What BFSI must do now to prepare for AI
Kate Eby: Hello, and welcome to the Industry Forward Podcast, where we explore the trends, technologies, and innovations transforming industries around the world. I’m your host, Kate Eby.
In today’s episode, we’re continuing our conversation with Dylan Tancill, Vice President of Sales and Global Head of Banking, Financial Services and Insurance at Siemens Digital Industries Software.
As financial institutions race to adopt AI, the conversation is shifting from experimentation to execution. Success is no longer just about deploying new technologies. It’s also about building trust and creating the data foundations that enable AI to deliver real business value.
We’ll explore how banks and insurers are navigating increasingly complex regulatory environments and what industry leaders should be doing today to prepare for the future of AI-driven financial services. We’ll also discuss the skills the next generation of professionals will need to succeed in a rapidly evolving and highly regulated industry.
Dylan, you talk a lot about the three reasons that banks buy software, one being to make money, one to save money, and the third big one to stay out of jail. I’d like to dive a little bit more into the compliance and regulatory pressures that are on the BFSI industry. Can you talk a little bit about what considerations BFSI has to take to remain compliant with regulators, especially in this current age of AI?
Dylan Tancill: Yes. I mean, what it comes down to for financial institutions is that being mostly right is not good enough. You have to be 100% accurate. And that’s where we talk about what needs to be done in a deterministic fashion versus what can be done in a probabilistic fashion using AI. Traditionally, when it comes to making decisions around lending and things of that nature, we’ve partnered with financial institutions to use deterministic models like traditional machine learning models to make those decisions because they have to get it 100% right. There are things that we’re seeing the ability for AI to move faster, find more insights, things of that nature. But if it has to be 100% right for the regulators or for the health of their business, that’s where it’s very important to keep a human in the loop.
So with everything that we do, we like to say that we treat regulation as a design principle and build from regulation out, rather than saying, we want to do something really cool with technology, with AI, and we’re going to treat regulation as an afterthought. And at the end of the process, we’re going to go, oh, does this comply with all the regulations? No, we take the exact opposite approach. We don’t view it as a constraint. We view our understanding of the regulations in this market very much as a strength. But everything needs to be explainable, auditable, have human oversight, have governance and guardrails, because one of the easiest factors to change the profitability of a bank or the tenure of a CEO is getting a big fine from a regulator for doing something wrong.
Kate Eby: It seems we talk a lot about digitalization and the digital twin. I’m going to start with, is the digital twin being used in BFSI and how is it useful?
Dylan Tancill: It’s funny, going back five years ago or so, when I was at Altair, we would publish lots of material around digital twin, because Altair played in a lot of the industrial industries that Siemens does as well. And I would be offered the opportunity with marketing to partner on digital twin stories for BFSI. What is exciting, and this is all because of AI, is we are starting to see RFPs be written in banking for true digital win. And the example that I’ve seen a few times now is for data center.
There are a lot of banks looking at the costs around AI, banks and insurance companies, that are looking at what should they own rather than lease from a data center vendor. And as they look at this, they’re also saying, well, we want to have complete understanding of our data center or of a potential data center. And they’re getting into simulating data centers at banks. And it’s really exciting to me because, I’ve been part of a simulation company, at least a company that has simulation solutions for the better part of a decade now, but we now have people looking at simulating everything that goes into a data center so that they can understand all of their costs related to electrification and cooling, et cetera, so they can make decisions on what they bring in-house versus what they do through a third party. And I think that’s a big change from my experience, and it’s all related to AI and the compute required for AI and the costs that come with that.
Kate Eby: You talked a lot about data centers, and I’m curious, based on your history and the fact that digital twin was kind of a story that had to be stretched to really be told in the BFSI industry in the past. Is that also true when it comes to a digital thread? And if not, what would a digital thread look like in a non-physical industry?
