The foundation of financial AI: data, governance, and trust- transcript
Kate Eby: Hello, and welcome to the Industry Forward podcast, where we explore key trends, transformative technologies, and real-world innovations that are reshaping fields from aerospace, industrial machinery, and semiconductors to pharmaceuticals and beyond. I’m Kate Eby, and I’ll be your host for today’s episode. Today, we’re taking a look at the banking, financial services, and insurance industry and examining its AI readiness. Joining me today is Dylan Tancill, Vice President of Sales, Global Head of Banking, Financial Services, and Insurance at Siemens Digital Industries Software. Thanks for joining me today, Dylan. Can you briefly introduce your background and experience in the BFSI industry?
Dylan Tancill: Yes, Kate, thanks for having me. I’m happy to be here. Yeah, my background starts in the financial service industry itself. I spent about the first decade of my career as a retail broker at Fidelity Investments. I’ve spent the last 12 years as a vendor to the industry, first at DataWatch, which was acquired by Altair, which has now been acquired by Siemens. So about 20 years of experience working either for or with the industry.
Kate Eby: That’s great. Before we jump in, I have to ask you a question that I’ve started asking all of our guests. And that is, what is your favorite fiction reference? A lot of people building toward the future have a favorite sci-fi reference. I’m guessing from your perspective, maybe it’s a little more banking related, but what would you say that is? And is it inspirational or cautionary?
Dylan Tancill: Oh, interesting. That’s a good question. Favorite fictional reference. I mean, if I wanted to keep it within the industry itself, it’s a movie based on a true story, but I think my favorite financial service piece of art is the movie The Big Short, because I worked through that time in the industry, and I think they did a really good job of capturing what had happened. If I wanted to go sci-fi, it’s probably a little bit ominous, and I don’t mean it to be that way. I almost mean it to be a little bit fun. But with where we’ve taken the business into AI, I do think of Space Odyssey and Space Odyssey 2000 and Hal A lot.
Kate Eby: I like it. I actually love both of those movies, especially The Big Short. All right, let’s go ahead and dive right in. How would you describe the current state of technology adoption in banking and insurance, especially since you just referenced Space Odyssey?
Dylan Tancill: Yeah, it’s a great place to start. I’d say one of the things that’s interesting and challenging about operating in this industry is banking, financial services, and insurance is both advanced and constrained. And I think that comes from the fact that they’ve often been early adopters of new technology, whether it was modeling and applying stats and math to business in the early days before they called it analytics, or getting into analytics and advanced analytics. And now with AI, that gives them a foundation and an experience to say, we’re going to try new things before other people. but it also means that they have massive legacy estates that can present challenges. A lot of what we work with our customers and partners on today is the concept of getting their data AI ready. Sometimes before they can get into full out agentic AI, they have to go through modernization efforts to make sure that their data is clean and curated and connected with context. They’re not struggling because they lack innovation, but they carry a lot of complexity, both operationally and in terms of regulations.
And I think that’s part of what Siemens brings to the table is that Siemens has 100 plus years of experience operating in some of the most highly regulated industries in the world. While BFSI is a slightly different business model, the idea that we have to be very conscious of these regulatory environments is nothing new for us.
Kate Eby: I want to get a little more into AI readiness, but before I do, let’s just take a step back. What would you say are some of the biggest technology challenges BFSI organizations are dealing with today?
Dylan Tancill: Well, one of the ways that we could frame it is I think that there’s a massive focus lately on total cost of ownership. Obviously, there’s plenty of press out there about the increasing costs of AI. And I think that the need to get into AI and get AI from pilots into production and actually drive value with AI to stay competitive is really making people take a look at very costly legacy systems. We have a lot of engagement around how can we bring down our total cost of ownership And it’s looking at infrastructure, storage, compute, but also software.
And this is where I find the broad portfolio of technology that we have at Siemens Software to be very interesting to banks and insurance companies, because they can look at one vendor who, through economies of scale and also just good business practices, can help them reduce the software side as they optimize their environments. The other big one that’s nothing new, but I think it’s ever-changing, and some of the fear around AI is only going to increase it, it’s regulatory compliance.
There’s certain banks out there, some of the ones that would be featured in the movie we heard talking about earlier, who make a lot of their decisions around what’s come at them lately from regulators. We used to say, and I still think it’s true to this day, that banks buy software from a vendor for three reasons. To make money, save money, or stay out of jail. And that last one of staying out of jail is tied a lot to regulatory compliance and what they’re hearing from their examiners. I think the total cost of ownership and optimizing that and regulatory compliance are still two of the biggest things, and AI’s only pushing that further. And then I think AI-driven innovation is a big one we hear.
And it’s not always because things can’t be done without AI, but there’s a lot of scrutiny lately from the market. For any publicly traded institution on what their CEOs and CFOs are talking about in earnings calls, Are they mentioning AI, agentic AI, improvements in automation through AI? So it’s a challenge whether people think it’s right or wrong. Just to meet the expectation of investors today, they’ve got to be finding ways to drive more value with AI.
