Ireland at HCL Tech. Thank you very much for being here. Again, following, uh, our first episode, which was really fascinating conversation all about the foundational elements of, um, building Gen AI in financial services. And um, one area that we talked, uh, about was the concept of data trust. And that's what we're going to be

Speaker A: looking at more in depth on today. Um, so from your perspective, how has

Speaker B: the concept of data trust evolved in financial services over the past few years? And what are some of the most significant challenges banks face today in establishing trust with customers?

Speaker C: Well, Jonathan, first of all, thank you for having me again in your second episode. I really enjoyed the conversation the first episode. It was quite, uh, interesting and exciting conversation. Uh, thank you for that. Uh, well, I think the question which you asked is a very, uh, fundamental question which banks are asking. And this is, uh, in my view, uh, the most important topic which needs to be addressed. So I think, um, in one hand, the organizations need to look at thinking, uh, data as a strategic asset, um, make sure that architectural perspective, it's set up in a proper way, it's managed, maintained in a proper way. There's a proper governance and discipline and adoption of best practices, uh, so that on a sustainable basis, uh, it is treated and managed as a strategic asset. Uh, on the other hand, uh, organizations need to invest, uh, to make it more robust in terms of security. Uh, and I think the CISO organization needs to play a very important role, uh, for the security part of it. The next part is as the organizations start adopting Large language models. We um, see that um, there is a need to train the model. Um, so you have the requirement of training the model and testing the model. So you need to make sure you have the right data sets uh, available to train and test. Because if you don't do that you have a risk of bias and hallucination, uh, which we see quite often. And that puts a question mark on the credibility of the output uh, it generates. So I think organizations need to make sure that um, uh there's a proper framework architecture available for training the model and testing the model, uh, and also have capabilities like explainability, observability, uh, those capabilities are in place.

Speaker B: You mentioned there about explainable AI. Now, um, how do you see explainable AI contributing to building uh, trust and customer trust, especially when third party providers implement AI systems?

Speaker C: Yeah, I think I will um, maybe give an example here. Uh, I think one of the large banks, European banks, we have uh, been engaged with them, um, you know, um, for their trade surveillance process where you have a lot of alerts, um and the analysts used to handle so many alerts and with the Genai you have the luxury of um, you know, making sure that um, the right alerts are chosen and the uh, analysts are far more productive um and far more efficient. Um and um, um we have successfully rolled out that program. I think it has delivered significant benefit. Uh but having said that you have to have sort of explainability uh, that uh, uh you know some of the false positives uh, which have been eliminated, uh, they are the false positives in a true sense. Um so at HCL Tech we have created a robust framework for explainable AI. Um and it's a combination of uh, tools and process. So whatever we try to do, we try to sort of bring people, process and technology together. Um so our explainable AI framework addressed that and with that we not only made sure that uh, the alerts are optimized, the analysts productivity is significantly improved, the quality of output is significantly improved but also the results are explainable. Uh, I think that's very important. Um, in the first phase, of course in the phase one you do want to have a bit of human uh, intervention, human checks and once you match the whole system and process uh, the models are properly trained, uh the data, with the right data sets I think you can move into a phase of complete uh, elimination. Um and then you have the output where you are not only able to explain but also you are able to sort of deliver the right output to the business.

Speaker B: So I just want to go at this point back to something we talked about in the first episode, which is about the governance. Um, now what are some more governance frameworks do you recommend for managing uh, Data Trust in this AI era? And what are some of the best practices for implementing responsible AI? And that's one of the big talking points at the moment.

Speaker C: At ah, the end of the day you have to have a sort of a center of excellence sort of capability. You have to create, uh, because you are talking about newer capability and um, newer uh, technologies, uh, best practices, um, and also element of continuous improvement because you are in a journey so you have to keep learning uh, on things. So you have to have that center of excellence mindset. The question is whether you want to do it centrally or you want to do it decentrally. Uh, so if you do it centrally, uh, you kind of create um, a central governance, but then you keep the business units bit far from what you want to do. Um, and if you want to make it a truly democratizing AI in sort of a way where everybody has sort of a liberty and freedom to do their own things, uh, you end up reinventing the will and maybe create uh, sort of architecture which could be more vulnerable for cyber threats. So you have to have a balance. So from AI governance point of view, I think the first I would say to create that framework, uh, of how the center of excellence would look like, what functions it would perform, what capabilities it should have. And then second element is how uh, the business units would be onboarded and what sort of do's and don'ts they would have, uh, to sort of leverage that capability to address their business problems or create newer solutions, newer opportunities for their business. Um, from a technology point of view also, uh, there are nuances because you know, we'll see that organizations will adopt uh, different large language LLMs, so you have different models. Um, and also there will be a phase where old and new would coexist. So which means that you need to trash, uh, between old and new. Um, and also you need to have, when there are more models involved, you need to have sort of a registry. So there is a technical element of AI governance. Um, uh, I think that that's something which we have a framework for that. Um, and I think that is in my view one of the critical success factors for AI adoption. Uh, and also to make sure that uh, when you adopt the AI, ah, you just don't um, increase your risk for possible uh, cyber threats.