Dylan Tancill: Yeah, we’ve had interesting debates among leadership at Siemens here because I think traditionally at Siemens, a digital thread has a very engineering specific definition. And there’s even been debates of should we come up with a different term for non-industrial like BFSI. But I think that the concept does apply. And in examples like the one I just gave you of a data center digital twin, we’re using multiple parts of the portfolio from simulation and visualization and AI to build that solution.
I do think it looks a lot like other digital twins in the sense that we’re using multiple solutions from Siemens to build a broader solution for a customer. But if you think about just at a high level, how manufacturers need to connect their requirements, the engineering, the production, and the operations, financial institutions, for much of what they do, they need to connect data, decisions, models, processes, and outcomes. And we’ve had a long history of playing across that entire spectrum with the solutions that we bring to market in data and AI.
Kate Eby: With that in mind, what would a potential digital thread look like? Obviously, you’re connecting all of that data, right? If you think about the digital thread from that perspective and the different aspects and areas that you need to connect together, it feels like even if it’s not physical, there’s still that opportunity for a digital thread.
Dylan Tancill: Yeah, absolutely. I mean, the one that I’ve been talking about when I’ve been out on the road at trade shows and visiting customers is this concept of having an enterprise trust layer. Because as you try to deploy AI for either speed or additional insights or anything with traditional processes that financial institutions have, like onboarding, lending, underwriting, fraud, and other financial crime related things, you are touching multiple areas of the business and multiple personas.
A lot of times these decisions get made in pockets, and then you’re stitching together solutions that were decided on in silos and that don’t necessarily speak well to each other. But we, across our technology stack, have this very unique ability to create an enterprise trust layer and serve multiple personas. and be one vendor to hold responsible for that whole process. So we do data integration through knowledge graph technology that’s usually going to be very IT-heavy persona. And we also have a framework for building and deploying AI agents and putting guardrails around the AI agents. And that would traditionally live with somebody with more of a data science background.
And then we have the ability to build custom application layers for the human in the loop to take their part in the process. And we can tailor each one to a specific use case and really improve the user interface and the user experience for what’s traditionally an analyst in some line of business who has to apply their domain expertise. I view this idea of an enterprise trust layer as something that we do that looks and feels like a digital thread in the sense that we’re bringing data together that’s going to touch multiple personas in multiple different ways to ultimately deliver a solution that hopefully means better, faster outcomes for our customers.
Kate Eby: Along those lines, in diving a little bit deeper, you talked earlier about simulation, especially as it relates to data centers, but how would you compare simulation to, or is there value in bringing simulation to some of the actuary models, for example, and the language of SaaS, which for our listeners who like me might be thinking software as a service, that actually has a different meaning when you get into the industry of BFSI?
Dylan Tancill: Yes. What you’ve described of testing AI agents before deploying or stress testing the customer journey or doing fraud and risk scenario modeling, that’s what we used to try to position as a potential digital twin. And I think it’s something that existed well before the term digital twin became popular. So that is somewhere that I was not introducing any new technology. I was simply branding something that we were already doing with our data analytics tooling as a digital twin.
But I do think that we’re seeing, largely because of cloud compute costs, we’re seeing people start to evaluate what they will bring on-premise versus what they will continue to lease from a data center or have in the cloud. And it’s introducing opportunities to apply Siemens’ actual simulation software to different processes and two different physical environments that we traditionally were not servicing in the past.
Kate Eby:
I’ll throw out the question. I feel like even the digital twin itself has really evolved in what that means, right? When we started talking about the digital twin years or even decades ago, something, it was a bit different than when we talk about the comprehensive digital twin or where it’s headed, right?
Dylan Tancill:
Yes. Yeah. I mean, financial services has been using elements of digital twins for decades, but we just didn’t call them digital twins. We’d call them stress testing, scenario analysis, simulation modeling is a term that I’d heard in the past. But I think a lot of people would think, well, that’s like kind of what if analysis. Simulation is usually involving a physical thing. And we are starting to now see opportunities to simulate physical environments for banks, which I think is really exciting.