Kate Eby: As you were talking, it sounds like there are some aspects of AI, whether agentic or otherwise, that are already being used in the BFSI industry. But you also talk a lot about this idea of AI readiness. I’d like to just delve into what exactly AI readiness means in the current and in the future.
Dylan Tancill: Yeah, absolutely. I mean, one of the most common topics you hear around AI and agentic AI is the risk of hallucinations, the risk of getting something wrong. For a long, time, we’ve worked with banks on traditional machine learning or data science. Using techniques like logistic regression or decision tree to predict something or to make a decision. Traditional machine learning and data science is largely deterministic. It’s going to give you the same answer every time with the same inputs. A lot of the fear around AI is that AI is probabilistic and it’s a people pleaser. And if you ask any of the LLMs the same question twice, you’re going to get at least somewhat different, if not very different answers. One way to control that and to limit the amount of variation in answers, or even to train AI to say, I don’t know, because it’s largely been trained to never say that. It will come up with some answer because, again, it’s a people pleaser. One of the ways to do that is by limiting the data available to AI to only curated data sets.
A lot of the work that we’re doing is building the foundation for that AI-ready data. There’s lots of different ways you could go about it. The way that we at Siemens go about it is through knowledge graph technology. Knowledge graph is really good at not only integrating data from all different data sources, but also doing it with context because it uses ontologies and semantic layers. And so, while a lot of different tools or programming languages could help you join two different data sources together, the join itself is just using one column that has a unique identifier to bring two different data sets together. It doesn’t tell each data set what the other columns are that have been brought together and how they relate to things. Knowledge graph gives context to the data by explaining the relationship. It also acts as a control mechanism.
If something’s not connected to the graph and you’re deploying AI on the knowledge graph, it can’t see those unconnected items. It doesn’t have the entire internet, the entire enterprise worth of data available to scramble around from hop to hop and maybe come up with something that is a hallucination or just an inaccurate answer. you can use the knowledge graph to bring the data together with context so that an AI agent can reason better and benefit from that context and the relationships, but also limit it to what you want it to be able to see. We hear the phrase very often from our customers of the first step for them is to have data quality and master data management in place, and then come to us with the, what we call curated data, and only the curated data do we connect through the knowledge graph. And then we deploy AI agents on the graph, and they’re limited to these nice clean data sources that have been brought together with context and relationship.
Kate Eby: It really is a spectrum when it comes to AI within the industry. It’s, you know, using AI now, but preparing to use AI more and more advantageously in the future and get correct results. Is that a fair assessment?
Dylan Tancill: Yeah, absolutely. I mean, a lot of what we work with at the banks, especially the large ones, is helping you get ready for AI and then putting guardrails on AI. Because the reality is that most large financial institutions have so many talented programmers, they don’t struggle to write machine learning models or AI models or AI agents. a lot of the big banks had a framework for AI agents before software vendors had caught up. But what they are lacking is getting from the pilot phase of playing around with AI into production. And it’s largely that they don’t have a problem building models. They have a problem trusting models and getting models into production. And so that’s where we look to add value because I’m not going to tell a bank that is essentially a technology company themselves with the number of programmers and other talented technologists that they employ that they don’t know how to do AI. What I’d like to tell them is, here’s what we’ve learned from working with over 800 financial institutions, all 10 of the top 10 global banks, seven of the top 10 global insurers, 15 of the top 20 financial service firms in the world. It’s that there’s things that have to be done foundationally and then in production to put guardrails in order to have successful deployments of AI.
Kate Eby: One of the things that I found interesting is we recently did a study with an analyst firm on where people really were when it came to industrial AI. So not just what people are using every day, not the ChatGPTs and Copilots that we use on our desktop, but getting to that trusted AI contextualized data aspect that you’ve been talking about. I’m curious from your perspective where the BFSI industry is in terms of that spectrum as compared to other industries?
Dylan Tancill: And are you talking about in terms of like where it’s successfully being applied?
Kate Eby: I would say not so much where it’s being successfully applied, although absolutely kind of that could illustrate where I’m going with this. It’s more about, if you look across industries from industrial machinery, from life sciences and pharmaceuticals, aerospace and defense, for example, where does BFSI kind of fall in terms of where it’s at when it comes to using AI?
Dylan Tancill: I would think that they would rank near the top of industries. Not that, you know, there’s not plenty of talented people working on AI in some of the other places that you mentioned, but they’ve just traditionally been such a early adopter of these new technologies. And I also think that they sort of lead from the front in terms of putting pressure on each other and on other industries to adopt AI. I also think that they’re in a more service-oriented industry, where there’s not as many physical things. It’s more about insights and analytics and predictions and things of that nature. And AI really lends itself to their work in a very obvious way.