Speaker B: One thing we also talked about in that first episode is the cultural piece, um, when it comes to this sort of thing. It's as much about minimizing friction between different areas of the business and not just creating this kind of centralized hub of AI activity and things. And, and everyone has to absolutely kind of should be a um, listening, you know, relationship between different areas.

Speaker C: Absolutely. I think, you know, this is what, this is why the democratizing AI for everyone is so important. Because if the capability is confined to only limited set of people, then you have those frictions. I think you need to create a sense of uh, ownership, uh, and a sense of uh, sort of people should feel that they have the access. Everyone should have the access to this capability. Uh, if one section in an organization, they have the access, others don't, then you have those problems and it can be a massive problem. So uh, I think democratization, uh, of A.I. uh, uh, ensuring that uh, A.I. for everyone. These are the foundational principles which the organizations need to adopt as part of their adoption journey.

Speaker B: And in line with all of that, the minimizing friction is so important because friction equals inefficiency.

Speaker C: Absolutely.

Speaker B: And might be one thing to have a bit of tension. If you're in a band, Fleetwood Mac, you know, you get rumors uh, from having uh, a bit of ah, friction going on. But that doesn't make for a healthy business. Now what are some steps that um, leaders in financial institutions need uh, to take to effectively, efficiently uh, implement AI?

Speaker C: I think the you know, um, defining from a leadership point of view, defining a vision and mission is very important. Having um, an effective communication strategy is very important. Um, having some of the foundational elements uh, in place is super important because you know, a small mistake, a small misstep can cause bigger issues. Um, you know, I think some of, uh. And that's why some of the, you know, explainability and these are so important. Um, also organizations need to keep in mind that uh, it's not a big bang. So there will be a phase where old and new would coexist and the organizations have to navigate and triage uh, between the two. So and at the same time, uh, deliver value uh, to the business. Um, I think one of the topic you spoke about, you asked is making sure that the employees, they don't feel um, that they are left behind. Um, and that is super important. Uh, AI for everyone is a very important concept. Uh, I think these are, I would say the principle, uh, the key principles for the organizations to keep in mind and make sure they are implemented in a religious manner for AI journey to be successful.

Speaker B: So looking ahead, what are some of the emerging Technologies or approaches that you, that you believe will be the most transformative for establishing data trust between uh banks, their partners and customers in global financial institutions.

Speaker C: I think going forward um we will see a uh move towards more agent AI um because you know especially in financial service still there are a lot of um, you know uh, handoffs and lot of manual processes, law of documentation of steps. Some of them are based on regulatory requirement, uh some are based on legacy and the problem which uh, which organizations have to to work with the legacy. Um I think we'll see more and more adoption of agentic AI uh so the agents are basically ah sort of a system which uh can act like human and it can take and take a task and can perform the task um in an effective efficient manner uh like a human would do. And also it learns from its action. So it's sort of uh evolution of the technology and we'll see and we are not, it's not something years ahead of us. I think the way the whole technology is developing, um, adoption is progressing. I think it's a matter of couple of years. We'll see agents uh doing many work which humans do as of today and

Speaker B: it's kind of the next step towards that ultimate goal of uh, artificial general intelligence.

Speaker C: Absolutely.

Speaker A: Yeah, absolutely.

Speaker B: Um, so you know I think this has been a really really fascinating conversation once again. Um, any last thoughts from you in terms of the future of data trust in banking within this entire generative AI conversation?

Speaker C: I think Jonathan this is a uh very very exciting topic. We at HCL Tech coming from a very strong engineering heritage, this is super exciting for us and um, uh it also provides so many opportunities uh to create superior business value for our customers. You know that um, we are living in an era of uh experience um the customers that stick to a company not uh, necessarily because of how robust the product is but the experience it gets from a particular organization. I think creating m a wow experience uh for the customers is a great opportunity for banks. At the same time, ah, looking at some of the traditional business processes um and making sure they are more efficient, um uh significantly reduce technical debt. Uh I think these are all foundational and fundamental uh opportunities for organizations to modernize at the same time deliver superior customer experience. So this technology is very exciting. It has a huge potential. Uh at the same time I do want to caution that there are risks um and organizations need to look at uh uh those risks, especially cyber uh risks and have the proper system and process and capability in place to address.

Speaker B: Well thank you uh again very much uh, Sudeep, for being us here. Fascinating, uh, conversation. Uh, sure, we could talk about it much longer. Um, but I think that's a perfect way of wrapping this up. So thank you again, Sudeep. And thank you for watching in our

Speaker A: next episode in this series, we'll be looking at how to scale Gen AI across the Enterprise.

Speaker B: Uh, but until then, take care.