Kate Eby: That’s great. I know we talked about this a little bit, what the future looks like, but it sounds like there are just so many moving parts with AI readiness, agentic AI, things that are going on now, things that are going on in the future. If you’re looking three to five years down the road, what is the BFSI landscape going to look like compared to where it is today?
Dylan Tancill: Well, I’ve got to be honest with you, three to five years seems the possibilities are endless with how much things are changing month to month here in the current landscape. But I do think that the financial institutions that win are going to be the ones that get the most real-world gains from AI. And those are going to be the institutions that are spending the time now to create these enterprise trust foundations, the AI ready data. I think there’s a lot of pressure on all types of institutions, but especially BFSI, to start driving real value out of AI and to get from the pilot phase into the production phase. And from the customers I have that are most successful, it’s those who are first getting their data in order.
The other thing is, you mentioned the language of SaaS earlier. So as you said, that’s S-A-S as opposed to S-A-A-S, software as a service. A lot of the modernization efforts that we’re involved in and total cost of ownership opportunities are related to things like the language of SaaS. SaaS is not the only example, but it’s a situation where the costs have increased greatly over the years. They’re not the only way to accomplish what that language does. You’ve got a lot of really good tooling out there to do low-code, no-code, as well as much more current and popular languages in open source like R and Python. And you have a talent problem. I’m old enough that when I went to college for economics, I used the language of SAS, but I think my brothers who are four and five years younger than me didn’t use it by the time they got there.
You’re going to have a waning pool of talent to support it, an increasing cost, and it’s not aligned with a lot of these current AI initiatives. So that cost becomes increasingly difficult to justify. So what we’ve seen a lot of people doing is taking legacy technology debt, like an estate for the language of SaaS, and modernizing it and reducing costs. And in that whole process, they free up budgeting that can be used for open source cloud and AI initiatives.
Kate Eby: Speaking of kind of the transition, not just in the industry, but obviously the education path that, or the skills that are required to succeed in the industry, what advice would you give someone looking to go into this industry who’s maybe just starting out in their college education or even just considering it from a high school perspective.
Dylan Tancill: I would say that’s another one where I, not to try to be too funny, but as I mentioned before, I’d repeat the comment of it is difficult to anticipate. If somebody’s a freshman in college today, what skills will they need when they graduate four years later? Because if you think about it, we were saying a couple of years ago that one of the most important things that you could learn was about prompt engineering. And now with the advancements in a lot of these LLMs, it’s become a lot less important. I think one thing that will remain true is banking and insurance is… almost akin to the technology industry. I’ve been out of the industry for 12 or 13 years now, but even over a decade ago when I was at Fidelity Investments, leadership would say all the time, we’re not a financial service company, we’re a technology company.
We spend X number of billions of dollars a year on technology, and we employ all of these talented programmers, et cetera, et cetera. I think a piece of advice that will not become outdated is to develop a strong understanding of data, traditional data science, AI, business processes and workflows, and then also learn how highly regulated organizations operate. That’s one of the biggest things that I think people sometimes struggle with or are surprised about when they work with a bank.
And I see that not just from peers that I had when I was in the industry, but I see it with our customers and with our employees. Because for a lot of what we do for financial institutions, we’re onboarding forward-deployed engineers from Siemens to go work at the bank. And they will end up with a at abcbank.com e-mail, and they’ll be logging into a VDI inside that bank’s environment. And as soon as they do that. They are now working for a bank or insurance company in a very highly regulated space.
We’ve had some people who said, wow, there’s a lot of requirements that they have to fill that are beyond what you would think of the normal scope of the project. We have a two-year project going on at a tier one bank right now, and they very often get pulled off of the actual work itself. to go do training that everybody at the bank has to take inside a certain timeline to remain compliant. I think if you’re in university or college and thinking about a career in financial services, I would get a good foundation for math, for data, but also a good understanding of how highly regulated industries work.