Kate Eby: Well, and I wonder if you look across some of the industries that I mentioned or used for reference, I feel like you have kind of that spectrum of the people who are way out on the forefront and would actually say that they’re using all the industrial AI capability that’s available and setting the stage for what’s coming next. And then you have people on the other end of the spectrum, whether it’s due to legacy systems or because they’re a really small shop or what have you, that are on, you know, they’re kind of in a wait and see mode. It sounds like with the BFSI industry, they really can’t afford to stay competitive without being on the forefront.
Dylan Tancill: Yes, I’ve experienced this firsthand. Yeah, I would think probably the only industry that’s out in front of BFSI in terms of adoption of AI and also the pressure to adopt AI is probably the technology sector themselves. And that makes sense. But then right behind the technology, I would probably say BFSI is going to be #2. And I think a large part of it, or what I’ve experienced firsthand, is this focus on what financial institutions are talking about AI publicly, whether it’s at a trade show or the classic example I’ve heard from multiple people is earnings calls. We work with all of the big global payment service companies. And we had an inbound call from one of our longtime contacts within the past year who said, I need to identify a use case in the CFO’s office that we can call AI. And he very plainly said that we can call AI. He said, I don’t care if we have to use some traditional machine learning techniques and automation techniques. We need to be able to brand this as AI.
And as we got involved in this project and dug into their motivations, it was that some analyst had published a paper or report looking at the CFOs of all the major payment providers and which ones were talking about AI versus which ones were not. And their particular CFO had zero mentions of public references to AI at this point. And so there is this pressure because the market, obviously the entire market, not just the FSI industry, but the entire market right now has had huge gains in terms of the stock indexes and whatnot related to the promise of AI.
If you believe in that statement, and then you see that one of the players in a particular segment of the industry like global payments, is not talking about AI at all while their competitors are, it stands to reason that their competitors might get an edge on them. I think that there’s a long history with new technology of these institutions putting pressure on themselves and of investors assuming that they will continue to be early adopters. If they’re not talking about it early on, it will cause problems.
Kate Eby: With that in mind, with them being on the forefront, how do you see AI reshaping banking and insurance operations as we move forward at this current rate of innovation?
Dylan Tancill: Yeah, some of the really cool initial wins that I’ve seen come around two different areas, really. You have financial crimes like fraud and insider trading and the things that they need to do to stay out of jail. And then the, I’d call it generally like research. We’ve got some firms that have used pieces of our technology as part of a solution to enable people to do research on the fly that previously would have taken hours to accumulate. It’s a little bit in the vein of the chatbot, which I think is sort of the cliche, low-hanging fruit.
But it’s impressive in the sense that the underlying data that’s being brought together is lots of unstructured and semi-structured data that was previously pretty hard to get at. And we’ve been doing a lot of that with our knowledge graph technology from the way that we use Elasticsearch. We can bring unstructured, semi-structured, and structured data together to enable people to query across vastly different data sources throughout the enterprise.
On the financial crime side, there’s a lot of good examples of deploying AI agents that will at least flag things that they think are potential bad behavior, intentionally or unintentionally. We have a knowledge graph deployed with a large US hedge fund where we brought together all of their communications data and all of their trading data. And the trading data is structured, but it’s high volume, tick database type of data that’s streaming all day.
And then the communications data is largely unstructured, recorded phone calls, emails, chats, and those things typically live in different systems. And we’re able to bring them together, again, with the context and relationship that makes an AI agent better able to reason across the data, but also the agents are just working 24 hours a day. They’re spotting stuff in real time. And then we’ve built a custom application layer on top of that allows the human in the loop at the hedge fund to quickly take action on those insights. And so, yeah, I think that broadly using terms here, research and financial crimes are two places that I’ve seen customers of ours having wins with AI.
Kate Eby: Thanks for joining us on this episode of Industry Forward. As we’ve heard, success with AI isn’t just about deploying the latest models or experimenting with new technologies. It’s about building a trusted foundation of high-quality, contextualized data, establishing the right governance, and creating the guardrails needed to move AI from pilot projects into production.
We also explored why BFSI remains one of the leading industries in AI adoption, driven by competitive pressures and regulatory demands. From combating financial crime to transforming research and decision-making, AI is already reshaping how financial institutions operate.
Dylan, thank you for sharing your expertise and for helping us better understand the opportunities and challenges ahead.
And thank you to our listeners for tuning in. Be sure to subscribe to Industry Forward for more conversations with industry leaders, innovators, and technology experts who are helping shape the future of business and technology.
Until next time, I’m Kate Eby. Thanks for listening.
Siemens Digital Industries Software helps organizations of all sizes digitally transform using software, hardware and services from the Siemens Xcelerator business platform. Siemens’ software and the comprehensive digital twin enable companies to optimize their design, engineering and manufacturing processes to turn today’s ideas into the sustainable products of the future. From chips to entire systems, from product to process, across all industries. Siemens Digital Industries Software – Accelerating transformation.