Kate Eby: So establish a good foundation, but stay curious and never stop learning.
Dylan Tancill: Absolutely. I mean, that’s the reason I’m giving you a somewhat generic answer is that I think these are things that will hold true no matter what comes next, but definitely stay curious and continue to learn because the space is changing all the time. And I think that there’s a lot of value that our customers see in bringing in top young talent because We’ve got people like myself here who worked in the industry 10, 15 years ago. I’ve hired quite a few people from the industry over time.
After a period of time, my experience is really a familiarity with the basics of the industry and some credibility when I go talk to leaders at financial institutions that I have done the job, I have lived in their world. But it’s the younger people we’ve hired away from banks and insurance companies that can really add the most benefit to the real work we do with AI, with knowledge graphs, with other pieces of our technology stack. I’ve joked before, my experience from 12 years ago doesn’t involve anything related to AI or knowledge graph, et cetera, et cetera. So I think the same way that I’ve really valued hiring young folks from the BFSI industry, they see the same value in hiring top talent because these people are going to come in creative and thinking outside of the box and able to help them make an impact in this landscape that’s just constantly changing.
Kate Eby: Other than seeking out the best talent coming out of college and universities today, what is the one thing that BFSI leaders should really be focusing on today to prepare for the future?
Dylan Tancill: Yeah, I think that, you know, in terms of successful implementations of the technologies that we serve the industry with, I think that you want to keep things simple. You don’t want to over-engineer things. And it really is about picking the right use case, especially as we talk a lot about AI. I think trying to take on big, lofty ideas is more risky in terms of failure or not being able to get out of the pilot phase because you can’t make it enterprise grade. You can’t make it regulatory, trustworthy. I think that the most successful customers I have doing real stuff with AI, not just traditional data science or automation and calling it AI, their success has largely been because they have picked the right use case that is large enough to be valuable, but small enough to wrap your head around and execute on in a reasonable timeline where you’re not going to get shut down because it’s months and months or years and years later and you haven’t delivered any product. So I think keeping things simple and not over-engineering is one of the traits for leaders in this industry that I see making the most gains with data and AI.
Kate Eby: I always find saying over-engineering to be a very polarizing phrase when we work with a lot of engineers, especially as someone who’s not got an engineering background, but it’s a favorite.
Dylan Tancill: That’s probably good advice. I borrowed that from one of the leaders at a financial institution that we work with. They’ve been particularly successful with our knowledge graph technology. And knowledge graph is, we see, like the foundation for getting AI-ready data, not only to make sure it’s all clean data, but because of the way it works with context. As I go to different trade shows, we go to the Gartner Data Analytics Summit, and last week I was at the Snowflake Summit. You hear vendors across this landscape talking about context, and Knowledge Graph does that really well. But Knowledge Graph’s also a place that can get very academic, almost religious.
There’s some deep technical pieces of it around the flavors of graph like RDF versus LPG and the semantic standards. And he says all the time that the people who are struggling with this are spending all their time trying to engineer the perfect ontology. And he tells a story about for the first success story we had together, they drew it on a whiteboard in a day. And they didn’t deploy it in production right away, of course. They deployed it in a lower environment and kept their existing solution running in production. But he said within three or four months, they had a graph up and running with AI agents deployed on it, improving the time to market and the number of insights they were able to pick up. And he said, yeah, I’ve worked with other people where the ontologist takes nine months to design the ontology.
And in the meantime, the leadership that’s supporting the project gets bored, finds another initiative, something else comes up or they just get frustrated and the initiative gets shut down. He’s big on pick the right use case and then don’t over-engineer the project. It’s better to get something up and running and then continue to test and iterate than to sit in the back room trying to come up with the perfect model for months and months.
Kate Eby: I’ve got to say, the word ontology finally made an appearance. And just for our listeners who maybe aren’t as familiar with that term, can you clarify what that is?
Dylan Tancill: Yeah, an ontology is an important part of a knowledge graph. It’s essentially a model of how things are related to one another in a domain. So we built ontologies for predicting fraud, for predicting insider trading. If you think about a bank, like an ontology might say, this customer owns this account, this account belongs to this branch, this loan is issued to this customer. A lot of this information would live in different systems. They might have a CRM and an ERP and lending platform, a core banking platform, et cetera.
And as I mentioned at one point earlier, you can do traditional joins across systems with lots of tooling, We offer tooling that you can do it with. You can do it with lots of programming languages too. You could write SQL or Python or R. But what the ontology does beyond just bringing different data together is explaining those relationships from items that live in different systems. So it’s not the data itself. It’s like a blueprint for the entities and relationships and defining that.
Kate Eby: It’s not just connecting data, it’s providing context for that data.
Dylan Tancill: Yes, yeah. So if you kind of break down into simple sources, an Excel spreadsheet would be data and a knowledge graph would be connected data. But the ontology, it’s like the dictionary or the rule book that explains what everything means from those connected sources.
Kate Eby: That is a great explanation. And I have to tell you, I joked about the word ontology finally making an appearance, but I have to say, Maybe this reflects poorly on me, but 20 years with Siemens and until about six months ago, I’d never heard that word used regularly. And now I hear it all the time. We were actually at a recent event and I joked about how many times the word ontology was used in mainstage presentations. And maybe that’s because I’ve been more focused on discrete industries as opposed to things like process and BFSI, where there’s less of a product and more of, well, process. But I find it interesting to just see that becoming such a much more regularly used term across our organization.
Dylan Tancill: Yes, it is, because it’s not a new term, but I think that AI and AI agents have really given knowledge graph technology, which includes semantic layers and ontologies, a new lease on life here. The technology’s been around since at least the 90s, but the way that an AI agent can benefit from the context that you get from an ontology is really improving their ability to reason, especially across multiple hops or traversals, as we call them. So I know it’s the foundation of our newly launched Intelligence Center X. As I mentioned, when I go to trade shows, everybody’s got the words context or semantic or ontology on their booths nowadays. So it’s a knowledge graph and the old semantic web technology is really having a fresh life here in the age of AI and AI agents.
Kate Eby: Is it fair to say then, if you’re kind of looking across the board, that ontologies are to the more process-related industries, almost the same as what a digital twin has historically been for some of the more discrete industries?
Dylan Tancill: That’s an interesting way of putting it. In terms of what I think of, and this may just be my experience from years past, but the BFSI concept of digital twin were really things like scenario analysis, stress testing, what if analysis, running Monte Carlo simulations, things like that, which don’t require knowledge graph technology. But I think that the knowledge graph technology is really becoming this like enterprise trust layer to combat some of the fears that people have about how they, you know, the threats, the perceived threats of AI and AI agents.
Kate Eby: Great. Dylan, unfortunately, we are running out of time for today’s discussion, but I do want to pause and see if you have any closing thoughts.
Dylan Tancill: Yeah, well, first, Kate, I would like to say I appreciate you having me. Really enjoyed the conversation. And yeah, I would say in terms of closing thoughts, you know, BFSI is undergoing one of the most significant technology shifts in its history. It’s a really exciting time to be a part of it. And Siemens is investing in BFSI because We believe that financial institutions will play a major role in shaping the next generation of digital transformation. We’ve got people, the domain expertise, and the technology to help BFSI really improve outcomes for themselves and their customers.
Kate Eby: That’s great, and I think a great way to end. Thank you again for joining us in this discussion, and a thank you to our listeners. We appreciate your time and hope that you enjoyed this episode of the Industry Forward Podcast. If you like what you hear, please follow the Industry Forward Podcast on your favorite podcast streaming platform. And we hope to see you on the next episode.
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.
